Radiomics in radiooncology - Challenging the medical physicist

Role of Radiomics-based Multiomics Panel in the Microenvironment and Prognosis of Hepatocellular Carcinoma

Hepatocellular carcinoma (HCC) is the most prevalent form of liver tumor, characterized by restricted therapeutic options and typically low long-term survival rates. Recently, immunotherapy has revolutionized HCC treatment, making the tumor microenvironment (TME) a research focus. Radiomics is increasingly crucial in HCC clinical decisions, offering advanced tools for TME characterization and prognosis assessment. Meanwhile, advancements in pathomics provide new insights into HCC's comprehensive traits and details. Advancements in genomics and transcriptomics enable the integration of radiomics and pathomics with genetic data to better understand HCC heterogeneity and its microenvironment, aiding prognostic assessments. This review provides a comprehensive overview of pivotal radiomics studies focused on TME prediction, underscoring the synergistic effects of integrating multiomics approaches for TME analysis and HCC outcome prediction. It critically examines the challenges and opportunities inherent in multiomics research, emphasizing its substantial significance in both research and clinical contexts.

ContraSurv: Enhancing Prognostic Assessment of Medical Images via Data-Efficient Weakly Supervised Contrastive Learning

Prognostic assessment remains a critical challenge in medical research, often limited by the lack of well-labeled data. In this work, we introduce ContraSurv, a weakly-supervised learning framework based on contrastive learning, designed to enhance prognostic predictions in 3D medical images. ContraSurv utilizes both the self-supervised information inherent in unlabeled data and the weakly-supervised cues present in censored data, refining its capacity to extract prognostic representations. For this purpose, we establish a Vision Transformer architecture optimized for our medical image datasets and introduce novel methodologies for both self-supervised and supervised contrastive learning for prognostic assessment. Additionally, we propose a specialized supervised contrastive loss function and introduce SurvMix, a novel data augmentation technique for survival analysis. Evaluations were conducted across three cancer types and two imaging modalities on three real-world datasets. The results confirmed the enhanced performance of ContraSurv over competing methods, particularly in data with a high censoring rate.

Robust machine learning challenge: An AIFM multicentric competition to spread knowledge, identify common pitfalls and recommend best practice

PURPOSE

A novel and unconventional approach to a machine learning challenge was designed to spread knowledge, identify robust methods and highlight potential pitfalls about machine learning within the Medical Physics community.

METHODS

A public dataset comprising 41 radiomic features and 535 patients was employed to assess the potential of radiomics in distinguishing between primary lung tumors and metastases. Each participant developed two classification models using: (i) all features (base model); (ii) only robust features (robust model). Both models were validated with cross-validation and on unseen data. The population stability index (PSI) was used as diagnostic metric for implementation issues. Performance was compared to reference. Base and robust models were compared in terms of performance and stability (coefficient of variation (CoV) of prediction probabilities).

RESULTS

PSI detected potential implementation errors in 70 % of models. The dataset exhibited strong imbalance. The average Gmean (i.e. an appropriate metric for imbalance) among all participants was 0.67 ± 0.01, significantly higher than reference Gmean of 0.50 ± 0.04. Robust models performances were slightly worse than base models (p < 0.05). Regarding stability, robust models exhibited lower median CoV on training set only.

CONCLUSION

AI4MP-Challenge models overperformed the reference, significantly improving the Gmean. Exclusion of less-robust features did not improve model robustness and it should be avoided when confounding effects are absent. Other methods, like harmonization or data augmentation, should be evaluated. This study demonstrated how the collaborative effort to foster knowledge on machine learning among medical physicists, through interactive sessions and exchange of information among participants, can result in improved models.

Three-dimensional iodine mapping quantified by dual-energy CT for predicting programmed death-ligand 1 expression in invasive pulmonary adenocarcinoma

We examined the association between texture features using three-dimensional (3D) io-dine density histogram on delayed phase of dual-energy CT (DECT) and expression of programmed death-ligand 1 (PD-L1) using immunostaining methods in non-small cell lung cancer. Consecutive 37 patients were scanned by DECT. Unenhanced and enhanced (3 min delay) images were obtained. 3D texture analysis was performed for each nodule to obtain 7 features (max, min, median, mean, standard deviation, skewness, and kurtosis) from iodine density mapping and extracellular volume (ECV). A pathologist evaluated a tumor proportion score (TPS, %) using PD-L1 immunostaining: PD-L1 high (TPS ≥ 50%) and low or negative expression (TPS < 50%). Associations between PD-L1 expression and each 8 parameter were evaluated using logistic regression analysis. The multivariate logistic regression analysis revealed that skewness and ECV were independent indicators associated with high PD-L1 expression (skewness: odds ratio [OR]  7.1 [95% CI 1.1, 45.6], p = 0.039; ECV: OR 6.6 [95% CI 1.1, 38.4], p = 0.037). In the receiver-operating characteristic analysis, the area under the curve of the combination of skewness and ECV was 0.83 (95% CI 0.67, 0.93) with sensitivity of 64% and specificity of 96%. Skewness from 3D iodine density histogram and ECV on dual energy CT were significant factors for predicting PD-L1 expression.

Research Trends and Evolution in Radiogenomics (2005-2023): Bibliometric Analysis

BACKGROUND

Radiogenomics is an emerging technology that integrates genomics and medical image-based radiomics, which is considered a promising approach toward achieving precision medicine.

OBJECTIVE

The aim of this study was to quantitatively analyze the research status, dynamic trends, and evolutionary trajectory in the radiogenomics field using bibliometric methods.

METHODS

The relevant literature published up to 2023 was retrieved from the Web of Science Core Collection. Excel was used to analyze the annual publication trend. VOSviewer was used for constructing the keywords co-occurrence network and the collaboration networks among countries and institutions. CiteSpace was used for citation keywords burst analysis and visualizing the references timeline.

RESULTS

A total of 3237 papers were included and exported in plain-text format. The annual number of publications showed an increasing annual trend. China and the United States have published the most papers in this field, with the highest number of citations in the United States and the highest average number per item in the Netherlands. Keywords burst analysis revealed that several keywords, including "big data," "magnetic resonance spectroscopy," "renal cell carcinoma," "stage," and "temozolomide," experienced a citation burst in recent years. The timeline views demonstrated that the references can be categorized into 8 clusters: lower-grade glioma, lung cancer histology, lung adenocarcinoma, breast cancer, radiation-induced lung injury, epidermal growth factor receptor mutation, late radiotherapy toxicity, and artificial intelligence.

CONCLUSIONS

The field of radiogenomics is attracting increasing attention from researchers worldwide, with the United States and the Netherlands being the most influential countries. Exploration of artificial intelligence methods based on big data to predict the response of tumors to various treatment methods represents a hot spot research topic in this field at present.

A Radiomics Model for the Differentiation of Intracranial Solitary Fibrous Tumor/Hemangiopericytoma and Meningioma Based on Multiparametric Magnetic Resonance Imaging

BACKGROUND

Although the imaging findings of intracranial solitary fibrous tumor (SFT)/hemangiopericytoma (HPC) and meningioma are similar, their treatment and prognosis are quite different. Accurate preoperative identification of these two types of tumors is crucial for individualized treatment.

OBJECTIVE

The aim of this study was to develop a radiomics model for the differentiation of intracranial SFT/HPC and meningioma based on multiparametric magnetic resonance imaging (mpMRI).

MATERIAL AND METHODS

A total of 99 patients from July 2012 to July 2018 with histologically and immunohistochemically confirmed SFT/HPC (n = 40) or meningiomas (n = 59) were retrospectively analyzed. A total of 1118 features were extracted based on its image shape, intensity and texture features. The logistic regression (LR) and multi-layer artificial neural network (ANN) classifiers were used to classify SFT/HPC and meningioma. The predictive performance was calculated using receiver operating characteristic curves (ROC).

RESULTS

We found significant difference in terms of sex between the SFT/HPC and meningioma group (χ2= 4.829, P < 0.05), but no significant difference was found in age (P > 0.05). The most significant radiomics features included five shape and four first-order level features. For the LR classifier, the prediction accuracy of SFT/HPC was 71.0% and meningioma was 78.7%. For the ANN classifier, the prediction accuracy of SFT/HPC was 83.9% and meningioma was 80.9%. Both of the two classifiers achieved a high accuracy rate, but ANN was better.

CONCLUSIONS

Radiomics features, especially when combined with an ANN classifier, can provide satisfactory performance in distinguishing SFT/HPC and meningioma.

The Radiation Therapist profession through the lens of new technology: A practice development paper based on the ESTRO Radiation Therapist Workshops

Technological advances in radiation therapy impact on the role and scope of practice of the radiation therapist. The European Society of Radiotherapy and Oncology (ESTRO) recently held two workshops on this topic and this position paper reflects the outcome of this workshop, which included radiation therapists from all global regions. Workflows, quality assurance, research, IGRT and ART as well as clinical decision making are the areas of radiation therapist practice that will be highly influenced by advancing technology in the near future. This position paper captures the opportunities that this will bring to the radiation therapist profession, to the practice of radiation therapy and ultimately to patient care.

The influence of artificial intelligence on the work of the medical physicist in radiotherapy practice: a short review

There have been many applications and influences of Artificial intelligence (AI) in many sectors and its professionals, that of radiotherapy and the medical physicist is no different. AI and technological advances have necessitated changing roles of medical physicists due to the development of modernized technology with image-guided accessories for the radiotherapy treatment of cancer patients. Given the changing role of medical physicists in ensuring patient safety and optimal care, AI can reshape radiotherapy practice now and in some years to come. Medical physicists' roles in radiotherapy practice have evolved to meet technology for the management of better patient care in the age of modern radiotherapy. This short review provides an insight into the influence of AI on the changing role of medical physicists in each specific chain of the workflow in radiotherapy in which they are involved.

The importance of planning CT-based imaging features for machine learning-based prediction of pain response

Patients suffering from painful spinal bone metastases (PSBMs) often undergo palliative radiation therapy (RT), with an efficacy of approximately two thirds of patients. In this exploratory investigation, we assessed the effectiveness of machine learning (ML) models trained on radiomics, semantic and clinical features to estimate complete pain response. Gross tumour volumes (GTV) and clinical target volumes (CTV) of 261 PSBMs were segmented on planning computed tomography (CT) scans. Radiomics, semantic and clinical features were collected for all patients. Random forest (RFC) and support vector machine (SVM) classifiers were compared using repeated nested cross-validation. The best radiomics classifier was trained on CTV with an area under the receiver-operator curve (AUROC) of 0.62 ± 0.01 (RFC; 95% confidence interval). The semantic model achieved a comparable AUROC of 0.63 ± 0.01 (RFC), significantly below the clinical model (SVM, AUROC: 0.80 ± 0.01); and slightly lower than the spinal instability neoplastic score (SINS; LR, AUROC: 0.65 ± 0.01). A combined model did not improve performance (AUROC: 0,74 ± 0,01). We could demonstrate that radiomics and semantic analyses of planning CTs allowed for limited prediction of therapy response to palliative RT. ML predictions based on established clinical parameters achieved the best results.

Preoperative contrast-enhanced CT-based radiomics signature for predicting hypoxia-inducible factor 1α expression in retroperitoneal sarcoma

AIM

To develop and test a contrast-enhanced computed tomography (CECT)-based radiomics signature (RS) to preoperatively predict hypoxia-inducible factor 1α (HIF-1α) expression in retroperitoneal sarcoma (RPS).

MATERIALS AND METHODS

This study included 129 patients with RPS retrospectively who underwent CECT, including 64 male and 65 female patients (55 [2-84] years). Participants were divided into a training set comprising 85 patients and a test set comprising 44 patients. Clinical data and CECT findings of all patients were collected. RS construction was performed by the minimum redundancy maximum relevance method and least absolute shrinkage and selection operator algorithm. The clinical information was analysed by univariate and multivariate logistic regression analysis. The RS and risk factors were included to build a radiomics nomogram. The predictive efficacy of different models was evaluated by accuracy, area under the receiver operating characteristic curve (AUC), and decision curve analysis.

RESULTS

The RS combined signature was constructed on the basis of multi-phase CECT and had an accuracy of 0.795 and an AUC of 0.719 (95% confidence interval [CI], 0.552-0.886) in the test set, which were higher than that of the radiomics nomogram (accuracy: 0.636; AUC: 0.702 [95% CI, 0.547-0.857]) and the clinical model (accuracy: 0.682; AUC: 0.486 [95% CI, 0.324-0.647]). The decision curve analysis showed that the RS combined signature provided better clinical application than the clinical model and radiomics nomogram.

CONCLUSIONS

The multi-phase CECT-based RS constructed can be used as a powerful tool for predicting HIF-1α expression in patients with RPS.

A Focus on the Synergy of Radiomics and RNA Sequencing in Breast Cancer

Radiological imaging is currently employed as the most effective technique for screening, diagnosis, and follow up of patients with breast cancer (BC), the most common type of tumor in women worldwide. However, the introduction of the omics sciences such as metabolomics, proteomics, and molecular genomics, have optimized the therapeutic path for patients and implementing novel information parallel to the mutational asset targetable by specific clinical treatments. Parallel to the "omics" clusters, radiological imaging has been gradually employed to generate a specific omics cluster termed "radiomics". Radiomics is a novel advanced approach to imaging, extracting quantitative, and ideally, reproducible data from radiological images using sophisticated mathematical analysis, including disease-specific patterns, that could not be detected by the human eye. Along with radiomics, radiogenomics, defined as the integration of "radiology" and "genomics", is an emerging field exploring the relationship between specific features extracted from radiological images and genetic or molecular traits of a particular disease to construct adequate predictive models. Accordingly, radiological characteristics of the tissue are supposed to mimic a defined genotype and phenotype and to better explore the heterogeneity and the dynamic evolution of the tumor over the time. Despite such improvements, we are still far from achieving approved and standardized protocols in clinical practice. Nevertheless, what can we learn by this emerging multidisciplinary clinical approach? This minireview provides a focused overview on the significance of radiomics integrated by RNA sequencing in BC. We will also discuss advances and future challenges of such radiomics-based approach.

Predicting Soft Tissue Sarcoma Response to Neoadjuvant Chemotherapy Using an MRI-Based Delta-Radiomics Approach

OBJECTIVES

To evaluate the performance of machine learning-augmented MRI-based radiomics models for predicting response to neoadjuvant chemotherapy (NAC) in soft tissue sarcomas.

METHODS

Forty-four subjects were identified retrospectively from patients who received NAC at our institution for pathologically proven soft tissue sarcomas. Only subjects who had both a baseline MRI prior to initiating chemotherapy and a post-treatment scan at least 2 months after initiating chemotherapy and prior to surgical resection were included. 3D ROIs were used to delineate whole-tumor volumes on pre- and post-treatment scans, from which 1708 radiomics features were extracted. Delta-radiomics features were calculated by subtraction of baseline from post-treatment values and used to distinguish treatment response through univariate analyses as well as machine learning-augmented radiomics analyses.

RESULTS

Though only 4.74% of variables overall reached significance at p ≤ 0.05 in univariate analyses, Laws Texture Energy (LTE)-derived metrics represented 46.04% of all such features reaching statistical significance. ROC analyses similarly failed to predict NAC response, with AUCs of 0.40 (95% CI 0.22-0.58) and 0.44 (95% CI 0.26-0.62) for RF and AdaBoost, respectively.

CONCLUSION

Overall, while our result was not able to separate NAC responders from non-responders, our analyses did identify a subset of LTE-derived metrics that show promise for further investigations. Future studies will likely benefit from larger sample size constructions so as to avoid the need for data filtering and feature selection techniques, which have the potential to significantly bias the machine learning procedures.

MRI-Based Computer-Aided Diagnostic Model to Predict Tumor Grading and Clinical Outcomes in Patients With Soft Tissue Sarcoma

BACKGROUND

MRI acts as a potential resource for exploration and interpretation to identify tumor characterization by advanced computer-aided diagnostic (CAD) methods.

PURPOSE

To evaluate and validate the performance of MRI-based CAD models for identifying low-grade and high-grade soft tissue sarcoma (STS) and for investigating survival prognostication.

STUDY TYPE

Retrospective.

SUBJECTS

A total of 540 patients (295 male/female: 295/245, median age: 42 years) with STSs.

FIELD SEQUENCE

5-T MRI with T1 WI sequence and fat-suppressed T2 -weighted (T2 FS) sequence.

ASSESSMENT

Manual regions of interests (ROIs) were delineated for generation of radiomic features. Automatic segmentation and pretrained convolutional neural networks (CNNs) were performed for deep learning (DL) analysis. The last fully connected layer at the top of CNNs was removed, and the global max pooling was added to transform feature maps to numeric values. Tumor grade was determined on histological specimens.

STATISTICAL TESTS

The support vector machine was adopted as the classifier for all MRI-based models. The DL signature was derived from the DL-MRI model with the highest area under the curve (AUC). The significant clinical variables, tumor location and size, integrated with radiomics and DL signatures were ready for construction of clinical-MRI nomogram to identify tumor grading. The prognostic value of clinical variables and these MRI-based signatures for overall survival (OS) was evaluated via Cox proportional hazard.

RESULTS

The clinical-MRI differentiation nomogram represented an AUC of 0.870 in the training cohort, and an AUC of 0.855, accuracy of 79.01%, sensitivity of 79.03%, and specificity of 78.95% in the validation cohort. The prognostic model showed good performance for OS with 3-year C-index of 0.681 and 0.642 and 5-year C-index of 0.722 and 0.676 in the training and validation cohorts.

DATA CONCLUSION

MRI-based CAD nomogram represents effective abilities in classification of low-grade and high-grade STSs. The MRI-based prognostic model yields favorable preoperative capacities to identify long-term survivals for STSs.

EVIDENCE LEVEL

3 TECHNICAL EFFICACY: Stage 4.

Value assessment of artificial intelligence in medical imaging: a scoping review

BACKGROUND

Artificial intelligence (AI) is seen as one of the major disrupting forces in the future healthcare system. However, the assessment of the value of these new technologies is still unclear, and no agreed international health technology assessment-based guideline exists. This study provides an overview of the available literature in the value assessment of AI in the field of medical imaging.

METHODS

We performed a systematic scoping review of published studies between January 2016 and September 2020 using 10 databases (Medline, Scopus, ProQuest, Google Scholar, and six related databases of grey literature). Information about the context (country, clinical area, and type of study) and mentioned domains with specific outcomes and items were extracted. An existing domain classification, from a European assessment framework, was used as a point of departure, and extracted data were grouped into domains and content analysis of data was performed covering predetermined themes.

RESULTS

Seventy-nine studies were included out of 5890 identified articles. An additional seven studies were identified by searching reference lists, and the analysis was performed on 86 included studies. Eleven domains were identified: (1) health problem and current use of technology, (2) technology aspects, (3) safety assessment, (4) clinical effectiveness, (5) economics, (6) ethical analysis, (7) organisational aspects, (8) patients and social aspects, (9) legal aspects, (10) development of AI algorithm, performance metrics and validation, and (11) other aspects. The frequency of mentioning a domain varied from 20 to 78% within the included papers. Only 15/86 studies were actual assessments of AI technologies. The majority of data were statements from reviews or papers voicing future needs or challenges of AI research, i.e. not actual outcomes of evaluations.

CONCLUSIONS

This review regarding value assessment of AI in medical imaging yielded 86 studies including 11 identified domains. The domain classification based on European assessment framework proved useful and current analysis added one new domain. Included studies had a broad range of essential domains about addressing AI technologies highlighting the importance of domains related to legal and ethical aspects.

Deep Learning Radiomics Nomogram to Predict Lung Metastasis in Soft-Tissue Sarcoma: A Multi-Center Study

Objectives

To build and evaluate a deep learning radiomics nomogram (DLRN) for preoperative prediction of lung metastasis (LM) status in patients with soft tissue sarcoma (STS).

Methods

In total, 242 patients with STS (training set, n=116; external validation set, n=126) who underwent magnetic resonance imaging were retrospectively enrolled in this study. We identified independent predictors for LM-status and evaluated their performance. The minimum redundancy maximum relevance (mRMR) method and least absolute shrinkage and selection operator (LASSO) algorithm were adopted to screen radiomics features. Logistic regression, decision tree, random forest, support vector machine (SVM), and adaptive boosting classifiers were compared for their ability to predict LM. To overcome the imbalanced distribution of the LM data, we retrained each machine-learning classifier using the synthetic minority over-sampling technique (SMOTE). A DLRN combining the independent clinical predictors with the best performing radiomics prediction signature (mRMR+LASSO+SVM+SMOTE) was established. Area under the receiver operating characteristics curve (AUC), calibration curves, and decision curve analysis (DCA) were used to assess the performance and clinical applicability of the models.

Result

Comparisons of the AUC values applied to the external validation set revealed that the DLRN model (AUC=0.833) showed better prediction performance than the clinical model (AUC=0.664) and radiomics model (AUC=0.799). The calibration curves indicated good calibration efficiency and the DCA showed the DLRN model to have greater clinical applicability than the other two models.

Conclusion

The DLRN was shown to be an accurate and efficient tool for LM-status prediction in STS.

Radiomics-guided radiation therapy: opportunities and challenges

Radiomics is an advanced image-processing framework, which extracts image features and considers them as biomarkers towards personalized medicine. Applications include disease detection, diagnosis, prognosis, and therapy response assessment/prediction. As radiation therapy aims for further individualized treatments, radiomics could play a critical role in various steps before, during and after treatment. Elucidation of the concept of radiomics-guided radiation therapy (RGRT) is the aim of this review, attempting to highlight opportunities and challenges underlying the use of radiomics to guide clinicians and physicists towards more effective radiation treatments. This work identifies the value of RGRT in various steps of radiotherapy from patient selection to follow-up, and subsequently provides recommendations to improve future radiotherapy using quantitative imaging features.

CT-Based Radiomics and Deep Learning for BRCA Mutation and Progression-Free Survival Prediction in Ovarian Cancer Using a Multicentric Dataset

PURPOSE

Build predictive radiomic models for early relapse and BRCA mutation based on a multicentric database of high-grade serous ovarian cancer (HGSOC) and validate them in a test set coming from different institutions.

METHODS

Preoperative CTs of patients with HGSOC treated at four referral centers were retrospectively acquired and manually segmented. Hand-crafted features and deep radiomics features were extracted respectively by dedicated software (MODDICOM) and a dedicated convolutional neural network (CNN). Features were selected with and without prior harmonization (ComBat harmonization), and models were built using different machine learning algorithms, including clinical variables.

RESULTS

We included 218 patients. Radiomic models showed low performance in predicting both BRCA mutation (AUC in test set between 0.46 and 0.59) and 1-year relapse (AUC in test set between 0.46 and 0.56); deep learning models demonstrated similar results (AUC in the test of 0.48 for BRCA and 0.50 for relapse). The inclusion of clinical variables improved the performance of the radiomic models to predict BRCA mutation (AUC in the test set of 0.74).

CONCLUSIONS

In our multicentric dataset, representative of a real-life clinical scenario, we could not find a good radiomic predicting model for PFS and BRCA mutational status, with both traditional radiomics and deep learning, but the combination of clinical and radiomic models improved model performance for the prediction of BRCA mutation. These findings highlight the need for standardization through the whole radiomic pipelines and robust multicentric external validations of results.

Radiomics in pulmonary neuroendocrine tumours (NETs)

Objectives

The aim of this single-centre, observational, retrospective study is to find a correlation using Radiomics between the analysis of CT texture features of primary lesion of neuroendocrine (NET) lung cancer subtypes (typical and atypical carcinoids, large and small cell neuroendocrine carcinoma), Ki-67 index and the presence of lymph nodal mediastinal metastases.

Methods

Twenty-seven patients (11 males and 16 females, aged between 48 and 81 years old—average age of 70,4 years) with histological diagnosis of pulmonary NET with known Ki-67 status and metastases who have performed pre-treatment CT in our department were included. All examinations were performed with the same CT scan (Sensation 16-slice, Siemens). The study protocol was a baseline scan followed by 70 s delay acquisition after administration of intravenous contrast medium. After segmentation of primary lesions, quantitative texture parameters of first and higher orders were extracted. Statistics nonparametric tests and linear correlation tests were conducted to evaluate the relationship between different textural characteristics and tumour subtypes.

Results

Statistically significant (p < 0.05) differences were seen in post-contrast enhanced CT in multiple first and higher-order extracted parameters regarding the correlation with classes of Ki-67 index values. Statistical analysis for direct acquisitions was not significant. Concerning the correlation with the presence of metastases, one histogram feature (Skewness) and one feature included in the Gray-Level Co-occurrence Matrix (ClusterShade) were significant on contrast-enhanced CT only.

Conclusions

CT texture analysis may be used as a valid tool for predicting the subtype of lung NET and its aggressiveness.

Machine learning application in personalised lung cancer recurrence and survivability prediction

Machine learning is an important artificial intelligence technique that is widely applied in cancer diagnosis and detection. More recently, with the rise of personalised and precision medicine, there is a growing trend towards machine learning applications for prognosis prediction. However, to date, building reliable prediction models of cancer outcomes in everyday clinical practice is still a hurdle. In this work, we integrate genomic, clinical and demographic data of lung adenocarcinoma (LUAD) and squamous cell carcinoma (LUSC) patients from The Cancer Genome Atlas (TCGA) and introduce copy number variation (CNV) and mutation information of 15 selected genes to generate predictive models for recurrence and survivability. We compare the accuracy and benefits of three well-established machine learning algorithms: decision tree methods, neural networks and support vector machines. Although the accuracy of predictive models using the decision tree method has no significant advantage, the tree models reveal the most important predictors among genomic information (e.g. KRAS, EGFR, TP53), clinical status (e.g. TNM stage and radiotherapy) and demographics (e.g. age and gender) and how they influence the prediction of recurrence and survivability for both early stage LUAD and LUSC. The machine learning models have the potential to help clinicians to make personalised decisions on aspects such as follow-up timeline and to assist with personalised planning of future social care needs.

Multi-center evaluation of artificial intelligent imaging and clinical models for predicting neoadjuvant chemotherapy response in breast cancer

BACKGROUND

Neoadjuvant chemotherapy (NAC) plays an important role in the management of locally advanced breast cancer. It allows for downstaging of tumors, potentially allowing for breast conservation. NAC also allows for in-vivo testing of the tumors' response to chemotherapy and provides important prognostic information. There are currently no clearly defined clinical models that incorporate imaging with clinical data to predict response to NAC. Thus, the aim of this work is to develop a predictive AI model based on routine CT imaging and clinical parameters to predict response to NAC.

METHODS

The CT scans of 324 patients with NAC from multiple centers in Singapore were used in this study. Four different radiomics models were built for predicting pathological complete response (pCR): first two were based on textural features extracted from peri-tumoral and tumoral regions, the third model based on novel space-resolved radiomics which extract feature maps using voxel-based radiomics and the fourth model based on deep learning (DL). Clinical parameters were included to build a final prognostic model.

RESULTS

The best performing models were based on space-resolved and DL approaches. Space-resolved radiomics improves the clinical AUCs of pCR prediction from 0.743 (0.650 to 0.831) to 0.775 (0.685 to 0.860) and our DL model improved it from 0.743 (0.650 to 0.831) to 0.772 (0.685 to 0.853). The tumoral radiomics model performs the worst with no improvement of the AUC from the clinical model. The peri-tumoral combined model gives moderate performance with an AUC of 0.765 (0.671 to 0.855).

CONCLUSIONS

Radiomics features extracted from diagnostic CT augment the predictive ability of pCR when combined with clinical features. The novel space-resolved radiomics and DL radiomics approaches outperformed conventional radiomics techniques.

MRI-based delta-radiomics predicts pathologic complete response in high-grade soft-tissue sarcoma patients treated with neoadjuvant therapy

PURPOSE

In high-grade soft-tissue sarcomas (STS) the standard of care encompasses multimodal therapy regimens. While there is a growing body of evidence for prognostic pretreatment radiomic models, we hypothesized that temporal changes in radiomic features following neoadjuvant treatment ("delta-radiomics") may be able to predict the pathological complete response (pCR).

METHODS

MRI scans (T1-weighted with fat-saturation and contrast-enhancement (T1FSGd) and T2-weighted with fat-saturation (T2FS)) of patients with STS of the extremities and trunk treated with neoadjuvant therapy were gathered from two independent institutions (training: 103, external testing: 53 patients). pCR was defined as <5% viable cells. After segmentation and preprocessing, 105 radiomic features were extracted. Delta-radiomic features were calculated by subtraction of features derived from MRI scans obtained before and after neoadjuvant therapy. After feature reduction, machine learning modeling was performed in 100 iterations of 3-fold nested cross-validation. Delta-radiomic models were compared with single timepoint models in the testing cohort.

RESULTS

The combined delta-radiomic models achieved the best area under the receiver operating characteristic curve (AUC) of 0.75. Pre-therapeutic tumor volume was the best conventional predictor (AUC 0.70). The T2FS-based delta-radiomic model had the most balanced classification performance with a balanced accuracy of 0.69. Delta-radiomic models achieved better reproducibility than single timepoint radiomic models, RECIST or the peri-therapeutic volume change. Delta-radiomic models were significantly associated with survival in multivariate Cox regression.

CONCLUSION

This exploratory analysis demonstrated that MRI-based delta-radiomics improves prediction of pCR over tumor volume and RECIST. Delta-radiomics may one day function as a biomarker for personalized treatment adaptations.

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.

Distinct tumor signatures using deep learning-based characterization of the peritumoral microenvironment in glioblastomas and brain metastases

Tumor types are classically distinguished based on biopsies of the tumor itself, as well as a radiological interpretation using diverse MRI modalities. In the current study, the overarching goal is to demonstrate that primary (glioblastomas) and secondary (brain metastases) malignancies can be differentiated based on the microstructure of the peritumoral region. This is achieved by exploiting the extracellular water differences between vasogenic edema and infiltrative tissue and training a convolutional neural network (CNN) on the Diffusion Tensor Imaging (DTI)-derived free water volume fraction. We obtained 85% accuracy in discriminating extracellular water differences between local patches in the peritumoral area of 66 glioblastomas and 40 metastatic patients in a cross-validation setting. On an independent test cohort consisting of 20 glioblastomas and 10 metastases, we got 93% accuracy in discriminating metastases from glioblastomas using majority voting on patches. This level of accuracy surpasses CNNs trained on other conventional DTI-based measures such as fractional anisotropy (FA) and mean diffusivity (MD), that have been used in other studies. Additionally, the CNN captures the peritumoral heterogeneity better than conventional texture features, including Gabor and radiomic features. Our results demonstrate that the extracellular water content of the peritumoral tissue, as captured by the free water volume fraction, is best able to characterize the differences between infiltrative and vasogenic peritumoral regions, paving the way for its use in classifying and benchmarking peritumoral tissue with varying degrees of infiltration.

Predicting Treatment Response of Neoadjuvant Chemoradiotherapy in Locally Advanced Rectal Cancer Using Amide Proton Transfer MRI Combined With Diffusion-Weighted Imaging

Objective

To evaluate amide proton weighted (APTw) MRI combined with diffusion-weighted imaging (DWI) in predicting neoadjuvant chemoradiotherapy (NCRT) response in patients with locally advanced rectal cancer (LARC).

Methods

53 patients with LARC were enrolled in this retrospective study. MR examination including APTw MRI and DWI was performed before and after NCRT. APTw SI, ADC value, tumor size, CEA level before and after NCRT were assessed. The difference of the above parameters between before and after NCRT was calculated. The tumor regression grading (TRG) was assessed by American Joint Committee on Cancer’s Cancer Staging Manual AJCC 8th score. The Shapiro-Wilk test, paired t-test and Wilcoxon Signed Ranks test, two-sample t-test, Mann-Whitney U test and multivariate analysis were used for statistical analysis.

Results

Of the 53 patients, 19 had good responses (TRG 0-1), 34 had poor responses (TRG 2-3). After NCRT, all the rectal tumors demonstrated decreased APT values, increased ADC values, reduced tumor volumes and CEA levels (all p < 0.001). Good responders demonstrated higher pre-APT values, higher Δ APT values, lower pre- ADC values and higher Δ tumor volumes than poor responders. Pre-APT combined with pre-ADC achieved the best diagnostic performance, with AUC of 0.895 (sensitivity of 85.29%, specificity of 89.47%, p < 0.001) in predicting good response to NCRT.

Conclusion

The combination of APTw and DWI may serve as a noninvasive biomarker for evaluating and identifying response to NCRT in LARC patients.

Development and External Validation of Deep-Learning-Based Tumor Grading Models in Soft-Tissue Sarcoma Patients Using MR Imaging

BACKGROUND

In patients with soft-tissue sarcomas, tumor grading constitutes a decisive factor to determine the best treatment decision. Tumor grading is obtained by pathological work-up after focal biopsies. Deep learning (DL)-based imaging analysis may pose an alternative way to characterize STS tissue. In this work, we sought to non-invasively differentiate tumor grading into low-grade (G1) and high-grade (G2/G3) STS using DL techniques based on MR-imaging.

METHODS

Contrast-enhanced T1-weighted fat-saturated (T1FSGd) MRI sequences and fat-saturated T2-weighted (T2FS) sequences were collected from two independent retrospective cohorts (training: 148 patients, testing: 158 patients). Tumor grading was determined following the French Federation of Cancer Centers Sarcoma Group in pre-therapeutic biopsies. DL models were developed using transfer learning based on the DenseNet 161 architecture.

RESULTS

The T1FSGd and T2FS-based DL models achieved area under the receiver operator characteristic curve (AUC) values of 0.75 and 0.76 on the test cohort, respectively. T1FSGd achieved the best F1-score of all models (0.90). The T2FS-based DL model was able to significantly risk-stratify for overall survival. Attention maps revealed relevant features within the tumor volume and in border regions.

CONCLUSIONS

MRI-based DL models are capable of predicting tumor grading with good reproducibility in external validation.

Quantitative assessment of acquisition imaging parameters on MRI radiomics features: a prospective anthropomorphic phantom study using a 3D-T2W-TSE sequence for MR-guided-radiotherapy

Background

MRI pulse sequences and imaging parameters substantially influence the variation of MRI radiomics features, thus impose a critical challenge on MRI radiomics reproducibility and reliability. This study aims to prospectively investigate the impact of various imaging parameters on MRI radiomics features in a 3D T2-weighted (T2W) turbo-spin-echo (TSE) pulse sequence for MR-guided-radiotherapy (MRgRT).

Methods

An anthropomorphic phantom was scanned using a 3D-T2W-TSE MRgRT sequence at 1.5T under a variety of acquisition imaging parameter changes. T1 and T2 relaxation times of the phantom were also measured. 93 first-order and texture radiomics features in the original and 14 transformed images, yielding 1,395 features in total, were extracted from 10 volumes-of-interest (VOIs). The percentage deviation (d%) of radiomics feature values from the baseline values and intra-class correlation coefficient (ICC) with the baseline were calculated. Robust radiomics features were identified based on the excellent agreement of radiomics feature values with the baseline, i.e., the averaged d% <5% and ICC >0.90 in all VOIs for all imaging parameter variations.

Results

The radiomics feature values changed considerably but to different degrees with different imaging parameter adjustments, in the ten VOIs. The deviation d% ranged from 0.02% to 321.3%, with a mean of 12.5% averaged for all original features in all ten VOIs. First-order and GLCM features were generally more robust to imaging parameters than other features in the original images. There were also significantly different radiomics feature values (ANOVA, P<0.001) between the original and the transformed images, exhibiting quite different robustness to imaging parameters. 330 out of 1395 features (23.7%) robust to imaging parameters were identified. GLCM and GLSZM features had the most (42.5%, 153/360) and least (3.8%, 9/240) robust features in the original and transformed images, respectively.

Conclusions

This study helps better understand the quantitative dependence of radiomics feature values on imaging parameters in a 3D-T2W-TSE sequence for MRgRT. Imaging parameter heterogeneity should be considered as a significant source of radiomics variability and uncertainty, which must be well harmonized for reliable clinical use. The identified robust features to imaging parameters are helpful for the pre-selection of radiomics features for reliable radiomics modeling.

Prognostic Assessment in High-Grade Soft-Tissue Sarcoma Patients: A Comparison of Semantic Image Analysis and Radiomics

BACKGROUND

In patients with soft-tissue sarcomas of the extremities, the treatment decision is currently regularly based on tumor grading and size. The imaging-based analysis may pose an alternative way to stratify patients' risk. In this work, we compared the value of MRI-based radiomics with expert-derived semantic imaging features for the prediction of overall survival (OS).

METHODS

Fat-saturated T2-weighted sequences (T2FS) and contrast-enhanced T1-weighted fat-saturated (T1FSGd) sequences were collected from two independent retrospective cohorts (training: 108 patients; testing: 71 patients). After preprocessing, 105 radiomic features were extracted. Semantic imaging features were determined by three independent radiologists. Three machine learning techniques (elastic net regression (ENR), least absolute shrinkage and selection operator, and random survival forest) were compared to predict OS.

RESULTS

ENR models achieved the best predictive performance. Histologies and clinical staging differed significantly between both cohorts. The semantic prognostic model achieved a predictive performance with a C-index of 0.58 within the test set. This was worse compared to a clinical staging system (C-index: 0.61) and the radiomic models (C-indices: T1FSGd: 0.64, T2FS: 0.63). Both radiomic models achieved significant patient stratification.

CONCLUSIONS

T2FS and T1FSGd-based radiomic models outperformed semantic imaging features for prognostic assessment.

Expanding the medical physicist curricular and professional programme to include Artificial Intelligence

PURPOSE

To provide a guideline curriculum related to Artificial Intelligence (AI), for the education and training of European Medical Physicists (MPs).

MATERIALS AND METHODS

The proposed curriculum consists of two levels: Basic (introducing MPs to the pillars of knowledge, development and applications of AI, in the context of medical imaging and radiation therapy) and Advanced. Both are common to the subspecialties (diagnostic and interventional radiology, nuclear medicine, and radiation oncology). The learning outcomes of the training are presented as knowledge, skills and competences (KSC approach).

RESULTS

For the Basic section, KSCs were stratified in four subsections: (1) Medical imaging analysis and AI Basics; (2) Implementation of AI applications in clinical practice; (3) Big data and enterprise imaging, and (4) Quality, Regulatory and Ethical Issues of AI processes. For the Advanced section instead, a common block was proposed to be further elaborated by each subspecialty core curriculum. The learning outcomes were also translated into a syllabus of a more traditional format, including practical applications.

CONCLUSIONS

This AI curriculum is the first attempt to create a guideline expanding the current educational framework for Medical Physicists in Europe. It should be considered as a document to top the sub-specialties' curriculums and adapted by national training and regulatory bodies. The proposed educational program can be implemented via the European School of Medical Physics Expert (ESMPE) course modules and - to some extent - also by the national competent EFOMP organizations, to reach widely the medical physicist community in Europe.

Impact of neoadjuvant androgen deprivation therapy on magnetic resonance imaging features in prostate cancer before radiotherapy

Background and purpose

In locally advanced prostate cancer (PC), androgen deprivation therapy (ADT) in combination with whole prostate radiotherapy (RT) is the standard treatment. ADT affects the prostate as well as the tumour on multiparametric magnetic resonance imaging (MRI) with decreased PC conspicuity and impaired localisation of the prostate lesion. Image texture analysis has been suggested to be of aid in separating tumour from normal tissue. The aim of the study was to investigate the impact of ADT on baseline defined MRI features in prostate cancer with the goal to investigate if it might be of use in radiotherapy planning.

Materials and methods

Fifty PC patients were included. Multiparametric MRI was performed before, and three months after ADT. At baseline, a tumour volume was delineated on apparent diffusion coefficient (ADC) maps with suspected tumour content and a reference volume in normal prostatic tissue. These volumes were transferred to MRIs after ADT and were analysed with first-order -and invariant Haralick -features.

Results

At baseline, the median value and several of the invariant Haralick features of ADC, showed a significant difference between tumour and reference volumes. After ADT, only ADC median value could significantly differentiate the two volumes.

Conclusions

Invariant Haralick -features could not distinguish between baseline MRI defined PC and normal tissue after ADT. First-order median value remained significantly different in tumour and reference volumes after ADT, but the difference was less pronounced than before ADT.

Magnetic Resonance Imaging-Based Radiomics Nomogram for Prediction of the Histopathological Grade of Soft Tissue Sarcomas: A Two-Center Study

BACKGROUND

Preoperative prediction of soft tissue sarcoma (STS) grade is important for treatment decisions. Therefore, formulation an STS grade model is strongly needed.

PURPOSE

To develop and test an magnetic resonance imaging (MRI)-based radiomics nomogram for predicting the grade of STS (low-grade vs. high grade).

STUDY TYPE

Retrospective POPULATION: One hundred and eighty patients with STS confirmed by pathologic results at two independent institutions were enrolled (training set, N = 109; external validation set, N = 71).

FIELD STRENGTH/SEQUENCE

Unenhanced T1-weighted (T1WI) and fat-suppressed T2-weighted images (FS-T2WI) were acquired at 1.5 T and 3.0 T.

ASSESSMENT

Clinical-MRI characteristics included age, gender, tumor-node-metastasis (TNM) stage, American Joint Committee on Cancer (AJCC) stage, progression-free survival (PFS), and MRI morphological features (ie, margin). Radiomics feature extraction were performed on T1WI and FS-T2WI images by minimum redundancy maximum relevance (MRMR) method and least absolute shrinkage and selection operator (LASSO) algorithm. The selected features constructed three radiomics signatures models (RS-T1， RS-FST2, and RS-Combined). Univariate and multivariate logistic regression analysis were applied for screening significant risk factors. Radiomics nomogram was constructed by incorporating the radiomics signature and risk factors.

STATISTICAL TESTS

Clinical-MRI characteristics were performed by a univariate analysis. Model performances (discrimination, calibration, and clinical usefulness) were validated in the external validation set. The RS-T1 model, RS-FST2 model, and RS-Combined model had an area under curves (AUCs) of 0.645, 0.641, and 0.829, respectively, in the external validation set. The radiomics nomogram, incorporating significant risk factors and the RS-Combined model had AUCs of 0.916 (95%CI, 0.866-0.966, training set) and 0.879 (95%CI, 0.791-0.967, external validation set), and demonstrated good calibration and good clinical utility.

DATA CONCLUSION

The proposed noninvasive MRI-based radiomics models showed good performance in differentiating low-grade from high-grade STSs.

LEVEL OF EVIDENCE

3 TECHNICAL EFFICACY STAGE: 2.

Artificial Intelligence and the Medical Physicist: Welcome to the Machine

Artificial intelligence (AI) is a branch of computer science dedicated to giving machines or computers the ability to perform human-like cognitive functions, such as learning, problem-solving, and decision making. Since it is showing superior performance than well-trained human beings in many areas, such as image classification, object detection, speech recognition, and decision-making, AI is expected to change profoundly every area of science, including healthcare and the clinical application of physics to healthcare, referred to as medical physics. As a result, the Italian Association of Medical Physics (AIFM) has created the “AI for Medical Physics” (AI4MP) group with the aims of coordinating the efforts, facilitating the communication, and sharing of the knowledge on AI of the medical physicists (MPs) in Italy. The purpose of this review is to summarize the main applications of AI in medical physics, describe the skills of the MPs in research and clinical applications of AI, and define the major challenges of AI in healthcare.

A deep survival interpretable radiomics model of hepatocellular carcinoma patients

This work aims to identify a new radiomics signature using imaging phenotypes and clinical variables for risk prediction of overall survival (OS) in hepatocellular carcinoma (HCC) patients treated with stereotactic body radiation therapy (SBRT). 167 patients were retrospectively analyzed with repeated nested cross-validation to mitigate overfitting issues. 56 radiomic features were extracted from pre-treatment contrast-enhanced (CE) CT images. 37 clinical factors were obtained from patients' electronic records. Variational autoencoders (VAE) based survival models were designed for radiomics and clinical features and a convolutional neural network (CNN) survival model was used for the CECT. Finally, radiomics, clinical and raw image deep learning network (DNN) models were combined to predict the risk probability for OS. The final models yielded c-indices of 0.579 (95%CI: 0.544-0.621), 0.629 (95%CI: 0.601-0.643), 0.581 (95%CI: 0.553-0.613) and 0.650 (95%CI: 0.635-0.683) for radiomics, clinical, image input and combined models on nested cross validation scheme, respectively. Integrated gradients method was used to interpret the trained models. Our interpretability analysis of the DNN showed that the top ranked features were clinical liver function and liver exclusive of tumor radiomics features, which suggests a prominent role of side effects and toxicities in liver outside the tumor region in determining the survival rate of these patients. In summary, novel deep radiomic analysis provides improved performance for risk assessment of HCC prognosis compared with Cox survival models and may facilitate stratification of HCC patients and personalization of their treatment strategies. Liver function was found to contribute most to the OS for these HCC patients and radiomics can aid in their management.

Enhanced childhood diseases treatment using computational models: Systematic review of intelligent experiments heading to precision medicine

INTRODUCTION

Precision or personalized Medicine (PM) is used for the prevention and treatment of diseases by considering a huge amount of information about individuals variables. Due to high volume of information, AI-based computational models are required. A large set of studies conducted to examine the PM approach to improve childhood clinical outcomes. Thus, the main goal of this study was to review the application of health information technology and especially artificial intelligence (AI) methods for the treatment of childhood disease using PM.

METHODS

PubMed, Scopus, Web of Science, and EMBASE databases were searched up to December 18, 2019. Articles that focused on informatics applications for childhood disease PM included in this study. Included papers were classified for qualitative analysis and interpreting results. The results were analyzed using Microsoft Excel 2019.

RESULTS

From 341 citations, 62 papers met our inclusion criteria. The number of published papers that used AI methods to apply for PM in childhood diseases increased from 2010 to 2019. Our results showed that most applied methods were related to machine learning discipline. In terms of clinical scope, the largest number of clinical articles are devoted to oncology. Besides, the analysis showed that genomics was the most PM approach used regarding childhood disease.

CONCLUSION

This systematic review examined papers that used AI methods for applying PM approaches in childhood diseases from medical informatics perspectives. Thus, it provided new insight to researchers who are interested in knowing research needs in this field.

A CT-based radiomics model to detect prostate cancer lymph node metastases in PSMA radioguided surgery patients

PURPOSE

In recurrent prostate carcinoma, determination of the site of recurrence is crucial to guide personalized therapy. In contrast to prostate-specific membrane antigen (PSMA)-positron emission tomography (PET) imaging, computed tomography (CT) has only limited capacity to detect lymph node metastases (LNM). We sought to develop a CT-based radiomic model to predict LNM status using a PSMA radioguided surgery (RGS) cohort with histological confirmation of all suspected lymph nodes (LNs).

METHODS

Eighty patients that received RGS for resection of PSMA PET/CT-positive LNMs were analyzed. Forty-seven patients (87 LNs) that received inhouse imaging were used as training cohort. Thirty-three patients (62 LNs) that received external imaging were used as testing cohort. As gold standard, histological confirmation was available for all LNs. After preprocessing, 156 radiomic features analyzing texture, shape, intensity, and local binary patterns (LBP) were extracted. The least absolute shrinkage and selection operator (radiomic models) and logistic regression (conventional parameters) were used for modeling.

RESULTS

Texture and shape features were largely correlated to LN volume. A combined radiomic model achieved the best predictive performance with a testing-AUC of 0.95. LBP features showed the highest contribution to model performance. This model significantly outperformed all conventional CT parameters including LN short diameter (AUC 0.84), LN volume (AUC 0.80), and an expert rating (AUC 0.67). In lymph node-specific decision curve analysis, there was a clinical net benefit above LN short diameter.

CONCLUSION

The best radiomic model outperformed conventional measures for detection of LNM demonstrating an incremental value of radiomic features.

Mini review: Personalization of the radiation therapy management of prostate cancer using MRI-based radiomics

Decisions on how to treat prostate cancer with radiation therapy are guideline-based but as such guidelines have been developed for populations of patients, this invariably leads to overly aggressive treatment in some patients and insufficient treatment in others. Heterogeneity within prostate tumors and in metastatic sites, even within the same patient, is believed to be a major cause of treatment failure. Radiomics biomarkers, more commonly referred to as radiomics 'features", provide readily available, cost-effective, non-invasive tools for screening, detecting tumors and serial monitoring of patients, including assessments of response to therapy and identification of therapeutic complications. Radiomics offers the potential to analyse whole tumors in 3D, as well as sub-regions or 'habitats' within tumors. Combining quantitative information from imaging with pathology, demographic details and other biomarkers will pave the way for personalised treatment selection and monitoring in prostate cancer. The aim of this review is to consider if MRI-based radiomics can bridge the gap between population-based management and personalised management of prostate cancer.

2D and 3D convolutional neural networks for outcome modelling of locally advanced head and neck squamous cell carcinoma

For treatment individualisation of patients with locally advanced head and neck squamous cell carcinoma (HNSCC) treated with primary radiochemotherapy, we explored the capabilities of different deep learning approaches for predicting loco-regional tumour control (LRC) from treatment-planning computed tomography images. Based on multicentre cohorts for exploration (206 patients) and independent validation (85 patients), multiple deep learning strategies including training of 3D- and 2D-convolutional neural networks (CNN) from scratch, transfer learning and extraction of deep autoencoder features were assessed and compared to a clinical model. Analyses were based on Cox proportional hazards regression and model performances were assessed by the concordance index (C-index) and the model's ability to stratify patients based on predicted hazards of LRC. Among all models, an ensemble of 3D-CNNs achieved the best performance (C-index 0.31) with a significant association to LRC on the independent validation cohort. It performed better than the clinical model including the tumour volume (C-index 0.39). Significant differences in LRC were observed between patient groups at low or high risk of tumour recurrence as predicted by the model ([Formula: see text]). This 3D-CNN ensemble will be further evaluated in a currently ongoing prospective validation study once follow-up is complete.

The Application of Artificial Intelligence in Prostate Cancer Management—What Improvements Can Be Expected? A Systematic Review

Artificial Intelligence (AI) is progressively remodeling our daily life. A large amount of information from “big data” now enables machines to perform predictions and improve our healthcare system. AI has the potential to reshape prostate cancer (PCa) management thanks to growing applications in the field. The purpose of this review is to provide a global overview of AI in PCa for urologists, pathologists, radiotherapists, and oncologists to consider future changes in their daily practice. A systematic review was performed, based on PubMed MEDLINE, Google Scholar, and DBLP databases for original studies published in English from January 2009 to January 2019 relevant to PCa, AI, Machine Learning, Artificial Neural Networks, Convolutional Neural Networks, and Natural-Language Processing. Only articles with full text accessible were considered. A total of 1008 articles were reviewed, and 48 articles were included. AI has potential applications in all fields of PCa management: analysis of genetic predispositions, diagnosis in imaging, and pathology to detect PCa or to differentiate between significant and non-significant PCa. AI also applies to PCa treatment, whether surgical intervention or radiotherapy, skills training, or assessment, to improve treatment modalities and outcome prediction. AI in PCa management has the potential to provide a useful role by predicting PCa more accurately, using a multiomic approach and risk-stratifying patients to provide personalized medicine.

Radiomics at a Glance: A Few Lessons Learned from Learning Approaches

Simple Summary

Radiomics has become a prominent component of medical imaging research and many studies show its specific value as a support tool for clinical decision-making processes. Radiomic data are typically analyzed with statistical and machine learning methods, which change depending on the disease context and the imaging modality. We found a certain bias in the literature towards the use of such methods and believe that this limitation may influence the capacity of producing accurate and reliable decisions. Therefore, in view of the relevance of various types of learning methods, we report their significance and discuss their unrevealed potential.

Abstract

Processing and modeling medical images have traditionally represented complex tasks requiring multidisciplinary collaboration. The advent of radiomics has assigned a central role to quantitative data analytics targeting medical image features algorithmically extracted from large volumes of images. Apart from the ultimate goal of supporting diagnostic, prognostic, and therapeutic decisions, radiomics is computationally attractive due to specific strengths: scalability, efficiency, and precision. Optimization is achieved by highly sophisticated statistical and machine learning algorithms, but it is especially deep learning that stands out as the leading inference approach. Various types of hybrid learning can be considered when building complex integrative approaches aimed to deliver gains in accuracy for both classification and prediction tasks. This perspective reviews some selected learning methods by focusing on both their significance for radiomics and their unveiled potential.

Dynamic contrast-enhanced MRI differentiates hepatocellular carcinoma from hepatic metastasis of rectal cancer by extracting pharmacokinetic parameters and radiomic features

The aim of the present study was to explore how dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) may differentiate hepatocellular carcinoma (HCC) from hepatic metastasis of rectal cancer (HMRC) by extracting pharmacokinetic parameters and radiomic features. A total of 75 patients, including 41 cases with HCC and 34 cases with HMRC, underwent DCE-MRI examination. Dual-input two-compartment extended Tofts tracer kinetic model attached to a specialized image post-processing software package from OmniKinetics; GE Healthcare was used to calculate the values of the pharmacokinetic parameters and radiomic features, which were extracted from the lesions at the same region of interest. These values were evaluated using Student's t-test and receiver operating characteristic curves, and discriminant models were built to differentiate between HCC and HRMC. The results identified statistically significant differences in the values of the pharmacokinetic parameters hepatic perfusion index (HPI), endothelial transfer constant (K^trans), initial area under the gadolinium concentration curve during the first 60 sec (IAUC) between the HCC and HRMC groups. In addition, statistically significant differences in 17 radiomic features were observed between the two groups (P<0.05). The areas under the receiver operating characteristic (ROC) curves of the pharmacokinetic parameters K^trans, IAUC and HPI were 0.73, 0.77 and 0.67, respectively. The range of the areas under the ROC curves of the 17 radiomic features with statistical differences was 0.63-0.79. In addition, when pharmacokinetic parameters and radiomic features were incorporated, the area under the ROC curve was 0.86. The accuracy of Fisher's discriminant analysis model based on radiomic features was 89.3%, and the leave-one-out cross-validation accuracy was 80.0%. In conclusion, DCE-MRI was demonstrated to be useful in the differential diagnosis of HCC and HMRC by extracting pharmacokinetic parameters and radiomic features, and incorporation of the two paths improved the diagnostic efficacy. A discriminant model based on radiomic features further enhanced the identification of HCC and HMRC.

Application and limitation of radiomics approach to prognostic prediction for lung stereotactic body radiotherapy using breath-hold CT images with random survival forest: A multi-institutional study

PURPOSE

To predict local recurrence (LR) and distant metastasis (DM) in early stage non-small cell lung cancer (NSCLC) patients after stereotactic body radiotherapy (SBRT) in multiple institutions using breath-hold computed tomography (CT)-based radiomic features with random survival forest.

METHODS

A total of 573 primary early stage NSCLC patients who underwent SBRT between January 2006 and March 2016 and met the eligibility criteria were included in this study. Patients were divided into two datasets: training (464 patients in 10 institutions) and test (109 patients in one institution) datasets. A total of 944 radiomic features were extracted from manually segmented gross tumor volumes (GTVs). Feature selection was performed by analyzing inter-segmentation reproducibility, GTV correlation, and inter-feature redundancy. Nine clinical factors, including histology and GTV size, were also used. Three prognostic models (clinical, radiomic, and combined) for LR and DM were constructed using random survival forest (RSF) to deal with total death as a competing risk in the training dataset. Robust models with optimal hyper-parameters were determined using fivefold cross-validation. The patients were dichotomized into two groups based on the median value of the patient-specific risk scores (high- and low-risk score groups). Gray's test was used to evaluate the statistical significance between the two risk score groups. The prognostic power was evaluated by the concordance index with the 95% confidence intervals (CI) via bootstrapping (2000 iterations).

RESULTS

The concordance indices at 3 yr of clinical, radiomic, and combined models for LR were 0.57 [CI: 0.39-0.75], 0.55 [CI: 0.38-0.73], and 0.61 [CI: 0.43-0.78], respectively, whereas those for DM were 0.59 [CI: 0.54-0.79], 0.67 [CI: 0.54-0.79], and 0.68 [CI: 0.55-0.81], respectively, in the test dataset. The combined DM model significantly discriminated its cumulative incidence between high- and low-risk score groups (P < 0.05). The variable importance of RSF in the combined model for DM indicated that two radiomic features were more important than other clinical factors. The feature maps generated on the basis of the most important radiomic feature had visual difference between high- and low-risk score groups.

CONCLUSIONS

The radiomics approach with RSF for competing risks using breath-hold CT-based radiomic features might predict DM in early stage NSCLC patients who underwent SBRT although that may not have potential to predict LR.

Automatic Tumor Segmentation With a Convolutional Neural Network in Multiparametric MRI: Influence of Distortion Correction

Precise tumor segmentation is a crucial task in radiation therapy planning. Convolutional neural networks (CNNs) are among the highest scoring automatic approaches for tumor segmentation. We investigate the difference in segmentation performance of geometrically distorted and corrected diffusion-weighted data using data of patients with head and neck tumors; 18 patients with head and neck tumors underwent multiparametric magnetic resonance imaging, including T2w, T1w, T2*, perfusion (k_trans), and apparent diffusion coefficient (ADC) measurements. Owing to strong geometrical distortions in diffusion-weighted echo planar imaging in the head and neck region, ADC data were additionally distortion corrected. To investigate the influence of geometrical correction, first 14 CNNs were trained on data with geometrically corrected ADC and another 14 CNNs were trained using data without the correction on different samples of 13 patients for training and 4 patients for validation each. The different sets were each trained from scratch using randomly initialized weights, but the training data distributions were pairwise equal for corrected and uncorrected data. Segmentation performance was evaluated on the remaining 1 test-patient for each of the 14 sets. The CNN segmentation performance scored an average Dice coefficient of 0.40 ± 0.18 for data including distortion-corrected ADC and 0.37 ± 0.21 for uncorrected data. Paired t test revealed that the performance was not significantly different (P = .313). Thus, geometrical distortion on diffusion-weighted imaging data in patients with head and neck tumor does not significantly impair CNN segmentation performance in use.

Glioblastomas and brain metastases differentiation following an MRI texture analysis-based radiomics approach

PURPOSE

To evaluate the potential of 2D texture features extracted from magnetic resonance (MR) images for differentiating brain metastasis (BM) and glioblastomas (GBM) following a radiomics approach.

METHODS

This retrospective study included 50 patients with BM and 50 with GBM who underwent T1-weighted MRI between December 2010 and January 2017. Eighty-eight rotation-invariant texture features were computed for each segmented lesion using six texture analysis methods. These features were also extracted from the four images obtained after applying the discrete wavelet transform (88 features × 4 images). Three feature selection methods and five predictive models were evaluated. A 5-fold cross-validation scheme was used to randomly split the study group into training (80 patients) and testing (20 patients), repeating the process ten times. Classification was evaluated computing the average area under the receiver operating characteristic curve. Sensibility, specificity and accuracy were also computed. The whole process was tested quantizing the images with different gray-level values to evaluate their influence in the final results.

RESULTS

Highest classification accuracy was obtained using the original images quantized with 128 gray-levels and a feature selection method based on the p-value. The best overall performance was achieved using a support vector machine model with a subset of 32 features (AUC = 0.896 ± 0.067, sensitivity of 82% and specificity of 80%). Naïve Bayes and k-nearest neighbors models showed also valuable results (AUC ≈ 0.8) with a lower number of features (<13), thus suggesting that these models may be more generalizable when using external validations.

CONCLUSION

The proposed radiomics MRI approach is able to discriminate between GBM and BM with high accuracy employing a set of 2D texture features, thus helping in the diagnosis of brain lesions in a fast and non-invasive way.

An Efficient Content-Based Image Retrieval System for the Diagnosis of Lung Diseases

The main problem in content-based image retrieval (CBIR) systems is the semantic gap which needs to be reduced for efficient retrieval. The common imaging signs (CISs) which appear in the patient's lung CT scan play a significant role in the identification of cancerous lung nodules and many other lung diseases. In this paper, we propose a new combination of descriptors for the effective retrieval of these imaging signs. First, we construct a feature database by combining local ternary pattern (LTP), local phase quantization (LPQ), and discrete wavelet transform. Next, joint mutual information (JMI)-based feature selection is deployed to reduce the redundancy and to select an optimal feature set for CISs retrieval. To this end, similarity measurement is performed by combining visual and semantic information in equal proportion to construct a balanced graph and the shortest path is computed for learning contextual similarity to obtain final similarity between each query and database image. The proposed system is evaluated on a publicly available database of lung CT imaging signs (LISS), and results are retrieved based on visual feature similarity comparison and graph-based similarity comparison. The proposed system achieves a mean average precision (MAP) of 60% and 0.48 AUC of precision-recall (P-R) graph using only visual features similarity comparison. These results further improve on graph-based similarity measure with a MAP of 70% and 0.58 AUC which shows the superiority of our proposed scheme.

Radiogenomics Based on PET Imaging

Radiogenomics or imaging genomics is a novel omics strategy of associating imaging data with genetic information, which has the potential to advance personalized medicine. Imaging features extracted from PET or PET/CT enable assessment of in vivo functional and physiological activity and provide comprehensive tumor information non-invasively. However, PET features are considered secondary to features on conventional imaging, and there has not yet been a review of the radiogenomic approach using PET features. This review article summarizes the current state of PET-based radiogenomic research for cancer, which discusses some of its limitations and directions for future study.

Radiographic Image Radiomics Feature Reproducibility: A Preliminary Study on the Impact of Field Size

RATIONALE AND OBJECTIVES

Radiomics is an approach to quantifying diseases. Recently, several studies have indicated that radiomics features are vulnerable against imaging parameters. The aim of this study is to assess how radiomics features change with radiographic field sizes, positions in the field size, and mAs.

MATERIALS AND METHODS

A large and small wood phantom and a cotton phantom were prepared and imaged in different field sizes, mAs, and placement in the radiographic field size. A region of interest was drawn on the image features, and twenty two features were extracted. Radiomics feature reproducibility was obtained based on coefficient of variation, Bland-Altman analysis, and intraclass correlation coefficient. Features with coefficient of variation ≤ 5%, intraclass correlation coefficient ≤ 90%, and 1% ≤ U/LRL ≤30% were introduced as robust features. U/LRL is upper/lower reproducibility limits in Bland-Altman.

RESULTS

For all field sizes and all phantoms, features including Difference Variance, Inverse Different Moment, Fraction, Long Run Emphasis, Run Length Non Uniformity, and Short Run Emphasis were found as highly reproducible features. For change in the position of field size, Fraction was the most reproducible in all field sizes and all phantoms. On the mAs change, we found that feature, Short Run Emphasis field 15 × 15 for small wood phantom, and Correlation in all field sizes for Cotton are the most reproducible features.

CONCLUSION

We demonstrated that radiomics features are strongly vulnerable against radiographic field size, positions in the radiation field, mAs, and phantom materials, and reproducibility analyses should be performed before each radiomics study. Moreover, these changing parameters should be considered, and their effects should be minimized in future radiomics studies.

Overlooked pitfalls in multi-class machine learning classification in radiation oncology and how to avoid them

In radiation oncology, Machine Learning classification publications are typically related to two outcome classes, e.g. the presence or absence of distant metastasis. However, multi-class classification problems also have great clinical relevance, e.g., predicting the grade of a treatment complication following lung irradiation. This work comprised two studies aimed at making work in this domain less prone to statistical blindsides. In multi-class classification, AUC is not defined, whereas correlation coefficients are. It may seem like solely quoting the correlation coefficient value (in lieu of the AUC value) is a suitable choice. In the first study, we illustrated using Monte Carlo (MC) models why this choice is misleading. We also considered the special case where the multiple classes are not ordinal, but nominal, and explained why Pearson or Spearman correlation coefficients are not only providing incomplete information but are actually meaningless. The second study concerned surrogate biomarkers for a clinical endpoint, which have purported benefits including potential for early assessment, being inexpensive, and being non-invasive. Using a MC experiment, we showed how conclusions derived from surrogate markers can be misleading. The simulated endpoint was radiation toxicity (scale of 0-5). The surrogate marker was the true toxicity grade plus a noise term. Five patient cohorts were simulated, including one control. Two of the cohorts were designed to have a statistically significant difference in toxicity. Under 1000 repeated experiments using the biomarker, these two cohorts were often found to be statistically indistinguishable, with the fraction of such occurrences rising with the level of noise.

Comparison of radiomic features in diagnostic CT images with and without contrast enhancement in the delayed phase for NSCLC patients

PURPOSE

To compare radiomic features extracted from diagnostic computed tomography (CT) images with and without contrast enhancement in delayed phase for non-small cell lung cancer (NSCLC) patients.

METHODS

Diagnostic CT images from 269 tumors [non-contrast CT, 188 (dataset NE); contrast-enhanced CT, 81 (dataset CE)] were enrolled in this study. Eighteen first-order and seventy-five texture features were extracted by setting five bin width levels for CT values. Reproducible features were selected by the intraclass correlation coefficient (ICC). Radiomic features were compared between datasets NE and CE. Subgroup analyses were performed based on the CT acquisition period, exposure value, and patient characteristics.

RESULTS

Eighty features were considered reproducible (0.5 ≤ ICC). Twelve of the sixteen first-order features, independent of the bin width levels, were statistically different between datasets NE and CE (p < 0.05), and the p-values of two first-order features depending on the bin width levels were reduced with narrower bin widths. Sixteen out of sixty-two features showed a significant difference, regardless of the bin width (p < 0.05). There were significant differences between datasets NE and CE with older age, lighter body weight, better performance status, being a smoker, larger gross tumor volume, and tumor location at central region.

CONCLUSIONS

Contrast enhancement in the delayed phase of CT images for NSCLC patients affected some of the radiomic features and the variability of radiomic features due to contrast uptake may depend largely on the patient characteristics.

Image-Guided Radiooncology: The Potential of Radiomics in Clinical Application

Medical imaging plays an imminent role in today's radiation oncology workflow. Predominantly based on semantic image analysis, malignant tumors are diagnosed, staged, and therapy decisions are made. The field of "radiomics" promises to extract complementary, objective information from medical images. In radiomics, predefined quantitative features including intensity statistics, texture, shape, or filtering techniques are combined into statistical or machine learning models to predict clinical or biological outcomes. Alternatively, deep neural networks can directly analyze medical images and provide predictions. A large number of research studies could demonstrate that radiomics prediction models may provide significant benefits in the radiation oncology workflow including diagnostics, tumor characterization, target volume segmentation, prognostic stratification, and prediction of therapy response or treatment-related toxicities. This chapter provides an overview of techniques within the radiomics toolbox, potential clinical application, and current limitations. A literature overview of four selected malignant entities including non-small cell lung cancer, head and neck squamous cell carcinomas, soft tissue sarcomas, and gliomas is given.

Prediction of malignant glioma grades using contrast-enhanced T1-weighted and T2-weighted magnetic resonance images based on a radiomic analysis

We conducted a feasibility study to predict malignant glioma grades via radiomic analysis using contrast-enhanced T1-weighted magnetic resonance images (CE-T1WIs) and T2-weighted magnetic resonance images (T2WIs). We proposed a framework and applied it to CE-T1WIs and T2WIs (with tumor region data) acquired preoperatively from 157 patients with malignant glioma (grade III: 55, grade IV: 102) as the primary dataset and 67 patients with malignant glioma (grade III: 22, grade IV: 45) as the validation dataset. Radiomic features such as size/shape, intensity, histogram, and texture features were extracted from the tumor regions on the CE-T1WIs and T2WIs. The Wilcoxon-Mann-Whitney (WMW) test and least absolute shrinkage and selection operator logistic regression (LASSO-LR) were employed to select the radiomic features. Various machine learning (ML) algorithms were used to construct prediction models for the malignant glioma grades using the selected radiomic features. Leave-one-out cross-validation (LOOCV) was implemented to evaluate the performance of the prediction models in the primary dataset. The selected radiomic features for all folds in the LOOCV of the primary dataset were used to perform an independent validation. As evaluation indices, accuracies, sensitivities, specificities, and values for the area under receiver operating characteristic curve (or simply the area under the curve (AUC)) for all prediction models were calculated. The mean AUC value for all prediction models constructed by the ML algorithms in the LOOCV of the primary dataset was 0.902 ± 0.024 (95% CI (confidence interval), 0.873-0.932). In the independent validation, the mean AUC value for all prediction models was 0.747 ± 0.034 (95% CI, 0.705-0.790). The results of this study suggest that the malignant glioma grades could be sufficiently and easily predicted by preparing the CE-T1WIs, T2WIs, and tumor delineations for each patient. Our proposed framework may be an effective tool for preoperatively grading malignant gliomas.