Radiomic features and multilayer perceptron network classifier: a robust MRI classification strategy for distinguishing glioblastoma from primary central nervous system lymphoma

PET radiomics in lung cancer: advances and translational challenges

Radiomics is an emerging field of medical imaging that aims at improving the accuracy of diagnosis, prognosis, treatment planning and monitoring non-invasively through the automated or semi-automated quantitative analysis of high-dimensional image features. Specifically in the field of nuclear medicine, radiomics utilizes imaging methods such as positron emission tomography (PET) and single photon emission computed tomography (SPECT) to evaluate biomarkers related to metabolism, blood flow, cellular activity and some biological pathways. Lung cancer ranks among the leading causes of cancer-related deaths globally, and radiomics analysis has shown great potential in guiding individualized therapy, assessing treatment response, and predicting clinical outcomes. In this review, we summarize the current state-of-the-art radiomics progress in lung cancer, highlighting the potential benefits and existing limitations of this approach. The radiomics workflow was introduced first including image acquisition, segmentation, feature extraction, and model building. Then the published literatures were described about radiomics-based prediction models for lung cancer diagnosis, differentiation, prognosis and efficacy evaluation. Finally, we discuss current challenges and provide insights into future directions and potential opportunities for integrating radiomics into routine clinical practice.

Deep Learning Radiomics Model of Contrast-Enhanced CT for Differentiating the Primary Source of Liver Metastases

RATIONALE AND OBJECTIVES

To develop and validate a deep learning radiomics (DLR) model based on contrast-enhanced computed tomography (CT) to identify the primary source of liver metastases.

MATERIALS AND METHODS

In total, 657 liver metastatic lesions, including breast cancer (BC), lung cancer (LC), colorectal cancer (CRC), gastric cancer (GC), and pancreatic cancer (PC), from 428 patients were collected at three clinical centers from January 2018 to October 2023 series. The lesions were randomly assigned to the training and validation sets in a 7:3 ratio. An additional 112 lesions from 61 patients at another clinical center served as an external test set. A DLR model based on contrast-enhanced CT of the liver was developed to distinguish the five pathological types of liver metastases. Stepwise classification was performed to improve the classification efficiency of the model. Lesions were first classified as digestive tract cancer (DTC) and non-digestive tract cancer (non-DTC). DTCs were divided into CRC, GC, and PC and non-DTCs were divided into LC and BC. To verify the feasibility of the DLR model, we trained classical machine learning (ML) models as comparison models. Model performance was evaluated using accuracy (ACC) and area under the receiver operating characteristic curve (AUC).

RESULTS

The classification model constructed by the DLR algorithm showed excellent performance in the classification task compared to ML models. Among the five categories task, highest ACC and average AUC were achieved at 0.563 and 0.796 in the validation set, respectively. In the DTC and non-DTC and the LC and BC classification tasks, AUC was achieved at 0.907 and 0.809 and ACC was achieved at 0.843 and 0.772, respectively. In the CRC, GC, and PC classification task, ACC and average AUC were the highest, at 0.714 and 0.811, respectively.

CONCLUSION

The DLR model is an effective method for identifying the primary source of liver metastases.

Radiomics and Clinical Features for Distinguishing Kidney Stone-Associated Urinary Tract Infection: A Comprehensive Analysis of Machine Learning Classification

Background

This study investigated the abilities of radiomics and clinical feature models to distinguish kidney stone-associated urinary tract infections (KS-UTIs) using computed tomography.

Methods

A retrospective analysis was conducted on a single-center dataset comprising computed tomography (CT) scans and corresponding clinical information from 461 patients with kidney stones. Radiomics features were extracted from CT images and underwent dimensionality reduction and selection. Multiple machine learning (Three types of shallow learning and four types of deep learning) algorithms were employed to construct radiomics and clinical models in this study. Performance evaluation and optimal model selection were done using receiver operating characteristic (ROC) curve analysis and Delong test. Univariate and multivariate logistic regression analyzed clinical and radiomics features to identify significant variables and develop a clinical model. A combined model integrating radiomics and clinical features was established. Model performance was assessed by ROC curve analysis, clinical utility was evaluated through decision curve analysis, and the accuracy of the model was analyzed via calibration curve.

Results

Multilayer perceptron (MLP) showed higher classification accuracy than other classifiers (area under the curve (AUC) for radiomics model: train 0.96, test 0.94; AUC for clinical model: train 0.95, test 0.91. The combined radiomics-clinical model performed best (AUC for combined model: train 0.98, test 0.95). Decision curve and calibration curve analyses confirmed the model's clinical efficacy and calibration.

Conclusions

This study showed the effectiveness of combining radiomics and clinical features from CT scans to identify KS-UTIs. A combined model using MLP exhibited strong classification abilities.

Beyond hand-crafted features for pretherapeutic molecular status identification of pediatric low-grade gliomas

The use of targeted agents in the treatment of pediatric low-grade gliomas (pLGGs) relies on the determination of molecular status. It has been shown that genetic alterations in pLGG can be identified non-invasively using MRI-based radiomic features or convolutional neural networks (CNNs). We aimed to build and assess a combined radiomics and CNN non-invasive pLGG molecular status identification model. This retrospective study used the tumor regions, manually segmented from T2-FLAIR MR images, of 336 patients treated for pLGG between 1999 and 2018. We designed a CNN and Random Forest radiomics model, along with a model relying on a combination of CNN and radiomic features, to predict the genetic status of pLGG. Additionally, we investigated whether CNNs could predict radiomic feature values from MR images. The combined model (mean AUC: 0.824) outperformed the radiomics model (0.802) and CNN (0.764). The differences in model performance were statistically significant (p-values < 0.05). The CNN was able to learn predictive radiomic features such as surface-to-volume ratio (average correlation: 0.864), and difference matrix dependence non-uniformity normalized (0.924) well but was unable to learn others such as run-length matrix variance (- 0.017) and non-uniformity normalized (- 0.042). Our results show that a model relying on both CNN and radiomic-based features performs better than either approach separately in differentiating the genetic status of pLGGs, and that CNNs are unable to express all handcrafted features.

How does deep learning/machine learning perform in comparison to radiologists in distinguishing glioblastomas (or grade IV astrocytomas) from primary CNS lymphomas?: a meta-analysis and systematic review

BACKGROUND

Several studies have been published comparing deep learning (DL)/machine learning (ML) to radiologists in differentiating PCNSLs from GBMs with equivocal results. We aimed to perform this meta-analysis to evaluate the diagnostic accuracy of ML/DL versus radiologists in classifying PCNSL versus GBM using MRI.

METHODOLOGY

The study was performed in accordance with PRISMA guidelines. Data was extracted and interpreted by two researchers with 12 and 23 years' experience, respectively, and QUADAS-2 tool was used for quality and risk-bias assessment. We constructed contingency tables to derive sensitivity, specificity accuracy, summary receiver operating characteristic (SROC) curve, and the area under the curve (AUC).

RESULTS

Our search identified 11 studies, of which 8 satisfied our inclusion criteria and restricted the analysis in each study to reporting the model showing highest accuracy, with a total sample size of 1159 patients. The random effects model showed a pooled sensitivity of 0.89 [95% CI:0.84-0.92] for ML and 0.82 [95% CI:0.76-0.87] for radiologists. Pooled specificity was 0.88 [95% CI: 0.84-0.91] for ML and 0.90 [95% CI: 0.81-0.95] for radiologists. Pooled accuracy was 0.88 [95% CI: 0.86-0.90] for ML and 0.86 [95% CI: 0.78-0.91] for radiologists. Pooled AUC of ML was 0.94 [95% CI:0.92-0.96]and for radiologists, it was 0.90 [95% CI: 0.84-0.93].

CONCLUSIONS

MRI-based ML/DL techniques can complement radiologists to improve the accuracy of classifying GBMs from PCNSL, possibly reduce the need for a biopsy, and avoid any unwanted neurosurgical resection of a PCNSL.

Improving the Generalizability and Performance of an Ultrasound Deep Learning Model Using Limited Multicenter Data for Lung Sliding Artifact Identification

Deep learning (DL) models for medical image classification frequently struggle to generalize to data from outside institutions. Additional clinical data are also rarely collected to comprehensively assess and understand model performance amongst subgroups. Following the development of a single-center model to identify the lung sliding artifact on lung ultrasound (LUS), we pursued a validation strategy using external LUS data. As annotated LUS data are relatively scarce-compared to other medical imaging data-we adopted a novel technique to optimize the use of limited external data to improve model generalizability. Externally acquired LUS data from three tertiary care centers, totaling 641 clips from 238 patients, were used to assess the baseline generalizability of our lung sliding model. We then employed our novel Threshold-Aware Accumulative Fine-Tuning (TAAFT) method to fine-tune the baseline model and determine the minimum amount of data required to achieve predefined performance goals. A subgroup analysis was also performed and Grad-CAM++ explanations were examined. The final model was fine-tuned on one-third of the external dataset to achieve 0.917 sensitivity, 0.817 specificity, and 0.920 area under the receiver operator characteristic curve (AUC) on the external validation dataset, exceeding our predefined performance goals. Subgroup analyses identified LUS characteristics that most greatly challenged the model's performance. Grad-CAM++ saliency maps highlighted clinically relevant regions on M-mode images. We report a multicenter study that exploits limited available external data to improve the generalizability and performance of our lung sliding model while identifying poorly performing subgroups to inform future iterative improvements. This approach may contribute to efficiencies for DL researchers working with smaller quantities of external validation data.

Exploring the Potential of Machine Learning Algorithms to Improve Diffusion Nuclear Magnetic Resonance Imaging Models Analysis

Purpose

This paper explores different machine learning (ML) algorithms for analyzing diffusion nuclear magnetic resonance imaging (dMRI) models when analytical fitting shows restrictions. It reviews various ML techniques for dMRI analysis and evaluates their performance on different b-values range datasets, comparing them with analytical methods.

Materials and Methods

After standard fitting for reference, four sets of diffusion-weighted nuclear magnetic resonance images were used to train/test various ML algorithms for prediction of diffusion coefficient (D), pseudo-diffusion coefficient (D*), perfusion fraction (f), and kurtosis (K). ML classification algorithms, including extra-tree classifier (ETC), logistic regression, C-support vector, extra-gradient boost, and multilayer perceptron (MLP), were used to determine the existence of diffusion parameters (D, D*, f, and K) within single voxels. Regression algorithms, including linear regression, polynomial regression, ridge, lasso, random forest (RF), elastic-net, and support-vector machines, were used to estimate the value of the diffusion parameters. Performance was evaluated using accuracy (ACC), area under the curve (AUC) tests, and cross-validation root mean square error (RMSECV). Computational timing was also assessed.

Results

ETC and MLP were the best classifiers, with 94.1% and 91.7%, respectively, for the ACC test and 98.7% and 96.3% for the AUC test. For parameter estimation, RF algorithm yielded the most accurate results The RMSECV percentages were: 8.39% for D, 3.57% for D*, 4.52% for f, and 3.53% for K. After the training phase, the ML methods demonstrated a substantial decrease in computational time, being approximately 232 times faster than the conventional methods.

Conclusions

The findings suggest that ML algorithms can enhance the efficiency of dMRI model analysis and offer new perspectives on the microstructural and functional organization of biological tissues. This paper also discusses the limitations and future directions of ML-based dMRI analysis.

Radiomics for differentiation of gliomas from primary central nervous system lymphomas: a systematic review and meta-analysis

Background and objective

Numerous radiomics-based models have been proposed to discriminate between central nervous system (CNS) gliomas and primary central nervous system lymphomas (PCNSLs). Given the heterogeneity of the existing models, we aimed to define their overall performance and identify the most critical variables to pilot future algorithms.

Methods

A systematic review of the literature and a meta-analysis were conducted, encompassing 12 studies and a total of 1779 patients, focusing on radiomics to differentiate gliomas from PCNSLs. A comprehensive literature search was performed through PubMed, Ovid MEDLINE, Ovid EMBASE, Web of Science, and Scopus databases. Overall sensitivity (SEN) and specificity (SPE) were estimated. Event rates were pooled using a random-effects meta-analysis, and the heterogeneity was assessed using the χ2 test.

Results

The overall SEN and SPE for differentiation between CNS gliomas and PCNSLs were 88% (95% CI = 0.83 - 0.91) and 87% (95% CI = 0.83 - 0.91), respectively. The best-performing features were the ones extracted from the Gray Level Run Length Matrix (GLRLM; ACC 97%), followed by those obtained from the Neighboring Gray Tone Difference Matrix (NGTDM; ACC 93%), and shape-based features (ACC 91%). The 18F-FDG-PET/CT was the best-performing imaging modality (ACC 97%), followed by the MRI CE-T1W (ACC 87% - 95%). Most studies applied a cross-validation analysis (92%).

Conclusion

The current SEN and SPE of radiomics to discriminate CNS gliomas from PCNSLs are high, making radiomics a helpful method to differentiate these tumor types. The best-performing features are the GLRLM, NGTDM, and shape-based features. The 18F-FDG-PET/CT imaging modality is the best-performing, while the MRI CE-T1W is the most used.

Fully automated MR-based virtual biopsy of primary CNS lymphomas

Background

Primary central nervous system lymphomas (PCNSL) pose a challenge as they may mimic gliomas on magnetic resonance imaging (MRI) imaging, compelling precise differentiation for appropriate treatment. This study focuses on developing an automated MRI-based workflow to distinguish between PCNSL and gliomas.

Methods

MRI examinations of 240 therapy-naive patients (141 males and 99 females, mean age: 55.16 years) with cerebral gliomas and PCNSLs (216 gliomas and 24 PCNSLs), each comprising a non-contrast T1-weighted, fluid-attenuated inversion recovery (FLAIR), and contrast-enhanced T1-weighted sequence were included in the study. HD-GLIO, a pre-trained segmentation network, was used to generate segmentations automatically. To validate the segmentation efficiency, 237 manual segmentations were prepared (213 gliomas and 24 PCNSLs). Subsequently, radiomics features were extracted following feature selection and training of an XGBoost algorithm for classification.

Results

The segmentation models for gliomas and PCNSLs achieved a mean Sørensen-Dice coefficient of 0.82 and 0.80 for whole tumors, respectively. Three classification models were developed in this study to differentiate gliomas from PCNSLs. The first model differentiated PCNSLs from gliomas, with an area under the curve (AUC) of 0.99 (F1-score: 0.75). The second model discriminated between high-grade gliomas and PCNSLs with an AUC of 0.91 (F1-score: 0.6), and the third model differentiated between low-grade gliomas and PCNSLs with an AUC of 0.95 (F1-score: 0.89).

Conclusions

This study serves as a pilot investigation presenting an automated virtual biopsy workflow that distinguishes PCNSLs from cerebral gliomas. Prior to clinical use, it is necessary to validate the results in a prospective multicenter setting with a larger number of PCNSL patients.

A Contrast-Enhanced CT-Based Deep Learning System for Preoperative Prediction of Colorectal Cancer Staging and RAS Mutation

PURPOSE

This study aimed to build a deep learning system using enhanced computed tomography (CT) portal-phase images for predicting colorectal cancer patients' preoperative staging and RAS gene mutation status.

METHODS

The contrast-enhanced CT image dataset comprises the CT portal-phase images from a retrospective cohort of 231 colorectal cancer patients. The deep learning system was developed via migration learning for colorectal cancer detection, staging, and RAS gene mutation status prediction. This study used pre-trained Yolov7, vision transformer (VIT), swin transformer (SWT), EfficientNetV2, and ConvNeXt. 4620, and contrast-enhanced CT images and annotated tumor bounding boxes were included in the tumor identification and staging dataset. A total of 19,700 contrast-enhanced CT images comprise the RAS gene mutation status prediction dataset.

RESULTS

In the validation cohort, the Yolov7-based detection model detected and staged tumors with a mean accuracy precision (IoU = 0.5) (mAP_0.5) of 0.98. The area under the receiver operating characteristic curve (AUC) in the test set and validation set for the VIT-based prediction model in predicting the mutation status of the RAS genes was 0.9591 and 0.9554, respectively. The detection network and prediction network of the deep learning system demonstrated great performance in explaining contrast-enhanced CT images.

CONCLUSION

In this study, a deep learning system was created based on the foundation of contrast-enhanced CT portal-phase imaging to preoperatively predict the stage and RAS mutation status of colorectal cancer patients. This system will help clinicians choose the best treatment option to increase colorectal cancer patients' chances of survival and quality of life.

Biomaterial-based in vitro 3D modeling of glioblastoma multiforme

Adult-onset brain cancers, such as glioblastomas, are particularly lethal. People with glioblastoma multiforme (GBM) do not anticipate living for more than 15 months if there is no cure. The results of conventional treatments over the past 20 years have been underwhelming. Tumor aggressiveness, location, and lack of systemic therapies that can penetrate the blood-brain barrier are all contributing factors. For GBM treatments that appear promising in preclinical studies, there is a considerable rate of failure in phase I and II clinical trials. Unfortunately, access becomes impossible due to the intricate architecture of tumors. In vitro, bioengineered cancer models are currently being used by researchers to study disease development, test novel therapies, and advance specialized medications. Many different techniques for creating in vitro systems have arisen over the past few decades due to developments in cellular and tissue engineering. Later-stage research may yield better results if in vitro models that resemble brain tissue and the blood-brain barrier are used. With the use of 3D preclinical models made available by biomaterials, researchers have discovered that it is possible to overcome these limitations. Innovative in vitro models for the treatment of GBM are possible using biomaterials and novel drug carriers. This review discusses the benefits and drawbacks of 3D in vitro glioblastoma modeling systems.

Computed Tomography-Based Deep Learning Nomogram Can Accurately Predict Lymph Node Metastasis in Gastric Cancer

BACKGROUND

Computed tomography is the most commonly used imaging modality for preoperative assessment of lymph node status, but the reported accuracy is unsatisfactory.

AIMS

To evaluate and verify the predictive performance of computed tomography deep learning on the presurgical evaluation of lymph node metastasis in patients with gastric cancer.

METHODS

347 patients were retrospectively selected (training cohort: 242, test cohort: 105). The enhanced computed tomography arterial phase images of gastric cancer were used for lesion segmentation, radiomics and deep learning feature extraction. Three methods were used for feature selection. Support vector machine (SVM) or random forest (RF) was used to build models. The classification performance of the models was evaluated using the area under the receiver operating characteristic curve (AUC). We also established a nomogram that included clinical predictors.

RESULTS

The model based on ResNet50-RF showed favorable classification performance and was verified in the test cohort (AUC = 0.9803). The nomogram based on deep learning feature scores and the lymph node status reported by computed tomography showed excellent discrimination. AUC of 0.9978 was achieved in the training cohort and verified in the test cohort (AUC = 0.9914). Decision analysis curve showed the value of nomogram in clinical application.

CONCLUSION

The computed tomography-based deep learning nomogram can accurately and effectively evaluate lymph node metastasis in patients with gastric cancer before surgery.

Machine learning for predicting accuracy of lung and liver tumor motion tracking using radiomic features

Background

Internal tumor motion is commonly predicted using external respiratory signals. However, the internal/external correlation is complex and patient-specific. The purpose of this study was to develop various models based on the radiomic features of computed tomography (CT) images to predict the accuracy of tumor motion tracking using external surrogates and to find accurate and reliable tracking algorithms.

Methods

Images obtained from a total of 108 and 71 patients pathologically diagnosed with lung and liver cancers, respectively, were examined. Real-time position monitoring motion was fitted to tumor motion, and samples with fitting errors greater than 2 mm were considered positive. Radiomic features were extracted from internal target volumes of average intensity projections, and cross-validation least absolute shrinkage and selection operator (LassoCV) was used to conduct feature selection. Based on the radiomic features, a total of 26 separate models (13 for the lung and 13 for the liver) were trained and tested. Area under the receiver operating characteristic curve (AUC), sensitivity, and specificity were used to assess performance. Relative standard deviation was used to assess stability.

Results

Thirty-three and 22 radiomic features were selected for the lung and liver, respectively. For the lung, the AUC varied from 0.848 (decision tree) to 0.941 [support vector classifier (SVC), logistic regression]; sensitivity varied from 0.723 (extreme gradient boosting) to 0.848 [linear support vector classifier (linearSVC)]; specificity varied from 0.834 (gaussian naive bayes) to 0.936 [multilayer perceptron (MLP), wide and deep (W&D)]; and MLP and W&D had better performance and stability than the median. For the liver, the AUC varied from 0.677 [light gradient boosting machine (Light)] to 0.892 (logistic regression); sensitivity varied from 0.717 (W&D) to 0.862 (MLP); specificity varied from 0.566 (Light) to 0.829 (linearSVC); and logistic regression, MLP, and SVC had better performance and stability than the median.

Conclusions

Respiratory-sensitive radiomic features extracted from CT images of lung and liver tumors were proved to contain sufficient information to establish an external/internal motion relationship. We developed a rapid and accurate method based on radiomics to classify the accuracy of monitoring a patient's external surface for lung and liver tumor tracking. Several machine learning algorithms-in particular, MLP-demonstrated excellent classification performance and stability.

Radiomics-Based Machine Learning Model for Predicting Overall and Progression-Free Survival in Rare Cancer: A Case Study for Primary CNS Lymphoma Patients

Primary Central Nervous System Lymphoma (PCNSL) is an aggressive neoplasm with a poor prognosis. Although therapeutic progresses have significantly improved Overall Survival (OS), a number of patients do not respond to HD–MTX-based chemotherapy (15–25%) or experience relapse (25–50%) after an initial response. The reasons underlying this poor response to therapy are unknown. Thus, there is an urgent need to develop improved predictive models for PCNSL. In this study, we investigated whether radiomics features can improve outcome prediction in patients with PCNSL. A total of 80 patients diagnosed with PCNSL were enrolled. A patient sub-group, with complete Magnetic Resonance Imaging (MRI) series, were selected for the stratification analysis. Following radiomics feature extraction and selection, different Machine Learning (ML) models were tested for OS and Progression-free Survival (PFS) prediction. To assess the stability of the selected features, images from 23 patients scanned at three different time points were used to compute the Interclass Correlation Coefficient (ICC) and to evaluate the reproducibility of each feature for both original and normalized images. Features extracted from Z-score normalized images were significantly more stable than those extracted from non-normalized images with an improvement of about 38% on average (p-value < 10−12). The area under the ROC curve (AUC) showed that radiomics-based prediction overcame prediction based on current clinical prognostic factors with an improvement of 23% for OS and 50% for PFS, respectively. These results indicate that radiomics features extracted from normalized MR images can improve prognosis stratification of PCNSL patients and pave the way for further study on its potential role to drive treatment choice.

The application of machine learning to imaging in hematological oncology: A scoping review

Background

Here, we conducted a scoping review to (i) establish which machine learning (ML) methods have been applied to hematological malignancy imaging; (ii) establish how ML is being applied to hematological cancer radiology; and (iii) identify addressable research gaps.

Methods

The review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Extension for Scoping Reviews guidelines. The inclusion criteria were (i) pediatric and adult patients with suspected or confirmed hematological malignancy undergoing imaging (population); (ii) any study using ML techniques to derive models using radiological images to apply to the clinical management of these patients (concept); and (iii) original research articles conducted in any setting globally (context). Quality Assessment of Diagnostic Accuracy Studies 2 criteria were used to assess diagnostic and segmentation studies, while the Newcastle-Ottawa scale was used to assess the quality of observational studies.

Results

Of 53 eligible studies, 33 applied diverse ML techniques to diagnose hematological malignancies or to differentiate them from other diseases, especially discriminating gliomas from primary central nervous system lymphomas (n=18); 11 applied ML to segmentation tasks, while 9 applied ML to prognostication or predicting therapeutic responses, especially for diffuse large B-cell lymphoma. All studies reported discrimination statistics, but no study calculated calibration statistics. Every diagnostic/segmentation study had a high risk of bias due to their case-control design; many studies failed to provide adequate details of the reference standard; and only a few studies used independent validation.

Conclusion

To deliver validated ML-based models to radiologists managing hematological malignancies, future studies should (i) adhere to standardized, high-quality reporting guidelines such as the Checklist for Artificial Intelligence in Medical Imaging; (ii) validate models in independent cohorts; (ii) standardize volume segmentation methods for segmentation tasks; (iv) establish comprehensive prospective studies that include different tumor grades, comparisons with radiologists, optimal imaging modalities, sequences, and planes; (v) include side-by-side comparisons of different methods; and (vi) include low- and middle-income countries in multicentric studies to enhance generalizability and reduce inequity.

Progress of radiological‑pathological workflows in the differential diagnosis between primary central nervous system lymphoma and high‑grade glioma (Review)

Primary central nervous system lymphoma (PCNSL) and high‑grade glioma (HGG) are distinct entities of the CNS with completely distinct treatments. The treatment of PCNSL is chemotherapy‑based, while surgery is the first choice for HGG. However, the clinical features of the two entities often overlap, and a clear pathological diagnosis is important for subsequent management, especially for the management of PCNSL. Stereotactic biopsy is recognized as one of the minimally invasive alternatives for evaluating the involvement of the CNS. However, in the case of limited tissue materials, the differential diagnosis between the two entities is still difficult. In addition, some patients are too ill to tolerate a needle biopsy. Therefore, combining imaging, histopathology and laboratory examinations is essential in order to make a clear diagnosis as soon as possible. The present study reviews the progress of comparative research on both imaging and laboratory tests based on the pathophysiological changes of the two entities, and proposes an integrative and optimized diagnostic process, with the purpose of building a better understanding for neurologists, hematologists, radiologists and pathologists.

Accurate preoperative staging and HER2 status prediction of gastric cancer by the deep learning system based on enhanced computed tomography

Purpose

To construct the deep learning system (DLS) based on enhanced computed tomography (CT) images for preoperative prediction of staging and human epidermal growth factor receptor 2 (HER2) status in gastric cancer patients.

Methods

The raw enhanced CT image dataset consisted of CT images of 389 patients in the retrospective cohort, The Cancer Imaging Archive (TCIA) cohort, and the prospective cohort. DLS was developed by transfer learning for tumor detection, staging, and HER2 status prediction. The pre-trained Yolov5, EfficientNet, EfficientNetV2, Vision Transformer (VIT), and Swin Transformer (SWT) were studied. The tumor detection and staging dataset consisted of 4860 enhanced CT images and annotated tumor bounding boxes. The HER2 state prediction dataset consisted of 38900 enhanced CT images.

Results

The DetectionNet based on Yolov5 realized tumor detection and staging and achieved a mean Average Precision (IoU=0.5) (mAP_0.5) of 0.909 in the external validation cohort. The VIT-based PredictionNet performed optimally in HER2 status prediction with the area under the receiver operating characteristics curve (AUC) of 0.9721 and 0.9995 in the TCIA cohort and prospective cohort, respectively. DLS included DetectionNet and PredictionNet had shown excellent performance in CT image interpretation.

Conclusion

This study developed the enhanced CT-based DLS to preoperatively predict the stage and HER2 status of gastric cancer patients, which will help in choosing the appropriate treatment to improve the survival of gastric cancer patients.

Fully automated radiomics-based machine learning models for multiclass classification of single brain tumors: Glioblastoma, lymphoma, and metastasis

BACKGROUND AND PURPOSE

To investigate the diagnostic performance of fully automated radiomics-based models for multiclass classification of a single enhancing brain tumor among glioblastoma, central nervous system lymphoma, and metastasis.

MATERIALS AND METHODS

The training and test sets were comprised of 538 cases (300 glioblastomas, 73 lymphomas, and 165 metastases) and 169 cases (101 glioblastomas, 29 lymphomas, and 39 metastases), respectively. After fully automated segmentation, radiomic features were extracted. Three conventional machine learning classifiers, including least absolute shrinkage and selection operator (LASSO), adaptive boosting (Adaboost), and support vector machine with the linear kernel (SVC), combined with one of four feature selection methods, including forward sequential feature selection, F score, mutual information, and LASSO, were trained. Additionally, one ensemble classifier based on the three classifiers was used. The diagnostic performance of the optimized models was tested in the test set using the accuracy, F1-macro score, and the area under the receiver operating characteristic curve (AUCROC).

RESULTS

The best performance was achieved when the LASSO was used as a feature selection method. In the test set, the best performance was achieved by the ensemble classifier, showing an accuracy of 76.3% (95% CI, 70.0-82.7), a F1-macro score of 0.704, and an AUCROC of 0.878.

CONCLUSION

Our fully automated radiomics-based models for multiclass classification might be useful for differential diagnosis of a single enhancing brain tumor with a good diagnostic performance and generalizability.

Deep learning for rare disease: A scoping review

Although individually rare, collectively more than 7,000 rare diseases affect about 10% of patients. Each of the rare diseases impacts the quality of life for patients and their families, and incurs significant societal costs. The low prevalence of each rare disease causes formidable challenges in accurately diagnosing and caring for these patients and engaging participants in research to advance treatments. Deep learning has advanced many scientific fields and has been applied to many healthcare tasks. This study reviewed the current uses of deep learning to advance rare disease research. Among the 332 reviewed articles, we found that deep learning has been actively used for rare neoplastic diseases (250/332), followed by rare genetic diseases (170/332) and rare neurological diseases (127/332). Convolutional neural networks (307/332) were the most frequently used deep learning architecture, presumably because image data were the most commonly available data type in rare disease research. Diagnosis is the main focus of rare disease research using deep learning (263/332). We summarized the challenges and future research directions for leveraging deep learning to advance rare disease research.

A hierarchical machine learning model based on Glioblastoma patients' clinical, biomedical, and image data to analyze their treatment plans

AIM OF STUDY

Glioblastoma Multiforme (GBM) is an aggressive brain cancer in adults that kills most patients in the first year due to ineffective treatment. Different clinical, biomedical, and image data features are needed to analyze GBM, increasing complexities. Besides, they lead to weak performances for machine learning models due to ignoring physicians' knowledge. Therefore, this paper proposes a hierarchical model based on Fuzzy C-mean (FCM) clustering, Wrapper feature selection, and twelve classifiers to analyze treatment plans.

METHODOLOGY/APPROACH

The proposed method finds the effectiveness of previous and current treatment plans, hierarchically determining the best decision for future treatment plans for GBM patients using clinical data, biomedical data, and different image data. A case study is presented based on the Cancer Genome Atlas Glioblastoma Multiforme dataset to prove the effectiveness of the proposed model. This dataset is analyzed using data preprocessing, experts' knowledge, and a feature reduction method based on the Principal Component Analysis. Then, the FCM clustering method is utilized to reinforce classifier learning.

OUTCOMES OF STUDY

The proposed model finds the best combination of Wrapper feature selection and classifier for each cluster based on different measures, including accuracy, sensitivity, specificity, precision, F-score, and G-mean according to a hierarchical structure. It has the best performance among other reinforced classifiers. Besides, this model is compatible with real-world medical processes for GBM patients based on clinical, biomedical, and image data.

Improving radiomic model reliability using robust features from perturbations for head-and-neck carcinoma

Background

Using high robust radiomic features in modeling is recommended, yet its impact on radiomic model is unclear. This study evaluated the radiomic model’s robustness and generalizability after screening out low-robust features before radiomic modeling. The results were validated with four datasets and two clinically relevant tasks.

Materials and methods

A total of 1,419 head-and-neck cancer patients’ computed tomography images, gross tumor volume segmentation, and clinically relevant outcomes (distant metastasis and local-regional recurrence) were collected from four publicly available datasets. The perturbation method was implemented to simulate images, and the radiomic feature robustness was quantified using intra-class correlation of coefficient (ICC). Three radiomic models were built using all features (ICC &gt; 0), good-robust features (ICC &gt; 0.75), and excellent-robust features (ICC &gt; 0.95), respectively. A filter-based feature selection and Ridge classification method were used to construct the radiomic models. Model performance was assessed with both robustness and generalizability. The robustness of the model was evaluated by the ICC, and the generalizability of the model was quantified by the train-test difference of Area Under the Receiver Operating Characteristic Curve (AUC).

Results

The average model robustness ICC improved significantly from 0.65 to 0.78 (P&lt; 0.0001) using good-robust features and to 0.91 (P&lt; 0.0001) using excellent-robust features. Model generalizability also showed a substantial increase, as a closer gap between training and testing AUC was observed where the mean train-test AUC difference was reduced from 0.21 to 0.18 (P&lt; 0.001) in good-robust features and to 0.12 (P&lt; 0.0001) in excellent-robust features. Furthermore, good-robust features yielded the best average AUC in the unseen datasets of 0.58 (P&lt; 0.001) over four datasets and clinical outcomes.

Conclusions

Including robust only features in radiomic modeling significantly improves model robustness and generalizability in unseen datasets. Yet, the robustness of radiomic model has to be verified despite building with robust radiomic features, and tightly restricted feature robustness may prevent the optimal model performance in the unseen dataset as it may lower the discrimination power of the model.

Deep Convolutional Neural Network for Nasopharyngeal Carcinoma Discrimination on MRI by Comparison of Hierarchical and Simple Layered Convolutional Neural Networks

Nasopharyngeal carcinoma (NPC) is one of the most common head and neck cancers. Early diagnosis plays a critical role in the treatment of NPC. To aid diagnosis, deep learning methods can provide interpretable clues for identifying NPC from magnetic resonance images (MRI). To identify the optimal models, we compared the discrimination performance of hierarchical and simple layered convolutional neural networks (CNN). Retrospectively, we collected the MRI images of patients and manually built the tailored NPC image dataset. We examined the performance of the representative CNN models including shallow CNN, ResNet50, ResNet101, and EfficientNet-B7. By fine-tuning, shallow CNN, ResNet50, ResNet101, and EfficientNet-B7 achieved the precision of 72.2%, 94.4%, 92.6%, and 88.4%, displaying the superiority of deep hierarchical neural networks. Among the examined models, ResNet50 with pre-trained weights demonstrated the best classification performance over other types of CNN with accuracy, precision, and an F1-score of 0.93, 0.94, and 0.93, respectively. The fine-tuned ResNet50 achieved the highest prediction performance and can be used as a potential tool for aiding the diagnosis of NPC tumors.

Computed tomography-based deep-learning prediction of lymph node metastasis risk in locally advanced gastric cancer

Purpose

Preoperative evaluation of lymph node metastasis (LNM) is the basis of personalized treatment of locally advanced gastric cancer (LAGC). We aim to develop and evaluate CT-based model using deep learning features to preoperatively predict LNM in LAGC.

Methods

A combined size of 523 patients who had pathologically confirmed LAGC were retrospectively collected between August 2012 and July 2019 from our hospital. Five pre-trained convolutional neural networks were exploited to extract deep learning features from pretreatment CT images. And the support vector machine (SVM) was employed as the classifier. We assessed the performance using the area under the receiver operating characteristics curve (AUC) and selected an optimal model, which was compared with a radiomics model developed from the training cohort. A clinical model was built with clinical factors only for baseline comparison.

Results

The optimal model with features extracted from ResNet yielded better performance with AUC of 0.796 [95% confidence interval (95% CI), 0.715-0.865] and accuracy of 75.2% (95% CI, 67.2%-81.5%) in the testing cohort, compared with 0.704 (0.625-0.783) and 61.8% (54.5%-69.9%) for the radiomics model. The predictive performance of all the radiological models were significantly better than the clinical model.

Conclusion

The novel and noninvasive deep learning approach could provide efficient and accurate prediction of lymph node metastasis in LAGC, and benefit clinical decision making of therapeutic strategy.

Machine Learning and Deep Learning CT-Based Models for Predicting the Primary Central Nervous System Lymphoma and Glioma Types: A Multicenter Retrospective Study

Purpose and Background

Distinguishing primary central nervous system lymphoma (PCNSL) and glioma on computed tomography (CT) is an important task since treatment options differ vastly from the two diseases. This study aims to explore various machine learning and deep learning methods based on radiomic features extracted from CT scans and end-to-end convolutional neural network (CNN) model to predict PCNSL and glioma types and compare the performance of different models.

Methods

A total of 101 patients from five Chinese medical centers with pathologically confirmed PCNSL and glioma were analyzed retrospectively, including 50 PCNSL and 51 glioma. After manual segmentation of the region of interest (ROI) on CT scans, 293 radiomic features of each patient were extracted. The radiomic features were used as input, and then, we established six machine learning models and one deep learning model and three readers to identify the two types of tumors. We also established a 2D CNN model using raw CT scans as input. The area under the receiver operating characteristic curve (AUC) and accuracy (ACC) were used to evaluate different models.

Results

The cohort was split into a training (70, 70% patients) and validation cohort (31,30% patients) according to the stratified sampling strategy. Among all models, the MLP performed best, with an accuracy of 0.886 and 0.903, sensitivity of 0.914 and 0.867, specificity of 0.857 and 0.937, and AUC of 0.957 and 0.908 in the training and validation cohorts, respectively, which was significantly higher than the three primary physician's diagnoses (ACCs ranged from 0.710 to 0.742, p &lt; 0.001 for all) and comparable with the senior radiologist (ACC 0.839, p = 0.988). Among all the machine learning models, the AUC ranged from 0.605 to 0.821 in the validation cohort. The end-to-end CNN model achieved an AUC of 0.839 and an ACC of 0.840 in the validation cohort, which had no significant difference in accuracy compared to the MLP model (p = 0.472) and the senior radiologist (p = 0.470).

Conclusion

The established PCNSL and glioma prediction model based on deep neural network methods from CT scans or radiomic features are feasible and provided high performance, which shows the potential to assist clinical decision-making.

A Systematic Review of the Current Status and Quality of Radiomics for Glioma Differential Diagnosis

Simple Summary

Gliomas can be difficult to discern clinically and radiologically from other brain lesions (either neoplastic or non-neoplastic) since their clinical manifestations as well as preoperative imaging features often overlap and appear misleading. Radiomics could be extremely helpful for non-invasive glioma differential diagnosis (DDx). However, implementation in clinical practice is still distant and concerns have been raised regarding the methodological quality of radiomic studies. In this context, we aimed to summarize the current status and quality of radiomic studies concerning glioma DDx in a systematic review. In total, 42 studies were selected and examined in our work. Our study revealed that, despite promising and encouraging results, current studies on radiomics for glioma DDx still lack the quality required to allow its introduction into clinical practice. This work could provide new insights and help to reach a consensus on the use of the radiomic approach for glioma DDx.

Abstract

Radiomics is a promising tool that may increase the value of imaging in differential diagnosis (DDx) of glioma. However, implementation in clinical practice is still distant and concerns have been raised regarding the methodological quality of radiomic studies. Therefore, we aimed to systematically review the current status of radiomic studies concerning glioma DDx, also using the radiomics quality score (RQS) to assess the quality of the methodology used in each study. A systematic literature search was performed to identify original articles focused on the use of radiomics for glioma DDx from 2015. Methodological quality was assessed using the RQS tool. Spearman’s correlation (ρ) analysis was performed to explore whether RQS was correlated with journal metrics and the characteristics of the studies. Finally, 42 articles were selected for the systematic qualitative analysis. Selected articles were grouped and summarized in terms of those on DDx between glioma and primary central nervous system lymphoma, those aiming at differentiating glioma from brain metastases, and those based on DDx of glioma and other brain diseases. Median RQS was 8.71 out 36, with a mean RQS of all studies of 24.21%. Our study revealed that, despite promising and encouraging results, current studies on radiomics for glioma DDx still lack the quality required to allow its introduction into clinical practice. This work could provide new insights and help to reach a consensus on the use of the radiomic approach for glioma DDx.

External Validation of Deep Learning Algorithms for Radiologic Diagnosis: A Systematic Review

Purpose

To assess generalizability of published deep learning (DL) algorithms for radiologic diagnosis.

Materials and Methods

In this systematic review, the PubMed database was searched for peer-reviewed studies of DL algorithms for image-based radiologic diagnosis that included external validation, published from January 1, 2015, through April 1, 2021. Studies using nonimaging features or incorporating non-DL methods for feature extraction or classification were excluded. Two reviewers independently evaluated studies for inclusion, and any discrepancies were resolved by consensus. Internal and external performance measures and pertinent study characteristics were extracted, and relationships among these data were examined using nonparametric statistics.

Results

Eighty-three studies reporting 86 algorithms were included. The vast majority (70 of 86, 81%) reported at least some decrease in external performance compared with internal performance, with nearly half (42 of 86, 49%) reporting at least a modest decrease (≥0.05 on the unit scale) and nearly a quarter (21 of 86, 24%) reporting a substantial decrease (≥0.10 on the unit scale). No study characteristics were found to be associated with the difference between internal and external performance.

Conclusion

Among published external validation studies of DL algorithms for image-based radiologic diagnosis, the vast majority demonstrated diminished algorithm performance on the external dataset, with some reporting a substantial performance decrease.Keywords: Meta-Analysis, Computer Applications-Detection/Diagnosis, Neural Networks, Computer Applications-General (Informatics), Epidemiology, Technology Assessment, Diagnosis, Informatics Supplemental material is available for this article. © RSNA, 2022.

Deep learning radiomics based on contrast enhanced computed tomography predicts microvascular invasion and survival outcome in early stage hepatocellular carcinoma

OBJECTIVE

To evaluate the performance of a deep learning (DL)-based radiomics strategy on contrast-enhanced computed tomography (CT) to predict microvascular invasion (MVI) status and clinical outcomes, recurrence-free survival (RFS) and overall survival (OS) in patients with early stage hepatocellular carcinoma (HCC) receiving surgical resection.

METHODS

All 283 eligible patients were included retrospectively between January 2008 and December 2015, and assigned into the training cohort (n = 198) and the testing cohort (n = 85). We extracted radiomics features via handcrafted radiomics analysis manually and DL analysis of pretrained convolutional neural networks via transfer learning automatically. Support vector machine was adopted as the classifier. A clinical-radiological model for MVI status integrated significant clinical features and the radiological signature generated from the radiological model with the optimal area under the receiver operating characteristics curve (AUC) in the testing cohort. Otherwise, DL-based prognostic models were constructed in prediction of recurrence and mortality via Cox proportional hazard analysis.

RESULTS

The clinical-radiological model for MVI represented an AUC of 0.909, accuracy of 96.47%, sensitivity of 90.91%, specificity of 97.30%, positive predictive value of 83.33%, and negative predictive value of 98.63% in the testing cohort. The clinical-radiological models for identification of RFS and OS outperformed prediction performance of the clinical model or the DL signature alone. The DL-based integrated model for prognostication showed great predictive value with significant classification and discrimination abilities after validation.

CONCLUSIONS

The integrated DL-based radiomics models achieved accurate preoperative prediction of MVI status, and might facilitate predicting tumor recurrence and mortality in order to optimize clinical decisions for patients with early stage HCC.

Prediction of multiple pH compartments by deep learning in magnetic resonance spectroscopy with hyperpolarized 13C-labelled zymonic acid

BACKGROUND

Hyperpolarization enhances the sensitivity of nuclear magnetic resonance experiments by between four and five orders of magnitude. Several hyperpolarized sensor molecules have been introduced that enable high sensitivity detection of metabolism and physiological parameters. However, hyperpolarized magnetic resonance spectroscopy imaging (MRSI) often suffers from poor signal-to-noise ratio and spectral analysis is complicated by peak overlap. Here, we study measurements of extracellular pH (pHe) by hyperpolarized zymonic acid, where multiple pHe compartments, such as those observed in healthy kidney or other heterogeneous tissue, result in a cluster of spectrally overlapping peaks, which is hard to resolve with conventional spectroscopy analysis routines.

METHODS

We investigate whether deep learning methods can yield improved pHe prediction in hyperpolarized zymonic acid spectra of multiple pHe compartments compared to conventional line fitting. As hyperpolarized 13C-MRSI data sets are often small, a convolutional neural network (CNN) and a multilayer perceptron (MLP) were trained with either a synthetic or a mixed (synthetic and augmented) data set of acquisitions from the kidneys of healthy mice.

RESULTS

Comparing the networks' performances compartment-wise on a synthetic test data set and eight real kidney data shows superior performance of CNN compared to MLP and equal or superior performance compared to conventional line fitting. For correct prediction of real kidney pHe values, training with a mixed data set containing only 0.5% real data shows a large improvement compared to training with synthetic data only. Using a manual segmentation approach, pH maps of kidney compartments can be improved by neural network predictions for voxels including three pH compartments.

CONCLUSION

The results of this study indicate that CNNs offer a reliable, accurate, fast and non-interactive method for analysis of hyperpolarized 13C MRS and MRSI data, where low amounts of acquired data can be complemented to achieve suitable network training.

Potential of the Non-Contrast-Enhanced Chest CT Radiomics to Distinguish Molecular Subtypes of Breast Cancer: A Retrospective Study

Objectives

The molecular subtype plays an important role in breast cancer, which is the main reference to guide treatment and is closely related to prognosis. The objective of this study was to explore the potential of the non-contrast-enhanced chest CT-based radiomics to predict breast cancer molecular subtypes non-invasively.

Methods

A total of 300 breast cancer patients (153 luminal types and 147 non-luminal types) who underwent routine chest CT examination were included in the study, of which 220 cases belonged to the training set and 80 cases to the time-independent test set. Identification of the molecular subtypes is based on immunohistochemical staining of postoperative tissue samples. The region of interest (ROI) of breast masses was delineated on the continuous slices of CT images. Forty-two models to predict the luminal type of breast cancer were established by the combination of six feature screening methods and seven machine learning classifiers; 5-fold cross-validation (cv) was used for internal validation. Finally, the optimal model was selected for external validation on the independent test set. In addition, we also took advantage of SHapley Additive exPlanations (SHAP) values to make explanations of the machine learning model.

Results

During internal validation, the area under the curve (AUC) values for different models ranged from 0.599 to 0.842, and the accuracy ranged from 0.540 to 0.775. Eventually, the LASSO_SVM combination was selected as the final model, which included 9 radiomics features. The AUC, accuracy, sensitivity, and specificity of the model to distinguish luminal from the non-luminal type were 0.842 [95% CI: 0.728−0.957], 0.773, 0.818, and 0.773 in the training set and 0.757 [95% CI: 0.640–0.866], 0.713, 0.767, and 0.676 in the test set.

Conclusion

The radiomics based on chest CT may provide a new idea for the identification of breast cancer molecular subtypes.

Novel computer aided diagnostic models on multimodality medical images to differentiate well differentiated liposarcomas from lipomas approached by deep learning methods

BACKGROUND

Deep learning methods have great potential to predict tumor characterization, such as histological diagnosis and genetic aberration. The objective of this study was to evaluate and validate the predictive performance of multimodality imaging-derived models using computer-aided diagnostic (CAD) methods for prediction of MDM2 gene amplification to identify well-differentiated liposarcoma (WDLPS) and lipoma.

MATERIALS AND METHODS

All 127 patients from two institutions were included with 89 patients in one institution for model training and 38 patients in the other institution for external validation between January 2012 and December 2018. For each modality, handcrafted radiomics analysis with manual segmentation was applied to extract 851 features for each modality, and six pretrained convolutional neural networks (CNNs) extracted 512-2048 deep learning features automatically. Extracted imaging-based features were selected via univariate filter selection methods and the recursive feature elimination algorithm, which were then classified by support vector machine for model construction. Integrated with two significant clinical variables, age and LDH level, a clinical-radiological model was constructed for identification WDLPS and lipoma. All differentiation models were evaluated using the area under the receiver operating characteristics curve (AUC) and their 95% confidence interval (CI).

RESULTS

The multimodality model on deep learning features extracted from ResNet50 algorithm (RN-DL model) performed great differentiation performance with an AUC of 0.995 (95% CI 0.987-1.000) for the training cohort, and an AUC of 0.950 (95% CI 0.886-1.000), accuracy of 92.11%, sensitivity of 95.00% (95% CI 73.06-99.74%), specificity of 88.89% (95% CI 63.93-98.05%) in external validation. The integrated clinical-radiological model represented an AUC of 0.996 (95% CI 0.989-1.000) for the training cohort, and an AUC of 0.942 (95% CI 0.867-1.000), accuracy of 86.84%, sensitivity of 95.00% (95% CI 73.06-99.74%), and specificity of 77.78% (95% CI 51.92-92.63%) in external validation.

CONCLUSIONS

Imaging-based multimodality models represent effective discrimination abilities between WDLPS and lipoma via CAD methods, and might be a practicable approach in assistance of treatment decision.

Machine Learning in Differentiating Gliomas from Primary CNS Lymphomas: A Systematic Review, Reporting Quality, and Risk of Bias Assessment

BACKGROUND

Differentiating gliomas and primary CNS lymphoma represents a diagnostic challenge with important therapeutic ramifications. Biopsy is the preferred method of diagnosis, while MR imaging in conjunction with machine learning has shown promising results in differentiating these tumors.

PURPOSE

Our aim was to evaluate the quality of reporting and risk of bias, assess data bases with which the machine learning classification algorithms were developed, the algorithms themselves, and their performance.

DATA SOURCES

Ovid EMBASE, Ovid MEDLINE, Cochrane Central Register of Controlled Trials, and the Web of Science Core Collection were searched according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines.

STUDY SELECTION

From 11,727 studies, 23 peer-reviewed studies used machine learning to differentiate primary CNS lymphoma from gliomas in 2276 patients.

DATA ANALYSIS

Characteristics of data sets and machine learning algorithms were extracted. A meta-analysis on a subset of studies was performed. Reporting quality and risk of bias were assessed using the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) and Prediction Model Study Risk Of Bias Assessment Tool.

DATA SYNTHESIS

The highest area under the receiver operating characteristic curve (0.961) and accuracy (91.2%) in external validation were achieved by logistic regression and support vector machines models using conventional radiomic features. Meta-analysis of machine learning classifiers using these features yielded a mean area under the receiver operating characteristic curve of 0.944 (95% CI, 0.898-0.99). The median TRIPOD score was 51.7%. The risk of bias was high for 16 studies.

LIMITATIONS

Exclusion of abstracts decreased the sensitivity in evaluating all published studies. Meta-analysis had high heterogeneity.

CONCLUSIONS

Machine learning-based methods of differentiating primary CNS lymphoma from gliomas have shown great potential, but most studies lack large, balanced data sets and external validation. Assessment of the studies identified multiple deficiencies in reporting quality and risk of bias. These factors reduce the generalizability and reproducibility of the findings.

Intracranial Aneurysm Rupture Risk Estimation With Multidimensional Feature Fusion

The rupture of aneurysms is the main cause of spontaneous subarachnoid hemorrhage (SAH), which is a serious life-threatening disease with high mortality and permanent disability rates. Therefore, it is highly desirable to evaluate the rupture risk of aneurysms. In this study, we proposed a novel semiautomatic prediction model for the rupture risk estimation of aneurysms based on the CADA dataset, including 108 datasets with 125 annotated aneurysms. The model consisted of multidimensional feature fusion, feature selection, and the construction of classification methods. For the multidimensional feature fusion, we extracted four kinds of features and combined them into the feature set, including morphological features, radiomics features, clinical features, and deep learning features. Specifically, we applied the feature extractor 3D EfficientNet-B0 to extract and analyze the classification capabilities of three different deep learning features, namely, no-sigmoid features, sigmoid features, and binarization features. In the experiment, we constructed five distinct classification models, among which the k-nearest neighbor classifier showed the best performance for aneurysm rupture risk estimation, reaching an F2-score of 0.789. Our results suggest that the full use of multidimensional feature fusion can improve the performance of aneurysm rupture risk assessment. Compared with other methods, our method achieves the state-of-the-art performance for aneurysm rupture risk assessment methods based on CADA 2020.

Beyond Glioma: The Utility of Radiomic Analysis for Non-Glial Intracranial Tumors

Simple Summary

Tumor qualities, such as growth rate, firmness, and intrusion into healthy tissue, can be very important for operation planning and further treatment. Radiomics is a promising new method that allows the determination of some of these qualities on images performed before surgery. In this article, we provide a review of the use of radiomics in various tumors of the central nervous system, such as metastases, lymphoma, meningioma, medulloblastoma, and pituitary tumors.

Abstract

The field of radiomics is rapidly expanding and gaining a valuable role in neuro-oncology. The possibilities related to the use of radiomic analysis, such as distinguishing types of malignancies, predicting tumor grade, determining the presence of particular molecular markers, consistency, therapy response, and prognosis, can considerably influence decision-making in medicine in the near future. Even though the main focus of radiomic analyses has been on glial CNS tumors, studies on other intracranial tumors have shown encouraging results. Therefore, as the main focus of this review, we performed an analysis of publications on PubMed and Web of Science databases, focusing on radiomics in CNS metastases, lymphoma, meningioma, medulloblastoma, and pituitary tumors.

Computed tomography-based deep-learning prediction of induction chemotherapy treatment response in locally advanced nasopharyngeal carcinoma

BACKGROUND

Deep learning methods have great potential to predict treatment response. The objective of this study was to evaluate and validate the predictive performance of the computed tomography (CT)-based model using deep learning features for identification of responders and nonresponders to induction chemotherapy (IC) in nasopharyngeal carcinoma (NPC).

MATERIALS AND METHODS

All eligible patients were included retrospectively between January 2012 and December 2018, and assigned to the training (n = 208) or the testing cohort (n = 89). We extracted deep learning features of six pretrained convolutional neural networks (CNNs) via transfer learning method, and handcrafted radiomics features manually. Support vector machine (SVM) was adopted as the classifier. All predictive models were evaluated using the area under the receiver operating characteristics curve (AUC), by which an optimal model was selected. We also built clinical and clinical-radiological models for comparison.

RESULTS

The model with features extracted from ResNet50 (RN-SVM) had optimal performance among all models with features extracted from pretrained CNNs with an AUC of 0.811, accuracy of 68.54%, sensitivity of 61.54%, specificity of 87.50%, positive predictive value (PPV) of 93.02%, and negative predictive value (NPV) of 45.65% in the testing cohort. The handcrafted radiomics model was slightly inferior to the RN-SVM model with an AUC of 0.663 and accuracy of 60.67% in the testing cohort. All the imaging-derived models had better predictive performance than the clinical model.

CONCLUSION

The noninvasive deep learning method could provide efficient prediction of treatment response to IC in locally advanced NPC and might be a practicable approach in therapeutic strategy decision-making.

Radiomics Features of the Spleen as Surrogates for CT-Based Lymphoma Diagnosis and Subtype Differentiation

Simple Summary

In malignant lymphoma an early and accurate diagnosis is essential for therapy initiation and patient outcome. Within the diagnostic process, imaging plays a crucial role in disease staging. However, an invasive biopsy is required for subtype classification. Involvement of the spleen, a major lymphoid organ, is frequent in malignant lymphoma; this may be reactive or due to infiltration by malignant cells. Using radiomics features of the spleen in a machine learning approach, we investigated the possibility of distinguishing malignant lymphoma patients from other cancer patients and to classify lymphoma subtypes in the case of disease presence. Recent studies have proven the value of radiomics analysis in differentiating lymphoma from non-lymphoma groups on involved sites. Supported by machine learning, imaging could gain importance as a noninvasive diagnostic tool for future lymphoma classification, offering more precise radiological information for an interdisciplinary approach regarding treatment planning.

Abstract

The spleen is often involved in malignant lymphoma, which manifests on CT as either splenomegaly or focal, hypodense lymphoma lesions. This study aimed to investigate the diagnostic value of radiomics features of the spleen in classifying malignant lymphoma against non-lymphoma as well as the determination of malignant lymphoma subtypes in the case of disease presence—in particular Hodgkin lymphoma (HL), diffuse large B-cell lymphoma (DLBCL), mantle-cell lymphoma (MCL), and follicular lymphoma (FL). Spleen segmentations of 326 patients (139 female, median age 54.1 +/− 18.7 years) were generated and 1317 radiomics features per patient were extracted. For subtype classification, we created four different binary differentiation tasks and addressed them with a Random Forest classifier using 10-fold cross-validation. To detect the most relevant features, permutation importance was analyzed. Classifier results using all features were: malignant lymphoma vs. non-lymphoma AUC = 0.86 (p < 0.01); HL vs. NHL AUC = 0.75 (p < 0.01); DLBCL vs. other NHL AUC = 0.65 (p < 0.01); MCL vs. FL AUC = 0.67 (p < 0.01). Classifying malignant lymphoma vs. non-lymphoma was also possible using only shape features AUC = 0.77 (p < 0.01), with the most important feature being sphericity. Based on only shape features, a significant AUC could be achieved for all tasks, however, best results were achieved combining shape and textural features. This study demonstrates the value of splenic imaging and radiomic analysis in the diagnostic process in malignant lymphoma detection and subtype classification.

Using Ultrasound-Based Multilayer Perceptron to Differentiate Early Breast Mucinous Cancer and its Subtypes From Fibroadenoma

Objectives

Mucinous breast cancer (MBC), particularly pure MBC (pMBC), often tend to be confused with fibroadenoma (FA) due to their similar images and firm masses, so some MBC cases are misdiagnosed to be FA, which may cause poor prognosis. We analyzed the ultrasonic features and aimed to identify the ability of multilayer perceptron (MLP) to classify early MBC and its subtypes and FA.

Materials and Methods

The study consisted of 193 patients diagnosed with pMBC, mMBC, or FA. The area under curve (AUC) was calculated to assess the effectiveness of age and 10 ultrasound features in differentiating MBC from FA. We used the pairwise comparison to examine the differences among MBC subtypes (pure and mixed types) and FA. We utilized the MLP to differentiate MBC and its subtypes from FA.

Results

The nine features with AUCs over 0.5 were as follows: age, echo pattern, shape, orientation, margin, echo rim, vascularity distribution, vascularity grade, and tumor size. In subtype analysis, the significant differences were obtained in 10 variables (p-value range, 0.000–0.037) among pMBC, mMBC, and FA, except posterior feature. Through MLP, the AUCs of predicting MBC and FA were both 0.919; the AUCs of predicting pMBC, mMBC, and FA were 0.875, 0.767, and 0.927, respectively.

Conclusion

Our study found that the MLP models based on ultrasonic characteristics and age can well distinguish MBC and its subtypes from FA. It may provide a critical insight into MBC preoperative clinical management.

A non-invasive, automated diagnosis of Menière's disease using radiomics and machine learning on conventional magnetic resonance imaging: A multicentric, case-controlled feasibility study

Purpose

This study investigated the feasibility of a new image analysis technique (radiomics) on conventional MRI for the computer-aided diagnosis of Menière’s disease.

Materials and methods

A retrospective, multicentric diagnostic case–control study was performed. This study included 120 patients with unilateral or bilateral Menière’s disease and 140 controls from four centers in the Netherlands and Belgium. Multiple radiomic features were extracted from conventional MRI scans and used to train a machine learning-based, multi-layer perceptron classification model to distinguish patients with Menière’s disease from controls. The primary outcomes were accuracy, sensitivity, specificity, positive predictive value, and negative predictive value of the classification model.

Results

The classification accuracy of the machine learning model on the test set was 82%, with a sensitivity of 83%, and a specificity of 82%. The positive and negative predictive values were 71%, and 90%, respectively.

Conclusion

The multi-layer perceptron classification model yielded a precise, high-diagnostic performance in identifying patients with Menière’s disease based on radiomic features extracted from conventional T2-weighted MRI scans. In the future, radiomics might serve as a fast and noninvasive decision support system, next to clinical evaluation in the diagnosis of Menière’s disease.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11547-021-01425-w.

Pathologic Complete Response Prediction after Neoadjuvant Chemoradiation Therapy for Rectal Cancer Using Radiomics and Deep Embedding Network of MRI

Assessment of magnetic resonance imaging (MRI) after neoadjuvant chemoradiation therapy (nCRT) is essential in rectal cancer staging and treatment planning. However, when predicting the pathologic complete response (pCR) after nCRT for rectal cancer, existing works either rely on simple quantitative evaluation based on radiomics features or partially analyze multi-parametric MRI. We propose an effective pCR prediction method based on novel multi-parametric MRI embedding. We first seek to extract volumetric features of tumors that can be found only by analyzing multiple MRI sequences jointly. Specifically, we encapsulate multiple MRI sequences into multi-sequence fusion images (MSFI) and generate MSFI embedding. We merge radiomics features, which capture important characteristics of tumors, with MSFI embedding to generate multi-parametric MRI embedding and then use it to predict pCR using a random forest classifier. Our extensive experiments demonstrate that using all given MRI sequences is the most effective regardless of the dimension reduction method. The proposed method outperformed any variants with different combinations of feature vectors and dimension reduction methods or different classification models. Comparative experiments demonstrate that it outperformed four competing baselines in terms of the AUC and F1-score. We use MRI sequences from 912 patients with rectal cancer, a much larger sample than in any existing work.

Machine learning for evolutive lymphoma and residual masses recognition in whole body diffusion weighted magnetic resonance images

BACKGROUND

After the treatment of the patients with malignant lymphoma, there may persist lesions that must be labeled either as evolutive lymphoma requiring new treatments or as residual masses. We present in this work, a machine learning-based computer-aided diagnosis (CAD) applied to whole-body diffusion-weighted magnetic resonance images.

METHODS

The database consists of a total of 1005 MRI images with evolutive lymphoma and residual masses. More specifically, we propose a novel approach that leverages: (1)-The complementarity of the functional and anatomical criteria of MRI images through a fusion step based on the discrete wavelet transforms (DWT). (2)- The automatic segmentation of the lesions, their localization, and their enumeration using the Chan-Vese algorithm. (3)- The generation of the parametric image which contains the apparent diffusion coefficient value named ADC map. (4)- The features selection through the application of the sequential forward selection (SFS), Entropy, Symmetric uncertainty and Gain Ratio algorithm on 72 extracted features. (5)- The classification of the lesions by applying five well known supervised machine learning classification algorithms: the back-propagation artificial neural network (ANN), the support vector machine (SVM), the K-nearest neighbours (K-NN), Relevance Vectors Machine (RVM), and the random forest (RF) compared to deep learning based on convolutional neural network (CNN). Moreover, this study is achieved with an evaluation of the classification using 335 DW-MR images where 80% of them are used for the training and the remaining 20% for the test.

RESULTS

The obtained accuracy for the five classifiers recorded a slight superiority to the proposed method based on the back-propagation 3-9-1 ANN model which reaches 96,5%. In addition, we compared the proposed method to five other works from the literature. The proposed method gives much better results in terms of SE, SP, accuracy, F₁-measure, and geometric-mean which reaches respectively 96.4%, 90.9%, 95.5%, 0.97, and 91.61%.

CONCLUSIONS

Our initial results suggest that Combining functional, anatomical, and morphological features of ROI's have very good accuracy (97.01%) for evolutive lymphoma and residual masses recognition when we based on the new proposed approach using the back-propagation 3-9-1 ANN model. Proposed method based on machine learning gives less than Deep learning CNN, which is 98.5%.

Filtration-Histogram Based Magnetic Resonance Texture Analysis (MRTA) for the Distinction of Primary Central Nervous System Lymphoma and Glioblastoma

Primary central nervous system lymphoma (PCNSL) has variable imaging appearances, which overlap with those of glioblastoma (GBM), thereby necessitating invasive tissue diagnosis. We aimed to investigate whether a rapid filtration histogram analysis of clinical MRI data supports the distinction of PCNSL from GBM. Ninety tumours (PCNSL n = 48, GBM n = 42) were analysed using pre-treatment MRI sequences (T₁-weighted contrast-enhanced (T₁CE), T₂-weighted (T₂), and apparent diffusion coefficient maps (ADC)). The segmentations were completed with proprietary texture analysis software (TexRAD version 3.3). Filtered (five filter sizes SSF = 2-6 mm) and unfiltered (SSF = 0) histogram parameters were compared using Mann-Whitney U non-parametric testing, with receiver operating characteristic (ROC) derived area under the curve (AUC) analysis for significant results. Across all (n = 90) tumours, the optimal algorithm performance was achieved using an unfiltered ADC mean and the mean of positive pixels (MPP), with a sensitivity of 83.8%, specificity of 8.9%, and AUC of 0.88. For subgroup analysis with >1/3 necrosis masses, ADC permitted the identification of PCNSL with a sensitivity of 96.9% and specificity of 100%. For T₁CE-derived regions, the distinction was less accurate, with a sensitivity of 71.4%, specificity of 77.1%, and AUC of 0.779. A role may exist for cross-sectional texture analysis without complex machine learning models to differentiate PCNSL from GBM. ADC appears the most suitable sequence, especially for necrotic lesion distinction.

Incorporating radiomics into clinical trials: expert consensus endorsed by the European Society of Radiology on considerations for data-driven compared to biologically driven quantitative biomarkers

Existing quantitative imaging biomarkers (QIBs) are associated with known biological tissue characteristics and follow a well-understood path of technical, biological and clinical validation before incorporation into clinical trials. In radiomics, novel data-driven processes extract numerous visually imperceptible statistical features from the imaging data with no a priori assumptions on their correlation with biological processes. The selection of relevant features (radiomic signature) and incorporation into clinical trials therefore requires additional considerations to ensure meaningful imaging endpoints. Also, the number of radiomic features tested means that power calculations would result in sample sizes impossible to achieve within clinical trials. This article examines how the process of standardising and validating data-driven imaging biomarkers differs from those based on biological associations. Radiomic signatures are best developed initially on datasets that represent diversity of acquisition protocols as well as diversity of disease and of normal findings, rather than within clinical trials with standardised and optimised protocols as this would risk the selection of radiomic features being linked to the imaging process rather than the pathology. Normalisation through discretisation and feature harmonisation are essential pre-processing steps. Biological correlation may be performed after the technical and clinical validity of a radiomic signature is established, but is not mandatory. Feature selection may be part of discovery within a radiomics-specific trial or represent exploratory endpoints within an established trial; a previously validated radiomic signature may even be used as a primary/secondary endpoint, particularly if associations are demonstrated with specific biological processes and pathways being targeted within clinical trials. KEY POINTS: • Data-driven processes like radiomics risk false discoveries due to high-dimensionality of the dataset compared to sample size, making adequate diversity of the data, cross-validation and external validation essential to mitigate the risks of spurious associations and overfitting. • Use of radiomic signatures within clinical trials requires multistep standardisation of image acquisition, image analysis and data mining processes. • Biological correlation may be established after clinical validation but is not mandatory.

Radiomic Based Machine Learning Performance for a Three Class Problem in Neuro-Oncology: Time to Test the Waters?

Prior radiomics studies have focused on two-class brain tumor classification, which limits generalizability. The performance of radiomics in differentiating the three most common malignant brain tumors (glioblastoma (GBM), primary central nervous system lymphoma (PCNSL), and metastatic disease) is assessed; factors affecting the model performance and usefulness of a single sequence versus multiparametric MRI (MP-MRI) remain largely unaddressed. This retrospective study included 253 patients (120 metastatic (lung and brain), 40 PCNSL, and 93 GBM). Radiomic features were extracted for whole a tumor mask (enhancing plus necrotic) and an edema mask (first pipeline), as well as for separate enhancing and necrotic and edema masks (second pipeline). Model performance was evaluated using MP-MRI, individual sequences, and the T1 contrast enhanced (T1-CE) sequence without the edema mask across 45 model/feature selection combinations. The second pipeline showed significantly high performance across all combinations (Brier score: 0.311-0.325). GBRM fit using the full feature set from the T1-CE sequence was the best model. The majority of the top models were built using a full feature set and inbuilt feature selection. No significant difference was seen between the top-performing models for MP-MRI (AUC 0.910) and T1-CE sequence with (AUC 0.908) and without edema masks (AUC 0.894). T1-CE is the single best sequence with comparable performance to that of multiparametric MRI (MP-MRI). Model performance varies based on tumor subregion and the combination of model/feature selection methods.

Machine Learning and Radiomics Applications in Esophageal Cancers Using Non-Invasive Imaging Methods-A Critical Review of Literature

Esophageal cancer (EC) is of public health significance as one of the leading causes of cancer death worldwide. Accurate staging, treatment planning and prognostication in EC patients are of vital importance. Recent advances in machine learning (ML) techniques demonstrate their potential to provide novel quantitative imaging markers in medical imaging. Radiomics approaches that could quantify medical images into high-dimensional data have been shown to improve the imaging-based classification system in characterizing the heterogeneity of primary tumors and lymph nodes in EC patients. In this review, we aim to provide a comprehensive summary of the evidence of the most recent developments in ML application in imaging pertinent to EC patient care. According to the published results, ML models evaluating treatment response and lymph node metastasis achieve reliable predictions, ranging from acceptable to outstanding in their validation groups. Patients stratified by ML models in different risk groups have a significant or borderline significant difference in survival outcomes. Prospective large multi-center studies are suggested to improve the generalizability of ML techniques with standardized imaging protocols and harmonization between different centers.

Machine learning applications to neuroimaging for glioma detection and classification: An artificial intelligence augmented systematic review

Glioma is the most common primary intraparenchymal tumor of the brain and the 5-year survival rate of high-grade glioma is poor. Magnetic resonance imaging (MRI) is essential for detecting, characterizing and monitoring brain tumors but definitive diagnosis still relies on surgical pathology. Machine learning has been applied to the analysis of MRI data in glioma research and has the potential to change clinical practice and improve patient outcomes. This systematic review synthesizes and analyzes the current state of machine learning applications to glioma MRI data and explores the use of machine learning for systematic review automation. Various datapoints were extracted from the 153 studies that met inclusion criteria and analyzed. Natural language processing (NLP) analysis involved keyword extraction, topic modeling and document classification. Machine learning has been applied to tumor grading and diagnosis, tumor segmentation, non-invasive genomic biomarker identification, detection of progression and patient survival prediction. Model performance was generally strong (AUC = 0.87 ± 0.09; sensitivity = 0.87 ± 0.10; specificity = 0.0.86 ± 0.10; precision = 0.88 ± 0.11). Convolutional neural network, support vector machine and random forest algorithms were top performers. Deep learning document classifiers yielded acceptable performance (mean 5-fold cross-validation AUC = 0.71). Machine learning tools and data resources were synthesized and summarized to facilitate future research. Machine learning has been widely applied to the processing of MRI data in glioma research and has demonstrated substantial utility. NLP and transfer learning resources enabled the successful development of a replicable method for automating the systematic review article screening process, which has potential for shortening the time from discovery to clinical application in medicine.

Radiomics-Based Differentiation between Glioblastoma, CNS Lymphoma, and Brain Metastases: Comparing Performance across MRI Sequences and Machine Learning Models

Prior radiomics studies have focused on two-class brain tumor classification, which limits generalizability. The performance of radiomics in differentiating the three most common malignant brain tumors (glioblastoma (GBM), primary central nervous system lymphoma (PCNSL), and metastatic disease) is assessed; factors affecting the model performance and usefulness of a single sequence versus multiparametric MRI (MP-MRI) remain largely unaddressed. This retrospective study included 253 patients (120 metastatic (lung and brain), 40 PCNSL, and 93 GBM). Radiomic features were extracted for whole a tumor mask (enhancing plus necrotic) and an edema mask (first pipeline), as well as for separate enhancing and necrotic and edema masks (second pipeline). Model performance was evaluated using MP-MRI, individual sequences, and the T1 contrast enhanced (T1-CE) sequence without the edema mask across 45 model/feature selection combinations. The second pipeline showed significantly high performance across all combinations (Brier score: 0.311–0.325). GBRM fit using the full feature set from the T1-CE sequence was the best model. The majority of the top models were built using a full feature set and inbuilt feature selection. No significant difference was seen between the top-performing models for MP-MRI (AUC 0.910) and T1-CE sequence with (AUC 0.908) and without edema masks (AUC 0.894). T1-CE is the single best sequence with comparable performance to that of multiparametric MRI (MP-MRI). Model performance varies based on tumor subregion and the combination of model/feature selection methods.

Radiomics-based differentiation between glioblastoma and primary central nervous system lymphoma: a comparison of diagnostic performance across different MRI sequences and machine learning techniques

OBJECTIVES

Despite the robust diagnostic performance of MRI-based radiomic features for differentiating between glioblastoma (GBM) and primary central nervous system lymphoma (PCNSL) reported on prior studies, the best sequence or a combination of sequences and model performance across various machine learning pipelines remain undefined. Herein, we compare the diagnostic performance of multiple radiomics-based models to differentiate GBM from PCNSL.

METHODS

Our retrospective study included 94 patients (34 with PCNSL and 60 with GBM). Model performance was assessed using various MRI sequences across 45 possible model and feature selection combinations for nine different sequence permutations. Predictive performance was assessed using fivefold repeated cross-validation with five repeats. The best and worst performing models were compared to assess differences in performance.

RESULTS

The predictive performance, both using individual and a combination of sequences, was fairly robust across multiple top performing models (AUC: 0.961-0.977) but did show considerable variation between the best and worst performing models. The top performing individual sequences had comparable performance to multiparametric models. The best prediction model in our study used a combination of ADC, FLAIR, and T1-CE achieving the highest AUC of 0.977, while the second ranked model used T1-CE and ADC, achieving a cross-validated AUC of 0.975.

CONCLUSION

Radiomics-based predictive accuracy can vary considerably, based on the model and feature selection methods as well as the combination of sequences used. Also, models derived from limited sequences show performance comparable to those derived from all five sequences.

KEY POINTS

• Radiomics-based diagnostic performance of various machine learning models for differentiating glioblastoma and PCNSL varies considerably. • ML models using limited or multiple MRI sequences can provide comparable performance, based on the chosen model. • Embedded feature selection models perform better than models using a priori feature reduction.

Differentiation of Glioma Mimicking Encephalitis and Encephalitis Using Multiparametric MR-Based Deep Learning

Background

Computational aid for diagnosis based on convolutional neural network (CNN) is promising to improve clinical diagnostic performance. Therefore, we applied pretrained CNN models in multiparametric magnetic resonance (MR) images to classify glioma mimicking encephalitis and encephalitis.

Methods

A data set containing 3064 MRI brain images from 164 patients with a final diagnosis of glioma (n = 56) and encephalitis (n = 108) patients and divided into training and testing sets. We applied three MRI modalities [fluid attenuated inversion recovery (FLAIR), contrast enhanced-T1 weighted imaging (CE-T1WI) and T2 weighted imaging (T2WI)] as the input data to build three pretrained deep CNN models (Alexnet, ResNet-50, and Inception-v3), and then compared their classification performance with radiologists' diagnostic performance. These models were evaluated by using the area under the receiver operator characteristic curve (AUC) of a five-fold cross-validation and the accuracy, sensitivity, specificity were analyzed.

Results

The three pretrained CNN models all had AUC values over 0.9 with excellent performance. The highest classification accuracy of 97.57% was achieved by the Inception-v3 model based on the T2WI data. In addition, Inception-v3 performed statistically significantly better than the Alexnet architecture (p<0.05). For Inception-v3 and ResNet-50 models, T2WI offered the highest accuracy, followed by CE-T1WI and FLAIR. The performance of Inception-v3 and ResNet-50 had a significant difference with radiologists (p<0.05), but there was no significant difference between the results of the Alexnet and those of a more experienced radiologist (p >0.05).

Conclusions

The pretrained CNN models can automatically and accurately classify these two diseases and further help to improving clinical diagnostic performance.

Deep Learning for Automatic Differential Diagnosis of Primary Central Nervous System Lymphoma and Glioblastoma: Multi-Parametric Magnetic Resonance Imaging Based Convolutional Neural Network Model

BACKGROUND

Differential diagnosis of primary central nervous system lymphoma (PCNSL) and glioblastoma (GBM) is useful to guide treatment strategies.

PURPOSE

To investigate the use of a convolutional neural network (CNN) model for differentiation of PCNSL and GBM without tumor delineation.

STUDY TYPE

Retrospective.

POPULATION

A total of 289 patients with PCNSL (136) or GBM (153) were included, the average age of the cohort was 54 years, and there were 173 men and 116 women.

FIELD STRENGTH/SEQUENCE

3.0 T Axial contrast-enhanced T₁ -weighted spin-echo inversion recovery sequence (CE-T₁ WI), T₂ -weighted fluid-attenuation inversion recovery sequence (FLAIR), and diffusion weighted imaging (DWI, b = 0 second/mm² , 1000 seconds/mm² ).

ASSESSMENT

A single-parametric CNN model was built using CE-T₁ WI, FLAIR, and the apparent diffusion coefficient (ADC) map derived from DWI, respectively. A decision-level fusion based multi-parametric CNN model (DF-CNN) was built by combining the predictions of single-parametric CNN models through logistic regression. An image-level fusion based multi-parametric CNN model (IF-CNN) was built using the integrated multi-parametric MR images. The radiomics models were developed. The diagnoses by three radiologists with 6 years (junior radiologist Y.Y.), 11 years (intermediate-level radiologist Y.T.), and 21 years (senior radiologist Y.L.) of experience were obtained.

STATISTICAL ANALYSIS

The 5-fold cross validation was used for model evaluation. The Pearson's chi-squared test was used to compare the accuracies. U-test and Fisher's exact test were used to compare clinical characteristics.

RESULTS

The CE-T₁ WI, FLAIR, and ADC based single-parametric CNN model had accuracy of 0.884, 0.782, and 0.700, respectively. The DF-CNN model had an accuracy of 0.899 which was higher than the IF-CNN model (0.830, P = 0.021), but had no significant difference in accuracy compared to the radiomics model (0.865, P = 0.255), and the senior radiologist (0.906, P = 0.886).

DATA CONCLUSION

A CNN model can differentiate PCNSL from GBM without tumor delineation, and comparable to the radiomics models and radiologists.

LEVEL OF EVIDENCE

4 TECHNICAL EFFICACY: Stage 2.

Glioblastoma and primary central nervous system lymphoma: differentiation using MRI derived first-order texture analysis - a machine learning study

OBJECTIVES

To evaluate the diagnostic performance of multiple machine learning classifier models derived from first-order histogram texture parameters extracted from T1-weighted contrast-enhanced images in differentiating glioblastoma and primary central nervous system lymphoma.

METHODS

Retrospective study with 97 glioblastoma and 46 primary central nervous system lymphoma patients. Thirty-six different combinations of classifier models and feature selection techniques were evaluated. Five-fold nested cross-validation was performed. Model performance was assessed for whole tumour and largest single slice using receiver operating characteristic curve.

RESULTS

The cross-validated model performance was relatively similar for the top performing models for both whole tumour and largest single slice (area under the curve 0.909-0.924). However, there was a considerable difference between the worst performing model (logistic regression with full feature set, area under the curve 0.737) and the highest performing model for whole tumour (least absolute shrinkage and selection operator model with correlation filter, area under the curve 0.924). For single slice, the multilayer perceptron model with correlation filter had the highest performance (area under the curve 0.914). No significant difference was seen between the diagnostic performance of the top performing model for both whole tumour and largest single slice.

CONCLUSIONS

T1 contrast-enhanced derived first-order texture analysis can differentiate between glioblastoma and primary central nervous system lymphoma with good diagnostic performance. The machine learning performance can vary significantly depending on the model and feature selection methods. Largest single slice and whole tumour analysis show comparable diagnostic performance.