MRI radiomics-based machine-learning classification of bone chondrosarcoma

Enhanced enchondroma detection from x-ray images using deep learning: A step towards accurate and cost-effective diagnosis

This study investigates the automated detection of enchondromas, benign cartilage tumors, from x-ray images using deep learning techniques. Enchondromas pose diagnostic challenges due to their potential for malignant transformation and overlapping radiographic features with other conditions. Leveraging a data set comprising 1645 x-ray images from 1173 patients, a deep-learning model implemented with Detectron2 achieved an accuracy of 0.9899 in detecting enchondromas. The study employed rigorous validation processes and compared its findings with the existing literature, highlighting the superior performance of the deep learning approach. Results indicate the potential of machine learning in improving diagnostic accuracy and reducing healthcare costs associated with advanced imaging modalities. The study underscores the significance of early and accurate detection of enchondromas for effective patient management and suggests avenues for further research in musculoskeletal tumor detection.

Bone Reporting and Data System (Bone-RADS) and Other Proposed Practice Guidelines for Reporting Bone Tumors

BACKGROUND

The purpose of this article is to review the different bone tumor radiology reporting systems [Bone Reporting and Data System (Bone-RADS), Osseous Tumor Reporting and Data System (OT-RADS), Solitary Bone Tumor Imaging Reporting and Data System (BTI-RADS), and Radiological Evaluation Score for Bone Tumors (REST)] and summarize their advantages and disadvantages.

METHODS

A selective search of PubMed was performed for literature regarding the definition and discussion of bone tumor reporting systems. No time frame was selected, but the search was particularly focused on current literature on musculoskeletal radiology lexicon.

RESULTS

To date, four major reporting systems has been proposed to standardize and systematize the reporting of imaging studies of bone tumors: Bone-RADS, OT-RADS, BTI-RADS, and REST. Both Bone-RADS and OT-RADS aid in the characterization and management of bone lesions on CT and MRI. OT-RADS and REST can be applied to MRI and radiography, respectively.

CONCLUSION

Radiologists play a central role in the detection and characterization of asymptomatic (or incidentally detected) and symptomatic bone tumors. There are several existing bone tumor reporting systems with various advantages and disadvantages including emphasis on lesion characterization as well as management of incidentally detected bone lesions.

KEY POINTS

· Four bone tumor reporting systems have been proposed thus far.. · Bone-RADS guides management of incidental bone lesions on CT and MRI.. · OT-RADS guides management of bone lesions on MRI with high accuracy.. · BTI-RADS classifies bone tumors on CT and MRI..

CITATION FORMAT

· Ghasemi A, Ahlawat S, . Bone Reporting and Data System (Bone-RADS) and Other Proposed Practice Guidelines for Reporting Bone Tumors. Fortschr Röntgenstr 2024; 196: 1134 - 1142.

A retrospective external validation study of the Birmingham Atypical Cartilage Tumour Imaging Protocol (BACTIP) for the management of solitary central cartilage tumours of the proximal humerus and around the knee

OBJECTIVES

This study aimed to externally validate the Birmingham Atypical Cartilage Tumour Imaging Protocol (BACTIP) recommendations for differentiation/follow-up of central cartilage tumours (CCTs) of the proximal humerus, distal femur, and proximal tibia and to propose BACTIP adaptations if the results provide new insights.

METHODS

MRIs of 123 patients (45 ± 11 years, 37 men) with an untreated CCT with MRI follow-up (n = 62) or histopathological confirmation (n = 61) were retrospectively/consecutively included and categorised following the BACTIP (2003-2020 / Ghent University Hospital/Belgium). Tumour length and endosteal scalloping differences between enchondroma, atypical cartilaginous tumour (ACT), and high-grade chondrosarcoma (CS II/III/dedifferentiated) were evaluated. ROC-curve analysis for differentiating benign from malignant CCTs and for evaluating the BACTIP was performed.

RESULTS

For lesion length and endosteal scalloping, ROC-AUCs were poor and fair-excellent, respectively, for differentiating different CCT groups (0.59-0.69 versus 0.73-0.91). The diagnostic performance of endosteal scalloping and the BACTIP was higher than that of lesion length. A 1° endosteal scalloping cut-off differentiated enchondroma from ACT + high-grade chondrosarcoma with a sensitivity of 90%, reducing the potential diagnostic delay. However, the specificity was 29%, inducing overmedicalisation (excessive follow-up). ROC-AUC of the BACTIP was poor for differentiating enchondroma from ACT (ROC-AUC = 0.69; 95%CI = 0.51-0.87; p = 0.041) and fair-good for differentiation between other CCT groups (ROC-AUC = 0.72-0.81). BACTIP recommendations were incorrect/unsafe in five ACTs and one CSII, potentially inducing diagnostic delay. Eleven enchondromas received unnecessary referrals/follow-up.

CONCLUSION

Although promising as a useful tool for management/follow-up of CCTs of the proximal humerus, distal femur, and proximal tibia, five ACTs and one chondrosarcoma grade II were discharged, potentially inducing diagnostic delay, which could be reduced by adapting BACTIP cut-off values.

CLINICAL RELEVANCE STATEMENT

Mostly, Birmingham Atypical Cartilage Tumour Imaging Protocol (BACTIP) assesses central cartilage tumours of the proximal humerus and the knee correctly. Both when using the BACTIP and when adapting cut-offs, caution should be taken for the trade-off between underdiagnosis/potential diagnostic delay in chondrosarcomas and overmedicalisation in enchondromas.

KEY POINTS

• This retrospective external validation confirms the Birmingham Atypical Cartilage Tumour Imaging Protocol as a useful tool for initial assessment and follow-up recommendation of central cartilage tumours in the proximal humerus and around the knee in the majority of cases. • Using only the Birmingham Atypical Cartilage Tumour Imaging Protocol, both atypical cartilaginous tumours and high-grade chondrosarcomas (grade II, grade III, and dedifferentiated chondrosarcomas) can be misdiagnosed, excluding them from specialist referral and further follow-up, thus creating a potential risk of delayed diagnosis and worse prognosis. • Adapted cut-offs to maximise detection of atypical cartilaginous tumours and high-grade chondrosarcomas, minimise underdiagnosis and reduce potential diagnostic delay in malignant tumours but increase unnecessary referral and follow-up of benign tumours.

Preoperative prediction of histopathological grading in patients with chondrosarcoma using MRI-based radiomics with semantic features

BACKGROUND

Distinguishing high-grade from low-grade chondrosarcoma is extremely vital not only for guiding the development of personalized surgical treatment but also for predicting the prognosis of patients. We aimed to establish and validate a magnetic resonance imaging (MRI)-based nomogram for predicting preoperative grading in patients with chondrosarcoma.

METHODS

Approximately 114 patients (60 and 54 cases with high-grade and low-grade chondrosarcoma, respectively) were recruited for this retrospective study. All patients were treated via surgery and histopathologically proven, and they were randomly divided into training (n = 80) and validation (n = 34) sets at a ratio of 7:3. Next, radiomics features were extracted from two sequences using the least absolute shrinkage and selection operator (LASSO) algorithms. The rad-scores were calculated and then subjected to logistic regression to develop a radiomics model. A nomogram combining independent predictive semantic features with radiomic by using multivariate logistic regression was established. The performance of each model was assessed by the receiver operating characteristic (ROC) curve analysis and the area under the curve, while clinical efficacy was evaluated via decision curve analysis (DCA).

RESULTS

Ultimately, six optimal radiomics signatures were extracted from T1-weighted imaging (T1WI) and T2-weighted imaging with fat suppression (T2WI-FS) sequences to develop the radiomics model. Tumour cartilage abundance, which emerged as an independent predictor, was significantly related to chondrosarcoma grading (p < 0.05). The AUC values of the radiomics model were 0.85 (95% CI, 0.76 to 0.95) in the training sets, and the corresponding AUC values in the validation sets were 0.82 (95% CI, 0.65 to 0.98), which were far superior to the clinical model AUC values of 0.68 (95% CI, 0.58 to 0.79) in the training sets and 0.72 (95% CI, 0.57 to 0.87) in the validation sets. The nomogram demonstrated good performance in the preoperative distinction of chondrosarcoma. The DCA analysis revealed that the nomogram model had a markedly higher clinical usefulness in predicting chondrosarcoma grading preoperatively than either the rad-score or clinical model alone.

CONCLUSION

The nomogram based on MRI radiomics combined with optimal independent factors had better performance for the preoperative differentiation between low-grade and high-grade chondrosarcoma and has potential as a noninvasive preoperative tool for personalizing clinical plans.

Enhancing radiographic image interpretation: WARES-PRS model for knee bone tumour detection

The early diagnosis of tumour is significant in biomedical research field to lower the severity level and restrict the process extension from cancer. Moreover, the detection of early sign of cancer is undertaken with extensive research efforts that dedicated to the disclosure and recognition of tumours. However, the limited data size as well as diverse appearance of images lowered the detection performance and failed to detect complex stage of tumour. So to solve these issues, a Weighted Adaptive Random Ensemble Support Vector-based Partial Reinforcement Search (WARES-PRS) algorithm is proposed that detected bone lesions accurately and also predicted the severity level stage efficiently. Further, the detection is performed with varied stages to diminish the presence of noise and undertaken effective classification. The performance is validated with CNUH dataset that enhanced image pre-processing tasks. Despite the proposed method uncover the mutual relationships between each pixel's local texture and the overall image's global context. The detection and classification efficiency is validated with various measures and the experimental results revealed that the detection accuracy is enhanced for the proposed approach by 98.5%. The outcomes of our study have exhibited a substantial contribution to assisting physicians in the detection of knee bone tumours.

Diagnostic Performance of Artificial Intelligence in Detection of Primary Malignant Bone Tumors: a Meta-Analysis

We aim to conduct a meta-analysis on studies that evaluated the diagnostic performance of artificial intelligence (AI) algorithms in the detection of primary bone tumors, distinguishing them from other bone lesions, and comparing them with clinician assessment. A systematic search was conducted using a combination of keywords related to bone tumors and AI. After extracting contingency tables from all included studies, we performed a meta-analysis using random-effects model to determine the pooled sensitivity and specificity, accompanied by their respective 95% confidence intervals (CI). Quality assessment was evaluated using a modified version of Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) and Prediction Model Study Risk of Bias Assessment Tool (PROBAST). The pooled sensitivities for AI algorithms and clinicians on internal validation test sets for detecting bone neoplasms were 84% (95% CI: 79.88) and 76% (95% CI: 64.85), and pooled specificities were 86% (95% CI: 81.90) and 64% (95% CI: 55.72), respectively. At external validation, the pooled sensitivity and specificity for AI algorithms were 84% (95% CI: 75.90) and 91% (95% CI: 83.96), respectively. The same numbers for clinicians were 85% (95% CI: 73.92) and 94% (95% CI: 89.97), respectively. The sensitivity and specificity for clinicians with AI assistance were 95% (95% CI: 86.98) and 57% (95% CI: 48.66). Caution is needed when interpreting findings due to potential limitations. Further research is needed to bridge this gap in scientific understanding and promote effective implementation for medical practice advancement.

X-rays radiomics-based machine learning classification of atypical cartilaginous tumour and high-grade chondrosarcoma of long bones

BACKGROUND

Atypical cartilaginous tumour (ACT) and high-grade chondrosarcoma (CS) of long bones are respectively managed with active surveillance or curettage and wide resection. Our aim was to determine diagnostic performance of X-rays radiomics-based machine learning for classification of ACT and high-grade CS of long bones.

METHODS

This retrospective, IRB-approved study included 150 patients with surgically treated and histology-proven lesions at two tertiary bone sarcoma centres. At centre 1, the dataset was split into training (n = 71 ACT, n = 24 high-grade CS) and internal test (n = 19 ACT, n = 6 high-grade CS) cohorts, respectively, based on the date of surgery. At centre 2, the dataset constituted the external test cohort (n = 12 ACT, n = 18 high-grade CS). Manual segmentation was performed on frontal view X-rays, using MRI or CT for preliminary identification of lesion margins. After image pre-processing, radiomic features were extracted. Dimensionality reduction included stability, coefficient of variation, and mutual information analyses. In the training cohort, after class balancing, a machine learning classifier (Support Vector Machine) was automatically tuned using nested 10-fold cross-validation. Then, it was tested on both the test cohorts and compared to two musculoskeletal radiologists' performance using McNemar's test.

FINDINGS

Five radiomic features (3 morphology, 2 texture) passed dimensionality reduction. After tuning on the training cohort (AUC = 0.75), the classifier had 80%, 83%, 79% and 80%, 89%, 67% accuracy, sensitivity, and specificity in the internal (temporally independent) and external (geographically independent) test cohorts, respectively, with no difference compared to the radiologists (p ≥ 0.617).

INTERPRETATION

X-rays radiomics-based machine learning accurately differentiates between ACT and high-grade CS of long bones.

FUNDING

AIRC Investigator Grant.

CT and MRI radiomics of bone and soft-tissue sarcomas: an updated systematic review of reproducibility and validation strategies

OBJECTIVE

To systematically review radiomic feature reproducibility and model validation strategies in recent studies dealing with CT and MRI radiomics of bone and soft-tissue sarcomas, thus updating a previous version of this review which included studies published up to 2020.

METHODS

A literature search was conducted on EMBASE and PubMed databases for papers published between January 2021 and March 2023. Data regarding radiomic feature reproducibility and model validation strategies were extracted and analyzed.

RESULTS

Out of 201 identified papers, 55 were included. They dealt with radiomics of bone (n = 23) or soft-tissue (n = 32) tumors. Thirty-two (out of 54 employing manual or semiautomatic segmentation, 59%) studies included a feature reproducibility analysis. Reproducibility was assessed based on intra/interobserver segmentation variability in 30 (55%) and geometrical transformations of the region of interest in 2 (4%) studies. At least one machine learning validation technique was used for model development in 34 (62%) papers, and K-fold cross-validation was employed most frequently. A clinical validation of the model was reported in 38 (69%) papers. It was performed using a separate dataset from the primary institution (internal test) in 22 (40%), an independent dataset from another institution (external test) in 14 (25%) and both in 2 (4%) studies.

CONCLUSIONS

Compared to papers published up to 2020, a clear improvement was noted with almost double publications reporting methodological aspects related to reproducibility and validation. Larger multicenter investigations including external clinical validation and the publication of databases in open-access repositories could further improve methodology and bring radiomics from a research area to the clinical stage.

CRITICAL RELEVANCE STATEMENT

An improvement in feature reproducibility and model validation strategies has been shown in this updated systematic review on radiomics of bone and soft-tissue sarcomas, highlighting efforts to enhance methodology and bring radiomics from a research area to the clinical stage.

KEY POINTS

• 2021-2023 radiomic studies on CT and MRI of musculoskeletal sarcomas were reviewed. • Feature reproducibility was assessed in more than half (59%) of the studies. • Model clinical validation was performed in 69% of the studies. • Internal (44%) and/or external (29%) test datasets were employed for clinical validation.

AI applications in musculoskeletal imaging: a narrative review

This narrative review focuses on clinical applications of artificial intelligence (AI) in musculoskeletal imaging. A range of musculoskeletal disorders are discussed using a clinical-based approach, including trauma, bone age estimation, osteoarthritis, bone and soft-tissue tumors, and orthopedic implant-related pathology. Several AI algorithms have been applied to fracture detection and classification, which are potentially helpful tools for radiologists and clinicians. In bone age assessment, AI methods have been applied to assist radiologists by automatizing workflow, thus reducing workload and inter-observer variability. AI may potentially aid radiologists in identifying and grading abnormal findings of osteoarthritis as well as predicting the onset or progression of this disease. Either alone or combined with radiomics, AI algorithms may potentially improve diagnosis and outcome prediction of bone and soft-tissue tumors. Finally, information regarding appropriate positioning of orthopedic implants and related complications may be obtained using AI algorithms. In conclusion, rather than replacing radiologists, the use of AI should instead help them to optimize workflow, augment diagnostic performance, and keep up with ever-increasing workload.Relevance statement This narrative review provides an overview of AI applications in musculoskeletal imaging. As the number of AI technologies continues to increase, it will be crucial for radiologists to play a role in their selection and application as well as to fully understand their potential value in clinical practice. Key points • AI may potentially assist musculoskeletal radiologists in several interpretative tasks.• AI applications to trauma, age estimation, osteoarthritis, tumors, and orthopedic implants are discussed.• AI should help radiologists to optimize workflow and augment diagnostic performance.

Development of A Radiomic Model for MGMT Promoter Methylation Detection in Glioblastoma Using Conventional MRI

The methylation of the O6-methylguanine-DNA methyltransferase (MGMT) promoter is a molecular marker associated with a better response to chemotherapy in patients with glioblastoma (GB). Standard pre-operative magnetic resonance imaging (MRI) analysis is not adequate to detect MGMT promoter methylation. This study aims to evaluate whether the radiomic features extracted from multiple tumor subregions using multiparametric MRI can predict MGMT promoter methylation status in GB patients. This retrospective single-institution study included a cohort of 277 GB patients whose 3D post-contrast T1-weighted images and 3D fluid-attenuated inversion recovery (FLAIR) images were acquired using two MRI scanners. Three separate regions of interest (ROIs) showing tumor enhancement, necrosis, and FLAIR hyperintensities were manually segmented for each patient. Two machine learning algorithms (support vector machine (SVM) and random forest) were built for MGMT promoter methylation prediction from a training cohort (196 patients) and tested on a separate validation cohort (81 patients), based on a set of automatically selected radiomic features, with and without demographic variables (i.e., patients' age and sex). In the training set, SVM based on the selected radiomic features of the three separate ROIs achieved the best performances, with an average of 83.0% (standard deviation: 5.7%) for accuracy and 0.894 (0.056) for the area under the curve (AUC) computed through cross-validation. In the test set, all classification performances dropped: the best was obtained by SVM based on the selected features extracted from the whole tumor lesion constructed by merging the three ROIs, with 64.2% (95% confidence interval: 52.8-74.6%) accuracy and 0.572 (0.439-0.705) for AUC. The performances did not change when the patients' age and sex were included with the radiomic features into the models. Our study confirms the presence of a subtle association between imaging characteristics and MGMT promoter methylation status. However, further verification of the strength of this association is needed, as the low diagnostic performance obtained in this validation cohort is not sufficiently robust to allow clinically meaningful predictions.

Artificial intelligence-based radiomics in bone tumors: Technical advances and clinical application

Radiomics is the extraction of predefined mathematic features from medical images for predicting variables of clinical interest. Recent research has demonstrated that radiomics can be processed by artificial intelligence algorithms to reveal complex patterns and trends for diagnosis, and prediction of prognosis and response to treatment modalities in various types of cancer. Artificial intelligence tools can utilize radiological images to solve next-generation issues in clinical decision making. Bone tumors can be classified as primary and secondary (metastatic) tumors. Osteosarcoma, Ewing sarcoma, and chondrosarcoma are the dominating primary tumors of bone. The development of bone tumor model systems and relevant research, and the assessment of novel treatment methods are ongoing to improve clinical outcomes, notably for patients with metastases. Artificial intelligence and radiomics have been utilized in almost full spectrum of clinical care of bone tumors. Radiomics models have achieved excellent performance in the diagnosis and grading of bone tumors. Furthermore, the models enable to predict overall survival, metastases, and recurrence. Radiomics features have exhibited promise in assisting therapeutic planning and evaluation, especially neoadjuvant chemotherapy. This review provides an overview of the evolution and opportunities for artificial intelligence in imaging, with a focus on hand-crafted features and deep learning-based radiomics approaches. We summarize the current application of artificial intelligence-based radiomics both in primary and metastatic bone tumors, and discuss the limitations and future opportunities of artificial intelligence-based radiomics in this field. In the era of personalized medicine, our in-depth understanding of emerging artificial intelligence-based radiomics approaches will bring innovative solutions to bone tumors and achieve clinical application.

Diagnostic value of machine learning-based computed tomography texture analysis for differentiating multiple myeloma from osteolytic metastatic bone lesions in the peripheral skeleton

OBJECTIVES

To report the diagnostic performance of machine learning-based CT texture analysis for differentiating multiple myeloma from osteolytic metastatic bone lesions in the peripheral skeleton.

METHODS

We retrospectively evaluated 172 patients with multiple myeloma (n = 70) and osteolytic metastatic bone lesions (n = 102) in the peripheral skeleton. Two radiologists individually used two-dimensional manual segmentation to extract texture features from non-contrast CT. In total, 762 radiomic features were extracted. Dimension reduction was performed in three stages: inter-observer agreement analysis, collinearity analysis, and feature selection. Data were randomly divided into training (n = 120) and test (n = 52) groups. Eight machine learning algorithms were used for model development. The primary performance metrics were the area under the receiver operating characteristic curve and accuracy.

RESULTS

In total, 476 of the 762 texture features demonstrated excellent interobserver agreement. The number of features was reduced to 22 after excluding those with strong collinearity. Of these features, six were included in the machine learning algorithms using the wrapper-based classifier-specific technique. When all eight machine learning algorithms were considered for differentiating multiple myeloma from osteolytic metastatic bone lesions in the peripheral skeleton, the area under the receiver operating characteristic curve and accuracy were 0.776-0.932 and 78.8-92.3%, respectively. The k-nearest neighbors model performed the best, with the area under the receiver operating characteristic curve and accuracy values of 0.902 and 92.3%, respectively.

CONCLUSION

Machine learning-based CT texture analysis is a promising method for discriminating multiple myeloma from osteolytic metastatic bone lesions.

MRI radiomics-based machine learning for classification of deep-seated lipoma and atypical lipomatous tumor of the extremities

PURPOSE

To determine diagnostic performance of MRI radiomics-based machine learning for classification of deep-seated lipoma and atypical lipomatous tumor (ALT) of the extremities.

MATERIAL AND METHODS

This retrospective study was performed at three tertiary sarcoma centers and included 150 patients with surgically treated and histology-proven lesions. The training-validation cohort consisted of 114 patients from centers 1 and 2 (n = 64 lipoma, n = 50 ALT). The external test cohort consisted of 36 patients from center 3 (n = 24 lipoma, n = 12 ALT). 3D segmentation was manually performed on T1- and T2-weighted MRI. After extraction and selection of radiomic features, three machine learning classifiers were trained and validated using nested fivefold cross-validation. The best-performing classifier according to previous analysis was evaluated and compared to an experienced musculoskeletal radiologist in the external test cohort.

RESULTS

Eight features passed feature selection and were incorporated into the machine learning models. After training and validation (74% ROC-AUC), the best-performing classifier (Random Forest) showed 92% sensitivity and 33% specificity in the external test cohort with no statistical difference compared to the radiologist (p = 0.474).

CONCLUSION

MRI radiomics-based machine learning may classify deep-seated lipoma and ALT of the extremities with high sensitivity and negative predictive value, thus potentially serving as a non-invasive screening tool to reduce unnecessary referral to tertiary tumor centers.

The diagnostic value of magnetic resonance imaging-based texture analysis in differentiating enchondroma and chondrosarcoma

OBJECTIVE

To assess the diagnostic performance of MRI-based texture analysis for differentiating enchondromas and chondrosarcomas, especially on fat-suppressed proton density (FS-PD) images.

MATERIALS AND METHODS

The whole tumor volumes of 23 chondrosarcomas and 24 enchondromas were manually segmented on both FS-PD and T1-weighted images. A total of 861 radiomic features were extracted. SelectKBest was used to select the features. The data were randomly split into training (n = 36) and test (n = 10) for T1-weighted and training (n = 37) and test (n = 10) for FS-PD datasets. Fivefold cross-validation was performed. Fifteen machine learning models were created using the training set. The best models for T1-weighted, FS-PD, and T1-weighted + FS-PD images were selected in terms of accuracy and area under the curve (AUC).

RESULTS

There were 7 men and 16 women in the chondrosarcoma group (mean ± standard deviation age, 45.65 ± 11.24) and 7 men and 17 women in the enchondroma group (mean ± standard deviation age, 46.17 ± 11.79). Naive Bayes was the best model for accuracy and AUC for T1-weighted images (AUC = 0.76, accuracy = 80%, recall = 80%, precision = 80%, F1 score = 80%). The best model for FS-PD images was the K neighbors classifier for accuracy and AUC (AUC = 1.00, accuracy = 80%, recall = 80%, precision = 100%, F1 score = 89%). The best model for T1-weighted + FS-PD images was logistic regression for accuracy and AUC (AUC = 0.84, accuracy = 80%, recall = 60%, precision = 100%, F1 score = 75%).

CONCLUSION

MRI-based machine learning models have promising results in the discrimination of enchondroma and chondrosarcoma based on radiomic features obtained from both FS-PD and T1-weighted images.

The use of radiomics and machine learning for the differentiation of chondrosarcoma from enchondroma

PURPOSE

To construct and compare machine learning models for differentiating chondrosarcoma from enchondroma using radiomic features from T1 and fat suppressed Proton density (PD) magnetic resonance imaging (MRI).

METHODS

Eighty-eight patients (57 with enchondroma, 31 with chondrosarcoma) were retrospectively included. Histogram matching and N4ITK MRI bias correction filters were applied. An experienced musculoskeletal radiologist and a senior resident in radiology performed manual segmentation. Voxel sizes were resampled. Laplacian of Gaussian filter and wavelet-based features were used. One thousand eight hundred eighty-eight features were obtained for each patient, with 944 from T1 and 944 from PD images. Sixty-four unstable features were removed. Seven machine learning models were used for classification.

RESULTS

Classification with all features showed neural network was the best model for both readers' datasets with area under the curve (AUC), classification accuracy (CA), and F1 score of 0.979, 0.984; 0.920, 0.932; and 0.889, 0.903, respectively. Four features, including one common to both readers, were selected using fast correlation based filter. The best performing models with selected features were gradient boosting for Fatih Erdem's dataset and neural network for Gülen Demirpolat's dataset with AUC, CA, and F1 score of 0.990, 0.979; 0.943, 0.955; 0.921, 0.933, respectively. Neural Network was the second-best model for FE's dataset based on AUC (0.984).

CONCLUSION

Using pathology as a gold standard, this study defined and compared seven well-performing models to distinguish enchondromas from chondrosarcomas and provided radiomic feature stability and reproducibility among the readers.

Application of Machine Learning for Differentiating Bone Malignancy on Imaging: A Systematic Review

An accurate diagnosis of bone tumours on imaging is crucial for appropriate and successful treatment. The advent of Artificial intelligence (AI) and machine learning methods to characterize and assess bone tumours on various imaging modalities may assist in the diagnostic workflow. The purpose of this review article is to summarise the most recent evidence for AI techniques using imaging for differentiating benign from malignant lesions, the characterization of various malignant bone lesions, and their potential clinical application. A systematic search through electronic databases (PubMed, MEDLINE, Web of Science, and clinicaltrials.gov) was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. A total of 34 articles were retrieved from the databases and the key findings were compiled and summarised. A total of 34 articles reported the use of AI techniques to distinguish between benign vs. malignant bone lesions, of which 12 (35.3%) focused on radiographs, 12 (35.3%) on MRI, 5 (14.7%) on CT and 5 (14.7%) on PET/CT. The overall reported accuracy, sensitivity, and specificity of AI in distinguishing between benign vs. malignant bone lesions ranges from 0.44–0.99, 0.63–1.00, and 0.73–0.96, respectively, with AUCs of 0.73–0.96. In conclusion, the use of AI to discriminate bone lesions on imaging has achieved a relatively good performance in various imaging modalities, with high sensitivity, specificity, and accuracy for distinguishing between benign vs. malignant lesions in several cohort studies. However, further research is necessary to test the clinical performance of these algorithms before they can be facilitated and integrated into routine clinical practice.

Classification of Chondrosarcoma: From Characteristic to Challenging Imaging Findings

Chondrosarcomas can be classified into various forms according to the presence or absence of a precursor lesion, location, and histological subtype. The new 2020 World Health Organization (WHO) Classification of Tumors of Soft Tissue and Bone classifies chondrogenic bone tumors as benign, intermediate (locally aggressive), or malignant, and separates atypical cartilaginous tumors (ACTs) and chondrosarcoma grade 1 (CS1) as intermediate and malignant tumors. respectively. Furthermore, the classification categorizes chondrosarcomas (including ACT) into eight subtypes: central conventional (grade 1 vs. 2–3), secondary peripheral (grade 1 vs. 2–3), periosteal, dedifferentiated, mesenchymal, and clear cell chondrosarcoma. Most chondrosarcomas are the low-grade, primary central conventional type. The rarer subtypes include clear cell, mesenchymal, and dedifferentiated chondrosarcomas. Comprehensive analysis of the characteristic imaging findings can help differentiate various forms of chondrosarcomas. However, distinguishing low-grade chondrosarcomas from enchondromas or high-grade chondrosarcomas is radiologically and histopathologically challenging, even for experienced radiologists and pathologists.

Texture Feature-Based Machine Learning Classification on MRI Image for Sepsis-Associated Encephalopathy Detection: A Pilot Study

Objective

The objective of this study was to assess the performance of combining MRI-based texture analysis with machine learning for differentiating sepsis-associated encephalopathy (SAE) from sepsis alone.

Method

Sixty-six MRI-T1WI images of an SAE patient and 125 images of patients with sepsis alone were collected. Frontal lobe, brain stem, hippocampus, and amygdala were selected as regions of interest (ROIs). 279 texture features of each ROI were obtained using MaZda software. After the dimension reduction, 30 highly discriminative features of each ROI were adopted to differentiate SAE from sepsis alone using the CatBoost model.

Results

The classification models of frontal, brain stem, hippocampus, and amygdala were constructed. The classification accuracy was above 0.83, and the area under the curve (AUC) exceeded 0.90 in the validation set.

Conclusion

The texture features differed between SAE patients and patients with sepsis alone in different anatomical locations, suggesting that MRI-based texture analysis with machine learning might be helpful in differentiating SAE from sepsis alone.

A systematic review of radiomics in chondrosarcoma: assessment of study quality and clinical value needs handy tools

OBJECTIVE

To evaluate the study quality and clinical value of radiomics studies on chondrosarcoma.

METHODS

PubMed, Embase, Web of Science, China National Knowledge Infrastructure, and Wanfang Data were searched for articles on radiomics for evaluating chondrosarcoma as of January 31, 2022. The study quality was assessed according to Radiomics Quality Score (RQS), Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) checklist, Image Biomarker Standardization Initiative (IBSI) guideline, and modified Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool. The level of evidence supporting clinical use of radiomics on chondrosarcoma differential diagnosis was determined based on meta-analyses.

RESULTS

Twelve articles were included. The median RQS was 10.5 (range, -3 to 15), with an adherence rate of 36%. The adherence rate was extremely low in domains of high-level evidence (0%), open science and data (17%), and imaging and segmentation (35%). The adherence rate of the TRIPOD checklist was 61%, and low for section of title and abstract (13%), introduction (42%), and results (56%). The reporting rate of pre-processing steps according to the IBSI guideline was 60%. The risk of bias and concern of application were mainly related to the index test. The meta-analysis on differential diagnosis of enchondromas vs. chondrosarcomas showed a diagnostic odds ratio of 43.90 (95% confidential interval, 25.33-76.10), which was rated as weak evidence.

CONCLUSIONS

The current scientific and reporting quality of radiomics studies on chondrosarcoma was insufficient. Radiomics has potential in facilitating the optimization of operation decision-making in chondrosarcoma.

KEY POINTS

• Among radiomics studies on chondrosarcoma, although differential diagnostic models showed promising performance, only pieces of weak level of evidence were reached with insufficient study quality. • Since the RQS rating, the TRIPOD checklist, and the IBSI guideline have largely overlapped with each other, it is necessary to establish one widely acceptable methodological and reporting guideline for radiomics research. • The TRIPOD model typing, the phase classification of image mining studies, and the level of evidence category are useful tools to assess the gap between academic research and clinical application, although their modifications for radiomics studies are needed.

A Computed Tomography Radiomics Nomogram in Differentiating Chordoma From Giant Cell Tumor in the Axial Skeleton

OBJECTIVE

The aim of the study is to develop and validate a computed tomography (CT) radiomics nomogram for preoperatively differentiating chordoma from giant cell tumor (GCT) in the axial skeleton.

METHODS

Seventy-three chordomas and 38 GCTs in axial skeleton were retrospectively included and were divided into a training cohort (n = 63) and a test cohort (n = 48). The radiomics features were extracted from CT images. A radiomics signature was developed by using the least absolute shrinkage and selection operator model, and a radiomics score (Rad-score) was acquired. By combining the Rad-score with independent clinical risk factors using multivariate logistic regression model, a radiomics nomogram was established. Calibration and receiver operator characteristic curves were used to assess the performance of the nomogram.

RESULTS

Five features were selected to construct the radiomics signature. The radiomics signature showed favorable discrimination in the training cohort (area under the curve [AUC], 0.860; 95% confidence interval [CI], 0.760-0.960) and the test cohort (AUC, 0.830; 95% CI, 0.710-0.950). Age and location were the independent clinical factors. The radiomics nomogram combining the Rad-score with independent clinical factors showed good discrimination capability in the training cohort (AUC, 0.930; 95% CI, 0.880-0.990) and the test cohort (AUC, 0.980; 95% CI, 0.940-1.000) and outperformed the radiomics signature (z = 2.768, P = 0.006) in the test cohort.

CONCLUSIONS

The CT radiomics nomogram shows good predictive efficacy in differentiating chordoma from GCT in the axial skeleton, which might facilitate clinical decision making.

Development and validation of a MRI-based combined radiomics nomogram for differentiation in chondrosarcoma

Objective

This study aims to develop and validate the performance of an unenhanced magnetic resonance imaging (MRI)-based combined radiomics nomogram for discrimination between low-grade and high-grade in chondrosarcoma.

Methods

A total of 102 patients with 44 in low-grade and 58 in high-grade chondrosarcoma were enrolled and divided into training set (n=72) and validation set (n=30) with a 7:3 ratio in this retrospective study. The demographics and unenhanced MRI imaging characteristics of the patients were evaluated to develop a clinic-radiological factors model. Radiomics features were extracted from T1-weighted (T1WI) images to construct radiomics signature and calculate radiomics score (Rad-score). According to multivariate logistic regression analysis, a combined radiomics nomogram based on MRI was constructed by integrating radiomics signature and independent clinic-radiological features. The performance of the combined radiomics nomogram was evaluated in terms of calibration, discrimination, and clinical usefulness.

Results

Using multivariate logistic regression analysis, only one clinic-radiological feature (marrow edema OR=0.29, 95% CI=0.11-0.76, P=0.012) was found to be independent predictors of differentiation in chondrosarcoma. Combined with the above clinic-radiological predictor and the radiomics signature constructed by LASSO [least absolute shrinkage and selection operator], a combined radiomics nomogram based on MRI was constructed, and its predictive performance was better than that of clinic-radiological factors model and radiomics signature, with the AUC [area under the curve] of the training set and the validation set were 0.78 (95%CI =0.67-0.89) and 0.77 (95%CI =0.59-0.94), respectively. DCA [decision curve analysis] showed that combined radiomics nomogram has potential clinical application value.

Conclusion

The MRI-based combined radiomics nomogram is a noninvasive preoperative prediction tool that combines clinic-radiological feature and radiomics signature and shows good predictive effect in distinguishing low-grade and high-grade bone chondrosarcoma, which may help clinicians to make accurate treatment plans.

Radiomics and Deep Learning for Disease Detection in Musculoskeletal Radiology: An Overview of Novel MRI- and CT-Based Approaches

ABSTRACT

Radiomics and machine learning-based methods offer exciting opportunities for improving diagnostic performance and efficiency in musculoskeletal radiology for various tasks, including acute injuries, chronic conditions, spinal abnormalities, and neoplasms. While early radiomics-based methods were often limited to a smaller number of higher-order image feature extractions, applying machine learning-based analytic models, multifactorial correlations, and classifiers now permits big data processing and testing thousands of features to identify relevant markers. A growing number of novel deep learning-based methods describe magnetic resonance imaging- and computed tomography-based algorithms for diagnosing anterior cruciate ligament tears, meniscus tears, articular cartilage defects, rotator cuff tears, fractures, metastatic skeletal disease, and soft tissue tumors. Initial radiomics and deep learning techniques have focused on binary detection tasks, such as determining the presence or absence of a single abnormality and differentiation of benign versus malignant. Newer-generation algorithms aim to include practically relevant multiclass characterization of detected abnormalities, such as typing and malignancy grading of neoplasms. So-called delta-radiomics assess tumor features before and after treatment, with temporal changes of radiomics features serving as surrogate markers for tumor responses to treatment. New approaches also predict treatment success rates, surgical resection completeness, and recurrence risk. Practice-relevant goals for the next generation of algorithms include diagnostic whole-organ and advanced classification capabilities. Important research objectives to fill current knowledge gaps include well-designed research studies to understand how diagnostic performances and suggested efficiency gains of isolated research settings translate into routine daily clinical practice. This article summarizes current radiomics- and machine learning-based magnetic resonance imaging and computed tomography approaches for musculoskeletal disease detection and offers a perspective on future goals and objectives.

3D vs. 2D MRI radiomics in skeletal Ewing sarcoma: Feature reproducibility and preliminary machine learning analysis on neoadjuvant chemotherapy response prediction

OBJECTIVE

The extent of response to neoadjuvant chemotherapy predicts survival in Ewing sarcoma. This study focuses on MRI radiomics of skeletal Ewing sarcoma and aims to investigate feature reproducibility and machine learning prediction of response to neoadjuvant chemotherapy.

MATERIALS AND METHODS

This retrospective study included thirty patients with biopsy-proven skeletal Ewing sarcoma, who were treated with neoadjuvant chemotherapy before surgery at two tertiary sarcoma centres. 7 patients were poor responders and 23 were good responders based on pathological assessment of the surgical specimen. On pre-treatment T1-weighted and T2-weighted MRI, 2D and 3D tumour segmentations were manually performed. Features were extracted from original and wavelet-transformed images. Feature reproducibility was assessed through small geometrical transformations of the regions of interest mimicking multiple manual delineations, and intraclass correlation coefficient >0.75 defined feature reproducibility. Feature selection also consisted of collinearity and significance analysis. After class balancing in the training cohort, three machine learning classifiers were trained and tested on unseen data using hold-out cross-validation.

RESULTS

1303 (77%) 3D and 620 (65%) 2D radiomic features were reproducible. 4 3D and 4 2D features passed feature selection. Logistic regression built upon 3D features achieved the best performance with 85% accuracy (AUC=0.9) in predicting response to neoadjuvant chemotherapy.

CONCLUSION

Compared to 2D approach, 3D MRI radiomics of Ewing sarcoma had superior reproducibility and higher accuracy in predicting response to neoadjuvant chemotherapy, particularly when using logistic regression classifier.

Primary bone tumor detection and classification in full-field bone radiographs via YOLO deep learning model

OBJECTIVES

Automatic bone lesions detection and classifications present a critical challenge and are essential to support radiologists in making an accurate diagnosis of bone lesions. In this paper, we aimed to develop a novel deep learning model called You Only Look Once (YOLO) to handle detecting and classifying bone lesions on full-field radiographs with limited manual intervention.

METHODS

In this retrospective study, we used 1085 bone tumor radiographs and 345 normal bone radiographs from two centers between January 2009 and December 2020 to train and test our YOLO deep learning (DL) model. The trained model detected bone lesions and then classified these radiographs into normal, benign, intermediate, or malignant types. The intersection over union (IoU) was used to assess the model's performance in the detection task. Confusion matrices and Cohen's kappa scores were used for evaluating classification performance. Two radiologists compared diagnostic performance with the trained model using the external validation set.

RESULTS

In the detection task, the model achieved accuracies of 86.36% and 85.37% in the internal and external validation sets, respectively. In the DL model, radiologist 1 and radiologist 2 achieved Cohen's kappa scores of 0.8187, 0.7927, and 0.9077 for four-way classification in the external validation set, respectively. The YOLO DL model illustrated a significantly higher accuracy for intermediate bone tumor classification than radiologist 1 (95.73% vs 88.08%, p = 0.004).

CONCLUSIONS

The developed YOLO DL model could be used to assist radiologists at all stages of bone lesion detection and classification in full-field bone radiographs.

KEY POINTS

• YOLO DL model can automatically detect bone neoplasms from full-field radiographs in one shot and then simultaneously classify radiographs into normal, benign, intermediate, or malignant. • The dataset used in this retrospective study includes normal bone radiographs. • YOLO can detect even some challenging cases with small volumes.

Automated data preparation for in vivo tumor characterization with machine learning

Background

This study proposes machine learning-driven data preparation (MLDP) for optimal data preparation (DP) prior to building prediction models for cancer cohorts.

Methods

A collection of well-established DP methods were incorporated for building the DP pipelines for various clinical cohorts prior to machine learning. Evolutionary algorithm principles combined with hyperparameter optimization were employed to iteratively select the best fitting subset of data preparation algorithms for the given dataset. The proposed method was validated for glioma and prostate single center cohorts by 100-fold Monte Carlo (MC) cross-validation scheme with 80-20% training-validation split ratio. In addition, a dual-center diffuse large B-cell lymphoma (DLBCL) cohort was utilized with Center 1 as training and Center 2 as independent validation datasets to predict cohort-specific clinical endpoints. Five machine learning (ML) classifiers were employed for building prediction models across all analyzed cohorts. Predictive performance was estimated by confusion matrix analytics over the validation sets of each cohort. The performance of each model with and without MLDP, as well as with manually-defined DP were compared in each of the four cohorts.

Results

Sixteen of twenty established predictive models demonstrated area under the receiver operator characteristics curve (AUC) performance increase utilizing the MLDP. The MLDP resulted in the highest performance increase for random forest (RF) (+0.16 AUC) and support vector machine (SVM) (+0.13 AUC) model schemes for predicting 36-months survival in the glioma cohort. Single center cohorts resulted in complex (6-7 DP steps) DP pipelines, with a high occurrence of outlier detection, feature selection and synthetic majority oversampling technique (SMOTE). In contrast, the optimal DP pipeline for the dual-center DLBCL cohort only included outlier detection and SMOTE DP steps.

Conclusions

This study demonstrates that data preparation prior to ML prediction model building in cancer cohorts shall be ML-driven itself, yielding optimal prediction models in both single and multi-centric settings.

Applications of machine learning for imaging-driven diagnosis of musculoskeletal malignancies-a scoping review

Musculoskeletal malignancies are a rare type of cancer. Consequently, sufficient imaging data for machine learning (ML) applications is difficult to obtain. The main purpose of this review was to investigate whether ML is already having an impact on imaging-driven diagnosis of musculoskeletal malignancies and what the respective reasons for this might be. A scoping review was conducted by a radiologist, an orthopaedic surgeon and a data scientist to identify suitable articles based on the PRISMA statement. Studies meeting the following criteria were included: primary malignant musculoskeletal tumours, machine/deep learning application, imaging data or data retrieved from images, human/preclinical, English language and original research. Initially, 480 articles were found and 38 met the eligibility criteria. Several continuous and discrete parameters related to publication, patient distribution, tumour specificities, ML methods, data and metrics were extracted from the final articles. For the synthesis, diagnosis-oriented studies were further examined by retrieving the number of patients and labels and metric scores. No significant correlations between metrics and mean number of samples were found. Several studies presented that ML could support imaging-driven diagnosis of musculoskeletal malignancies in distinct cases. However, data quality and quantity must be increased to achieve clinically relevant results. Compared to the experience of an expert radiologist, the studies used small datasets and mostly included only one type of data. Key to critical advancement of ML models for rare diseases such as musculoskeletal malignancies is a systematic, structured data collection and the establishment of (inter)national networks to obtain substantial datasets in the future. KEY POINTS: • Machine learning does not yet significantly impact imaging-driven diagnosis for musculoskeletal malignancies compared to other disciplines such as lung, breast or CNS cancer. • Research in the area of musculoskeletal tumour imaging and machine learning is still very limited. • Machine learning in musculoskeletal tumour imaging is impeded by insufficient availability of data and rarity of the disease.

Diffusion-weighted MRI radiomics of spine bone tumors: feature stability and machine learning-based classification performance

PURPOSE

To evaluate stability and machine learning-based classification performance of radiomic features of spine bone tumors using diffusion- and T2-weighted magnetic resonance imaging (MRI).

MATERIAL AND METHODS

This retrospective study included 101 patients with histology-proven spine bone tumor (22 benign; 38 primary malignant; 41 metastatic). All tumor volumes were manually segmented on morphologic T2-weighted sequences. The same region of interest (ROI) was used to perform radiomic analysis on ADC map. A total of 1702 radiomic features was considered. Feature stability was assessed through small geometrical transformations of the ROIs mimicking multiple manual delineations. Intraclass correlation coefficient (ICC) quantified feature stability. Feature selection consisted of stability-based (ICC > 0.75) and significance-based selections (ranking features by decreasing Mann-Whitney p-value). Class balancing was performed to oversample the minority (i.e., benign) class. Selected features were used to train and test a support vector machine (SVM) to discriminate benign from malignant spine tumors using tenfold cross-validation.

RESULTS

A total of 76.4% radiomic features were stable. The quality metrics for the SVM were evaluated as a function of the number of selected features. The radiomic model with the best performance and the lowest number of features for classifying tumor types included 8 features. The metrics were 78% sensitivity, 68% specificity, 76% accuracy and AUC 0.78.

CONCLUSION

SVM classifiers based on radiomic features extracted from T2- and diffusion-weighted imaging with ADC map are promising for classification of spine bone tumors. Radiomic features of spine bone tumors show good reproducibility rates.

Radiomics of Musculoskeletal Sarcomas: A Narrative Review

Bone and soft-tissue primary malignant tumors or sarcomas are a large, diverse group of mesenchymal-derived malignancies. They represent a model for intra- and intertumoral heterogeneities, making them particularly suitable for radiomics analyses. Radiomic features offer information on cancer phenotype as well as the tumor microenvironment which, combined with other pertinent data such as genomics and proteomics and correlated with outcomes data, can produce accurate, robust, evidence-based, clinical-decision support systems. Our purpose in this narrative review is to offer an overview of radiomics studies dealing with Magnetic Resonance Imaging (MRI)-based radiomics models of bone and soft-tissue sarcomas that could help distinguish different histotypes, low-grade from high-grade sarcomas, predict response to multimodality therapy, and thus better tailor patients’ treatments and finally improve their survivals. Although showing promising results, interobserver segmentation variability, feature reproducibility, and model validation are three main challenges of radiomics that need to be addressed in order to translate radiomics studies to clinical applications. These efforts, together with a better knowledge and application of the “Radiomics Quality Score” and Image Biomarker Standardization Initiative reporting guidelines, could improve the quality of sarcoma radiomics studies and facilitate radiomics towards clinical translation.

Sarcopenia: imaging assessment and clinical application

Sarcopenia is a progressive, generalized skeletal muscle disorder characterized by reduction of muscle mass and strength. It is associated with increased adverse outcomes including falls, fractures, physical disability, and mortality, particularly, in elderly patients. Nowadays, sarcopenia has become a specific imaging biomarker able to predict clinical outcomes of patients. Muscle fibre reduction has shown to be an unfavourable pre-operative predictive factor in patients with cancer, and is associated with worse clinical outcomes in terms of postoperative complications, morbidity, mortality, and lower tolerance of chemoradiation therapy. Several imaging modalities, including dual-energy X-ray absorptiometry, CT, MRI, and US can be used to estimate muscle mass and quality to reach the diagnosis of sarcopenia. This article reviews the clinical implications of sarcopenia, how this condition can be assessed through different imaging modalities, and future perspectives of imaging of sarcopenia.

Development and evaluation of machine learning models based on X-ray radiomics for the classification and differentiation of malignant and benign bone tumors

OBJECTIVES

To develop and validate machine learning models to distinguish between benign and malignant bone lesions and compare the performance to radiologists.

METHODS

In 880 patients (age 33.1 ± 19.4 years, 395 women) diagnosed with malignant (n = 213, 24.2%) or benign (n = 667, 75.8%) primary bone tumors, preoperative radiographs were obtained, and the diagnosis was established using histopathology. Data was split 70%/15%/15% for training, validation, and internal testing. Additionally, 96 patients from another institution were obtained for external testing. Machine learning models were developed and validated using radiomic features and demographic information. The performance of each model was evaluated on the test sets for accuracy, area under the curve (AUC) from receiver operating characteristics, sensitivity, and specificity. For comparison, the external test set was evaluated by two radiology residents and two radiologists who specialized in musculoskeletal tumor imaging.

RESULTS

The best machine learning model was based on an artificial neural network (ANN) combining both radiomic and demographic information achieving 80% and 75% accuracy at 75% and 90% sensitivity with 0.79 and 0.90 AUC on the internal and external test set, respectively. In comparison, the radiology residents achieved 71% and 65% accuracy at 61% and 35% sensitivity while the radiologists specialized in musculoskeletal tumor imaging achieved an 84% and 83% accuracy at 90% and 81% sensitivity, respectively.

CONCLUSIONS

An ANN combining radiomic features and demographic information showed the best performance in distinguishing between benign and malignant bone lesions. The model showed lower accuracy compared to specialized radiologists, while accuracy was higher or similar compared to residents.

KEY POINTS

• The developed machine learning model could differentiate benign from malignant bone tumors using radiography with an AUC of 0.90 on the external test set. • Machine learning models that used radiomic features or demographic information alone performed worse than those that used both radiomic features and demographic information as input, highlighting the importance of building comprehensive machine learning models. • An artificial neural network that combined both radiomic and demographic information achieved the best performance and its performance was compared to radiology readers on an external test set.

MRI radiomics-based machine learning classification of atypical cartilaginous tumour and grade II chondrosarcoma of long bones

Background

Atypical cartilaginous tumour (ACT) and grade II chondrosarcoma (CS2) of long bones are respectively managed with watchful waiting or curettage and wide resection. Preoperatively, imaging diagnosis can be challenging due to interobserver variability and biopsy suffers from sample errors. The aim of this study is to determine diagnostic performance of MRI radiomics-based machine learning in differentiating ACT from CS2 of long bones.

Methods

One-hundred-fifty-eight patients with surgically treated and histology-proven cartilaginous bone tumours were retrospectively included at two tertiary bone tumour centres. The training cohort consisted of 93 MRI scans from centre 1 (n=74 ACT; n=19 CS2). The external test cohort consisted of 65 MRI scans from centre 2 (n=45 ACT; n=20 CS2). Bidimensional segmentation was performed on T1-weighted MRI. Radiomic features were extracted. After dimensionality reduction and class balancing in centre 1, a machine-learning classifier (Extra Trees Classifier) was tuned on the training cohort using 10-fold cross-validation and tested on the external test cohort. In centre 2, its performance was compared with an experienced musculoskeletal oncology radiologist using McNemar's test.

Findings

After tuning on the training cohort (AUC=0.88), the machine-learning classifier had 92% accuracy (60/65, AUC=0.94) in identifying the lesions in the external test cohort. Its accuracies in correctly classifying ACT and CS2 were 98% (44/45) and 80% (16/20), respectively. The radiologist had 98% accuracy (64/65) with no difference compared to the classifier (p=0.134).

Interpretation

Machine learning showed high accuracy in classifying ACT and CS2 of long bones based on MRI radiomic features.

Funding

ESSR Young Researchers Grant.

Whole-body MRI radiomics model to predict relapsed/refractory Hodgkin Lymphoma: A preliminary study

PURPOSE

A strong prognostic score that enables a stratification of newly diagnosed Hodgkin Lymphoma (HL) to identify patients at high risk of refractory/relapsed disease is still needed. Our aim was to investigate the potential value of a radiomics analysis pipeline from whole-body MRI (WB-MRI) exams for clinical outcome prediction in patients with HL.

MATERIALS AND METHODS

Index lesions from baseline WB-MRIs of 40 patients (22 females; mean age 31.7 ± 11.4 years) with newly diagnosed HL treated by ABVD chemotherapy regimen were manually segmented on T1-weighted, STIR, and DWI images for texture analysis feature extraction. A machine learning approach based on the Extra Trees classifier and incorporating clinical variables, ¹⁸F-FDG-PET/CT-derived metabolic tumor volume, and WB-MRI radiomics features was tested using cross-validation to predict refractory/relapsed disease.

RESULTS

Relapsed disease was observed in 10/40 patients (25%), two of whom died due to progression of disease and graft versus host disease, while eight reached the complete remission. In total, 1403 clinical and radiomics features were extracted, of which 11 clinical variables and 171 radiomics parameters from both original and filtered images were selected. The 3 best performing Extra Trees classifier models obtained an equivalent highest mean accuracy of 0.78 and standard deviation of 0.09, with a mean AUC of 0.82 and standard deviation of 0.08.

CONCLUSIONS

Our preliminary results demonstrate that a combined machine learning and texture analysis model to predict refractory/relapsed HL on WB-MRI exams is feasible and may help in the clinical outcome prediction in HL patients.

Radiomics and Machine Learning Can Differentiate Transient Osteoporosis from Avascular Necrosis of the Hip

Differentiation between transient osteoporosis (TOH) and avascular necrosis (AVN) of the hip is a longstanding challenge in musculoskeletal radiology. The purpose of this study was to utilize MRI-based radiomics and machine learning (ML) for accurate differentiation between the two entities. A total of 109 hips with TOH and 104 hips with AVN were retrospectively included. Femoral heads and necks with segmented radiomics features were extracted. Three ML classifiers (XGboost, CatBoost and SVM) using 38 relevant radiomics features were trained on 70% and validated on 30% of the dataset. ML performance was compared to two musculoskeletal radiologists, a general radiologist and two radiology residents. XGboost achieved the best performance with an area under the curve (AUC) of 93.7% (95% CI from 87.7 to 99.8%) among ML models. MSK radiologists achieved an AUC of 90.6% (95% CI from 86.7% to 94.5%) and 88.3% (95% CI from 84% to 92.7%), respectively, similar to residents. The general radiologist achieved an AUC of 84.5% (95% CI from 80% to 89%), significantly lower than of XGboost (p = 0.017). In conclusion, radiomics-based ML achieved a performance similar to MSK radiologists and significantly higher compared to general radiologists in differentiating between TOH and AVN.

Effects of Interobserver Variability on 2D and 3D CT- and MRI-Based Texture Feature Reproducibility of Cartilaginous Bone Tumors

This study aims to investigate the influence of interobserver manual segmentation variability on the reproducibility of 2D and 3D unenhanced computed tomography (CT)- and magnetic resonance imaging (MRI)-based texture analysis. Thirty patients with cartilaginous bone tumors (10 enchondromas, 10 atypical cartilaginous tumors, 10 chondrosarcomas) were retrospectively included. Three radiologists independently performed manual contour-focused segmentation on unenhanced CT and T1-weighted and T2-weighted MRI by drawing both a 2D region of interest (ROI) on the slice showing the largest tumor area and a 3D ROI including the whole tumor volume. Additionally, a marginal erosion was applied to both 2D and 3D segmentations to evaluate the influence of segmentation margins. A total of 783 and 1132 features were extracted from original and filtered 2D and 3D images, respectively. Intraclass correlation coefficient ≥ 0.75 defined feature stability. In 2D vs. 3D contour-focused segmentation, the rates of stable features were 74.71% vs. 86.57% (p < 0.001), 77.14% vs. 80.04% (p = 0.142), and 95.66% vs. 94.97% (p = 0.554) for CT and T1-weighted and T2-weighted images, respectively. Margin shrinkage did not improve 2D (p = 0.343) and performed worse than 3D (p < 0.001) contour-focused segmentation in terms of feature stability. In 2D vs. 3D contour-focused segmentation, matching stable features derived from CT and MRI were 65.8% vs. 68.7% (p = 0.191), and those derived from T1-weighted and T2-weighted images were 76.0% vs. 78.2% (p = 0.285). 2D and 3D radiomic features of cartilaginous bone tumors extracted from unenhanced CT and MRI are reproducible, although some degree of interobserver segmentation variability highlights the need for reliability analysis in future studies.

Vertebral MRI-based radiomics model to differentiate multiple myeloma from metastases: influence of features number on logistic regression model performance

OBJECTIVES

This study aimed to use the most frequent features to establish a vertebral MRI-based radiomics model that could differentiate multiple myeloma (MM) from metastases and compare the model performance with different features number.

METHODS

We retrospectively analyzed conventional MRI (T1WI and fat-suppression T2WI) of 103 MM patients and 138 patients with metastases. The feature selection process included four steps. The first three steps defined as conventional feature selection (CFS), carried out 50 times (ten times with 5-fold cross-validation), included variance threshold, SelectKBest, and least absolute shrinkage and selection operator. The most frequent fixed features were selected for modeling during the last step. The number of events per independent variable (EPV) is the number of patients in a smaller subgroup divided by the number of radiomics features considered in developing the prediction model. The EPV values considered were 5, 10, 15, and 20. Therefore, we constructed four models using the top 16, 8, 6, and 4 most frequent features, respectively. The models constructed with features selected by CFS were also compared.

RESULTS

The AUCs of 20EPV-Model, 15EPV-Model, and CSF-Model (AUC = 0.71, 0.81, and 0.78) were poor than 10EPV-Model (AUC = 0.84, p < 0.001). The AUC of 10EPV-Model was comparable with 5EPV-Model (AUC = 0.85, p = 0.480).

CONCLUSIONS

The radiomics model constructed with an appropriate small number of the most frequent features could well distinguish metastases from MM based on conventional vertebral MRI. Based on our results, we recommend following the 10 EPV as the rule of thumb for feature selection.

KEY POINTS

• The developed radiomics model could distinguish metastases from multiple myeloma based on conventional vertebral MRI. • An accurate model based on just a handful of the most frequent features could be constructed by utilizing multiple feature reduction techniques. • An event per independent variable value of 10 is recommended as a rule of thumb for modeling feature selection.

State of the Art in Artificial Intelligence and Radiomics in Hepatocellular Carcinoma

The most common liver malignancy is hepatocellular carcinoma (HCC), which is also associated with high mortality. Often HCC develops in a chronic liver disease setting, and early diagnosis as well as accurate screening of high-risk patients is crucial for appropriate and effective management of these patients. While imaging characteristics of HCC are well-defined in the diagnostic phase, challenging cases still occur, and current prognostic and predictive models are limited in their accuracy. Radiomics and machine learning (ML) offer new tools to address these issues and may lead to scientific breakthroughs with the potential to impact clinical practice and improve patient outcomes. In this review, we will present an overview of these technologies in the setting of HCC imaging across different modalities and a range of applications. These include lesion segmentation, diagnosis, prognostic modeling and prediction of treatment response. Finally, limitations preventing clinical application of radiomics and ML at the present time are discussed, together with necessary future developments to bring the field forward and outside of a purely academic endeavor.

CT and MRI radiomics of bone and soft-tissue sarcomas: a systematic review of reproducibility and validation strategies

Background

Feature reproducibility and model validation are two main challenges of radiomics. This study aims to systematically review radiomic feature reproducibility and predictive model validation strategies in studies dealing with CT and MRI radiomics of bone and soft-tissue sarcomas. The ultimate goal is to promote achieving a consensus on these aspects in radiomic workflows and facilitate clinical transferability.

Results

Out of 278 identified papers, forty-nine papers published between 2008 and 2020 were included. They dealt with radiomics of bone (n = 12) or soft-tissue (n = 37) tumors. Eighteen (37%) studies included a feature reproducibility analysis. Inter-/intra-reader segmentation variability was the theme of reproducibility analysis in 16 (33%) investigations, outnumbering the analyses focused on image acquisition or post-processing (n = 2, 4%). The intraclass correlation coefficient was the most commonly used statistical method to assess reproducibility, which ranged from 0.6 and 0.9. At least one machine learning validation technique was used for model development in 25 (51%) papers, and K-fold cross-validation was the most commonly employed. A clinical validation of the model was reported in 19 (39%) papers. It was performed using a separate dataset from the primary institution (i.e., internal validation) in 14 (29%) studies and an independent dataset related to different scanners or from another institution (i.e., independent validation) in 5 (10%) studies.

Conclusions

The issues of radiomic feature reproducibility and model validation varied largely among the studies dealing with musculoskeletal sarcomas and should be addressed in future investigations to bring the field of radiomics from a preclinical research area to the clinical stage.

CT radiomics-based machine learning classification of atypical cartilaginous tumours and appendicular chondrosarcomas

Background

Clinical management ranges from surveillance or curettage to wide resection for atypical to higher-grade cartilaginous tumours, respectively. Our aim was to investigate the performance of computed tomography (CT) radiomics-based machine learning for classification of atypical cartilaginous tumours and higher-grade chondrosarcomas of long bones.

Methods

One-hundred-twenty patients with histology-proven lesions were retrospectively included. The training cohort consisted of 84 CT scans from centre 1 (n=55 G1 or atypical cartilaginous tumours; n=29 G2-G4 chondrosarcomas). The external test cohort consisted of the CT component of 36 positron emission tomography-CT scans from centre 2 (n=16 G1 or atypical cartilaginous tumours; n=20 G2-G4 chondrosarcomas). Bidimensional segmentation was performed on preoperative CT. Radiomic features were extracted. After dimensionality reduction and class balancing in centre 1, the performance of a machine-learning classifier (LogitBoost) was assessed on the training cohort using 10-fold cross-validation and on the external test cohort. In centre 2, its performance was compared with preoperative biopsy and an experienced radiologist using McNemar's test.

Findings

The classifier had 81% (AUC=0.89) and 75% (AUC=0.78) accuracy in identifying the lesions in the training and external test cohorts, respectively. Specifically, its accuracy in classifying atypical cartilaginous tumours and higher-grade chondrosarcomas was 84% and 78% in the training cohort, and 81% and 70% in the external test cohort, respectively. Preoperative biopsy had 64% (AUC=0.66) accuracy (p=0.29). The radiologist had 81% accuracy (p=0.75).

Interpretation

Machine learning showed good accuracy in classifying atypical and higher-grade cartilaginous tumours of long bones based on preoperative CT radiomic features.

Funding

ESSR Young Researchers Grant.

An update in musculoskeletal tumors: from quantitative imaging to radiomics

In the last two decades, relevant progress has been made in the diagnosis of musculoskeletal tumors due to the development of new imaging tools, such as diffusion-weighted imaging, diffusion kurtosis imaging, magnetic resonance spectroscopy, and diffusion tensor imaging. Another important role has been played by the development of artificial intelligence software based on complex algorithms, which employ computing power in the detection of specific tumor types. The aim of this article is to report the most advanced imaging techniques focusing on their advantages in clinical practice.

Radiomic Machine Learning Classifiers in Spine Bone Tumors: A Multi-Software, Multi-Scanner Study

PURPOSE

Spinal lesion differential diagnosis remains challenging even in MRI. Radiomics and machine learning (ML) have proven useful even in absence of a standardized data mining pipeline. We aimed to assess ML diagnostic performance in spinal lesion differential diagnosis, employing radiomic data extracted by different software.

METHODS

Patients undergoing MRI for a vertebral lesion were retrospectively analyzed (n = 146, 67 males, 79 females; mean age 63 ± 16 years, range 8-89 years) and constituted the train (n = 100) and internal test cohorts (n = 46). Part of the latter had additional prior exams which constituted a multi-scanner, external test cohort (n = 35). Lesions were labeled as benign or malignant (2-label classification), and benign, primary malignant or metastases (3-label classification) for classification analyses. Features extracted via 3D Slicer heterogeneityCAD module (hCAD) and PyRadiomics were independently used to compare different combinations of feature selection methods and ML classifiers (n = 19).

RESULTS

In total, 90 and 1548 features were extracted by hCAD and PyRadiomics, respectively. The best feature selection method-ML algorithm combination was selected by 10 iterations of 10-fold cross-validation in the training data. For the 2-label classification ML obtained 94% accuracy in the internal test cohort, using hCAD data, and 86% in the external one. For the 3-label classification, PyRadiomics data allowed for 80% and 69% accuracy in the internal and external test sets, respectively.

CONCLUSIONS

MRI radiomics combined with ML may be useful in spinal lesion assessment. More robust pre-processing led to better consistency despite scanner and protocol heterogeneity.

Diagnostic Performance and Radiation Dose of the EOS System to Image Enchondromatosis: A Phantom Study

Background: Radiation doses and capability of EOS, conventional radiography (CR), and computed tomography (CT) to detect and measure enchondromas in a dedicated five-year-old anthropomorphic phantom were compared. Methods: To simulate enchondromas, minced pieces of chicken bone and cartilage were packed in conventional kitchen plastic foil to create ovoidal/rounded masses and randomly hung on the phantom. The phantom was imaged five times with CR, CT, and EOS, each time changing the number and position of inserts. All images were reviewed by a senior radiologist and a radiology resident. Results: EOS and CR detected all inserts in 4/5 cases (80%), while in one case 1/17 inserts was not seen. Excellent agreement of EOS with CR (88% reproducibility; bias = 14 mm; repeatability coefficient (CoR) 2.9; 95% CI from −2.8 to 3.1 mm; p = 0.5) and CT (81% reproducibility; bias = 15 mm; CoR 5.2; 95% CI from −5.5 to 5.2 mm; p = 0.7) was found. EOS showed 71% interobserver reproducibility (CoR 7.2; bias = 0.6 mm; 95% CI from −6.6 to 7.8 mm; p = 0.25). The EOS-Fast radiation dose was also significantly lower than the median radiation dose of CR (644.7 (599.4–651.97) mGy•cm2, p = 0.004). Conclusions: Low-dose EOS has the same capability as CR to detect and measure enchondroma-like inserts on a phantom and may be considered to monitor patients with multiple enchondromas.

Magnetic Resonance Imaging Role in the Differentiation Between Atypical Cartilaginous Tumors and High-Grade Chondrosarcoma: An Updated Systematic Review

Chondrosarcoma (CS) is a malignant tumor affecting the bones while atypical cartilaginous tumors (ACTs) are chondral tumors with moderate cellularity, mild atypia, and with myxoid changes and mild metastatic ability. Both can have one of the worst prognoses if not identified early enough. Magnetic resonance imaging (MRI) has been one of the modalities to detect such tumors and we aim to determine the common characteristic and features to be seen while screening for CS or ACTs. We conducted a systematic review of the previously published reports that investigated the diagnostic ability of MRI and the reported characteristics that can differentiate between ACTs and high-grade chondrosarcomas (HGCS). A comprehensive relevant database search was performed to include all the relevant studies. Among these studies, seven studies investigated the overall accuracy in the classification of the different chondroma types. Additionally, many studies reported the characteristic findings of each tumor according to the MRI results. These characteristics mainly included trapped fat, bone marrow edema, cortical damage, and soft-tissue expansion. Therefore, further attention should be given to these criteria for better assessment, differentiation, and favorable outcomes. MRI can efficiently identify some of the characteristics of both ACTs and HGCS. However, combining it with other radiological modalities may lead to a better differentiation. The detection of ACTs and HGCS lesions with MRI solely has been doubted before in the literature.