Non-small cell lung cancer: identifying prognostic imaging biomarkers by leveraging public gene expression microarray data--methods and preliminary results

Proceedings of the Clinical Microbiology Open 2024: artificial intelligence applications in clinical microbiology

The Clinical Microbiology Open (CMO) is a meeting sponsored by the American Society for Microbiology (ASM) in collaboration with its Corporate Council and Clinical and Public Health Microbiology representatives, which is held to discuss topics that are relevant to both industry and practicing clinical microbiologists. The 2024 CMO was held in Oceanside, California on February 1 and 2. Participants included clinical and public health laboratory directors, representatives from government agencies, and biotechnology industry partners. The group engaged in discussions with the theme, "The Lab of the Future." One of the primary topics discussed was artificial intelligence (AI) opportunities in clinical microbiology laboratories. This report summarizes the discussion and sentiment of the group regarding AI tools, opportunities and challenges of AI in clinical laboratories, and potential future directions for AI in clinical microbiology practice.

Impact of harmonization and oversampling methods on radiomics analysis of multi-center imbalanced datasets: application to PET-based prediction of lung cancer subtypes

BACKGROUND

Medical imaging data frequently encounter image-generation heterogeneity and class imbalance properties, challenging strong generalized predictive performances with data-driven machine-learning methods. The purpose of this study was to investigate the impact of harmonization and oversampling methods on multi-center imbalanced datasets, with specific application to PET-based radiomics modeling for histologic subtype prediction in non-small cell lung cancer (NSCLC).

METHODS

The retrospective study included 245 patients with adenocarcinoma (ADC) and 78 patients with squamous cell carcinoma (SCC) from 4 centers. Utilizing 1502 radiomics features per patient, we trained, validated, and tested 4 machine-learning classifiers, to investigate the effect of no harmonization (NoH) or 4 feature harmonization methods, paired with no oversampling (NoO) or 5 oversampling methods on subtype prediction. Model performance was evaluated using the average area under the ROC curve (AUROC) and G-mean via 5 times 5-fold cross-validations. Statistical comparisons of the combined models against baseline (NoH + NoO) were performed for each fold of cross-validation using the DeLong test.

RESULTS

The number of cross-combinations with both AUROC and G-mean outperforming baseline in validation and testing was 15, 4, 2, and 7 (out of 29) for random forest (RF), linear discriminant analysis (LDA), logistic regression (LR), and support vector machine (SVM), respectively. ComBat harmonization combined with oversampling (SMOTE) via RF yielded better performance than baseline (AUROC and G-mean of validation: 0.725 vs. 0.608 and 0.625 vs. 0.398; testing: 0.637 vs. 0.567 and 0.506 vs. 0.287), though statistical significances were not observed.

CONCLUSIONS

Applying harmonization and oversampling methods in multi-center imbalanced datasets can improve NSCLC-subtype prediction, but the effect varies widely across classifiers. We have created open-source comparisons of harmonization and oversampling on different classifiers for comprehensive evaluations in different studies.

Imaging genomics of cancer: a bibliometric analysis and review

BACKGROUND

Imaging genomics is a burgeoning field that seeks to connections between medical imaging and genomic features. It has been widely applied to explore heterogeneity and predict responsiveness and disease progression in cancer. This review aims to assess current applications and advancements of imaging genomics in cancer.

METHODS

Literature on imaging genomics in cancer was retrieved and selected from PubMed, Web of Science, and Embase before July 2024. Detail information of articles, such as systems and imaging features, were extracted and analyzed. Citation information was extracted from Web of Science and Scopus. Additionally, a bibliometric analysis of the included studies was conducted using the Bibliometrix R package and VOSviewer.

RESULTS

A total of 370 articles were included in the study. The annual growth rate of articles on imaging genomics in cancer is 24.88%. China (133) and the USA (107) were the most productive countries. The top 2 keywords plus were "survival" and "classification". The current research mainly focuses on the central nervous system (121) and the genitourinary system (110, including 44 breast cancer articles). Despite different systems utilizing different imaging modalities, more than half of the studies in each system employed radiomics features.

CONCLUSIONS

Publication databases provide data support for imaging genomics research. The development of artificial intelligence algorithms, especially in feature extraction and model construction, has significantly advanced this field. It is conducive to enhancing the related-models' interpretability. Nonetheless, challenges such as the sample size and the standardization of feature extraction and model construction must overcome. And the research trends revealed in this study will guide the development of imaging genomics in the future and contribute to more accurate cancer diagnosis and treatment in the clinic.

Recurrence-free survival prediction for anal squamous cell carcinoma after chemoradiotherapy using planning CT-based radiomics model

OBJECTIVES

Approximately 30% of non-metastatic anal squamous cell carcinoma (ASCC) patients will experience recurrence after chemoradiotherapy (CRT), and currently available clinical variables are poor predictors of treatment response. We aimed to develop a model leveraging information extracted from radiation pretreatment planning CT to predict recurrence-free survival (RFS) in ASCC patients after CRT.

METHODS

Radiomics features were extracted from planning CT images of 96 ASCC patients. Following pre-feature selection, the optimal feature set was selected via step-forward feature selection with a multivariate Cox proportional hazard model. The RFS prediction was generated from a radiomics-clinical combined model based on an optimal feature set with 5 repeats of nested 5-fold cross validation. The risk stratification ability of the proposed model was evaluated with Kaplan-Meier analysis.

RESULTS

Shape- and texture-based radiomics features significantly predicted RFS. Compared to a clinical-only model, radiomics-clinical combined model achieves better performance in the testing cohort with higher concordance index (0.80 vs 0.73) and AUC (0.84 vs 0.78 for 1-year RFS, 0.84 vs 0.79 for 2-year RFS, and 0.85 vs 0.81 for 3-year RFS), leading to distinctive high- and low-risk of recurrence groups (P < .001).

CONCLUSIONS

A treatment planning CT based radiomics and clinical combined model had improved prognostic performance in predicting RFS for ASCC patients treated with CRT as compared to a model using clinical features only.

ADVANCES IN KNOWLEDGE

The use of radiomics from planning CT is promising in assisting in personalized management in ASCC. The study outcomes support the role of planning CT-based radiomics as potential imaging biomarker.

From Images to Genes: Radiogenomics Based on Artificial Intelligence to Achieve Non-Invasive Precision Medicine in Cancer Patients

With the increasing demand for precision medicine in cancer patients, radiogenomics emerges as a promising frontier. Radiogenomics is originally defined as a methodology for associating gene expression information from high-throughput technologies with imaging phenotypes. However, with advancements in medical imaging, high-throughput omics technologies, and artificial intelligence, both the concept and application of radiogenomics have significantly broadened. In this review, the history of radiogenomics is enumerated, related omics technologies, the five basic workflows and their applications across tumors, the role of AI in radiogenomics, the opportunities and challenges from tumor heterogeneity, and the applications of radiogenomics in tumor immune microenvironment. The application of radiogenomics in positron emission tomography and the role of radiogenomics in multi-omics studies is also discussed. Finally, the challenges faced by clinical transformation, along with future trends in this field is discussed.

Advancements in Artificial Intelligence-Enhanced Imaging Diagnostics for the Management of Liver Disease-Applications and Challenges in Personalized Care

Liver disease can significantly impact life expectancy, making early diagnosis and therapeutic intervention critical challenges in medical care. Imaging diagnostics play a crucial role in diagnosing and managing liver diseases. Recently, the application of artificial intelligence (AI) in medical imaging analysis has become indispensable in healthcare. AI, trained on vast datasets of medical images, has sometimes demonstrated diagnostic accuracy that surpasses that of human experts. AI-assisted imaging diagnostics are expected to contribute significantly to the standardization of diagnostic quality. Furthermore, AI has the potential to identify image features that are imperceptible to humans, thereby playing an essential role in clinical decision-making. This capability enables physicians to make more accurate diagnoses and develop effective treatment strategies, ultimately improving patient outcomes. Additionally, AI is anticipated to become a powerful tool in personalized medicine. By integrating individual patient imaging data with clinical information, AI can propose optimal plans for treatment, making it an essential component in the provision of the most appropriate care for each patient. Current reports highlight the advantages of AI in managing liver diseases. As AI technology continues to evolve, it is expected to advance personalized diagnostics and treatments and contribute to overall improvements in healthcare quality.

Combination of FDG PET/CT radiomics and clinical parameters for outcome prediction in patients with non-Hodgkin's lymphoma

PURPOSE

The purposes was to build model incorporating PET +  computed tomography (CT) radiomics features from baseline PET/CT + clinical parameters to predict outcomes in patients with non-Hodgkin lymphomas.

METHODS

Cohort of 138 patients with complete clinical parameters and follow up times of 25.3 months recorded. Textural analysis of PET and manual correlating contouring in CT images analyzed using LIFE X software. Defined outcomes were overall survival (OS), disease free-survival, radiotherapy, and unfavorable response (defined as disease progression) assessed by end of therapy PET/CT or contrast CT. Univariable and multivariable analysis performed to assess association between PET, CT, and clinical.

RESULTS

Male ( P  = 0.030), abnormal lymphocytes ( P  = 0.030), lower value of PET entropy ( P  = 0.030), higher value of SHAPE sphericity ( P  = 0.002) were significantly associated with worse OS. Advanced stage (III or IV, P  = 0.013), abnormal lymphocytes ( P  = 0.032), higher value of CT gray-level run length matrix (GLRLM) LRLGE mean ( P  = 0.010), higher value of PET gray-level co-occurrence matrix energy angular second moment ( P  < 0.001), and neighborhood gray-level different matrix (NGLDM) busyness mean ( P  < 0.001) were significant predictors of shorter DFS. Abnormal lymphocyte ( P  = 0.033), lower value of CT NGLDM coarseness ( P  = 0.082), and higher value of PET GLRLM gray-level nonuniformity zone mean ( P  = 0.040) were significant predictors of unfavorable response to chemotherapy. Area under the curve for the three models (clinical alone, clinical + PET parameters, and clinical + PET + CT parameters) were 0.626, 0.716, and 0.759, respectively.

Histological Subtype Classification of Non-Small Cell Lung Cancer with Radiomics and 3D Convolutional Neural Networks

Non-small cell lung carcinoma (NSCLC) is the most common type of pulmonary cancer, one of the deadliest malignant tumors worldwide. Given the increased emphasis on the precise management of lung cancer, identifying various subtypes of NSCLC has become pivotal for enhancing diagnostic standards and patient prognosis. In response to the challenges presented by traditional clinical diagnostic methods for NSCLC pathology subtypes, which are invasive, rely on physician experience, and consume medical resources, we explore the potential of radiomics and deep learning to automatically and non-invasively identify NSCLC subtypes from computed tomography (CT) images. An integrated model is proposed that investigates both radiomic features and deep learning features and makes comprehensive decisions based on the combination of these two features. To extract deep features, a three-dimensional convolutional neural network (3D CNN) is proposed to fully utilize the 3D nature of CT images while radiomic features are extracted by radiomics. These two types of features are combined and classified with multi-head attention (MHA) in our proposed model. To our knowledge, this is the first work that integrates different learning methods and features from varied sources in histological subtype classification of lung cancer. Experiments are organized on a mixed dataset comprising NSCLC Radiomics and Radiogenomics. The results show that our proposed model achieves 0.88 in accuracy and 0.89 in the area under the receiver operating characteristic curve (AUC) when distinguishing lung adenocarcinoma (ADC) and lung squamous cell carcinoma (SqCC), indicating the potential of being a non-invasive way for predicting histological subtypes of lung cancer.

Machine Learning Enabled Prediction of Biologically Relevant Gene Expression Using CT-Based Radiomic Features in Non-Small Cell Lung Cancer

BACKGROUND

Non-small-cell lung cancer (NSCLC) remains a global health challenge, driving morbidity and mortality. The emerging field of radiogenomics utilizes statistical methods to correlate radiographic tumor features with genomic characteristics from biopsy samples. Radiomic techniques automate the precise extraction of imaging features from tumor regions in radiographic scans, which are subjected to machine learning (ML) to predict genomic attributes.

METHODS

In a retrospective study of two NSCLC patient cohorts separated by 5 years, we performed a radiogenomic analysis of previously disseminated data from 2018 (n = 116) and newly acquired data from 2023 (n = 44) using RNA sequencing and lung CT images. Combining the data from two cohorts post binarization (of gene expression) or batch normalization (of radiomic features) in each cohort proved to be a better approach as compared to training the model on one cohort and validating on the other.

RESULTS

Our ML-based radiogenomic modeling identified specific imaging features-wavelet, three-dimensional local binary patterns, and logarithmic sigma of gray-level variance-as predictive indicators for high (1) vs. low (0) gene expression of pivotal NSCLC-related genes: SLC35C1, BCL2L1, and MAPK1. These genes have recognized implications in a variety of biological pathways and mechanisms of drug resistance pertinent to NSCLC.

CONCLUSION

The successful integration of heterogeneous radiogenomic datasets underscores the potential of imaging biomarkers in uncovering NSCLC biological processes through gene expression profiles.

radMLBench: A dataset collection for benchmarking in radiomics

BACKGROUND

New machine learning methods and techniques are frequently introduced in radiomics, but they are often tested on a single dataset, which makes it challenging to assess their true benefit. Currently, there is a lack of a larger, publicly accessible dataset collection on which such assessments could be performed. In this study, a collection of radiomics datasets with binary outcomes in tabular form was curated to allow benchmarking of machine learning methods and techniques.

METHODS

A variety of journals and online sources were searched to identify tabular radiomics data with binary outcomes, which were then compiled into a homogeneous data collection that is easily accessible via Python. To illustrate the utility of the dataset collection, it was applied to investigate whether feature decorrelation prior to feature selection could improve predictive performance in a radiomics pipeline.

RESULTS

A total of 50 radiomic datasets were collected, with sample sizes ranging from 51 to 969 and 101 to 11165 features. Using this data, it was observed that decorrelating features did not yield any significant improvement on average.

CONCLUSIONS

A large collection of datasets, easily accessible via Python, suitable for benchmarking and evaluating new machine learning techniques and methods was curated. Its utility was exemplified by demonstrating that feature decorrelation prior to feature selection does not, on average, lead to significant performance gains and could be omitted, thereby increasing the robustness and reliability of the radiomics pipeline.

AI-Generated Annotations Dataset for Diverse Cancer Radiology Collections in NCI Image Data Commons

The National Cancer Institute (NCI) Image Data Commons (IDC) offers publicly available cancer radiology collections for cloud computing, crucial for developing advanced imaging tools and algorithms. Despite their potential, these collections are minimally annotated; only 4% of DICOM studies in collections considered in the project had existing segmentation annotations. This project increases the quantity of segmentations in various IDC collections. We produced high-quality, AI-generated imaging annotations dataset of tissues, organs, and/or cancers for 11 distinct IDC image collections. These collections contain images from a variety of modalities, including computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography (PET). The collections cover various body parts, such as the chest, breast, kidneys, prostate, and liver. A portion of the AI annotations were reviewed and corrected by a radiologist to assess the performance of the AI models. Both the AI's and the radiologist's annotations were encoded in conformance to the Digital Imaging and Communications in Medicine (DICOM) standard, allowing for seamless integration into the IDC collections as third-party analysis collections. All the models, images and annotations are publicly accessible.

Imaging biobanks: operational limits, medical-legal and ethical reflections

The extraordinary growth of health technologies has determined an increasing interest in biobanks that represent a unique wealth for research, experimentation, and validation of new therapies. "Human" biobanks are repositories of various types of human biological samples. Through years the paradigm has shifted from spontaneous collections of biological material all over the world to institutional, organized, and well-structured forms. Imaging biobanks represent a novel field and are defined by European Society of Radiology as: "organized databases of medical images, and associated imaging biomarkers shared among multiple researchers, linked to other biorepositories". Modern radiology and nuclear medicine can provide multiple imaging biomarkers, that express the phenotype related to certain diseases, especially in oncology. Imaging biobanks, not a mere catalogue of bioimages associated to clinical data, involve advanced computer technologies to implement the emergent field of radiomics and radiogenomics. Since Europe hosts most of the biobanks, juridical and ethical framework, with a specific referral to Italy, is analyzed. Linking imaging biobanks to traditional ones appears to be a crucial step that needs to be driven by medical imaging community under clear juridical and ethical guidelines.

Network analysis of histopathological image features and genomics data improving prognosis performance in clear cell renal cell carcinoma

INTRODUCTION

Clear cell renal cell carcinoma is the most common type of kidney cancer, but the prediction of prognosis remains a challenge.

METHODS

We collected whole-slide histopathological images, corresponding clinical and genetic information from the The Cancer Imaging Archive and The Cancer Genome Atlas databases and randomly divided patients into training (n = 197) and validation (n = 84) cohorts. After feature extraction by CellProfiler, we used 2 different machine learning techniques (Least Absolute Shrinkage and Selector Operation-regularized Cox and Support Vector Machine-Recursive Feature Elimination) and weighted gene co-expression network analysis to select prognosis-related image features and genes, respectively. These features and genes were integrated into a joint model using random forest and used to create a nomogram that combines other predictive indicators.

RESULTS

A total of 4 overlapped features were identified, represented by the computed histopathological risk score in the random forest model, and showed predictive value for overall survival (test set: 1-year area under the curves (AUC) = 0.726, 3-year AUC = 0.727, and 5-year AUC = 0.764). The histopathological-genetic risk score (HGRS) integrating the genetic information computed performed better than the model that used image features only (test set: 1-year AUC = 0.682, 3-year AUC = 0.734, and 5-year AUC = 0.78). The nomogram (gender, stage, and HGRS) achieved the highest net benefit according to decision curve analysis compared to HGRS or clinical model.

CONCLUSION

This study developed a histopathological-genetic-related nomogram by combining histopathological features and clinical predictors, providing a more comprehensive prognostic assessment for clear cell renal cell carcinoma patients.

A machine learning radiomics based on enhanced computed tomography to predict neoadjuvant immunotherapy for resectable esophageal squamous cell carcinoma

Background

Patients with resectable esophageal squamous cell carcinoma (ESCC) receiving neoadjuvant immunotherapy (NIT) display variable treatment responses. The purpose of this study is to establish and validate a radiomics based on enhanced computed tomography (CT) and combined with clinical data to predict the major pathological response to NIT in ESCC patients.

Methods

This retrospective study included 82 ESCC patients who were randomly divided into the training group (n = 57) and the validation group (n = 25). Radiomic features were derived from the tumor region in enhanced CT images obtained before treatment. After feature reduction and screening, radiomics was established. Logistic regression analysis was conducted to select clinical variables. The predictive model integrating radiomics and clinical data was constructed and presented as a nomogram. Area under curve (AUC) was applied to evaluate the predictive ability of the models, and decision curve analysis (DCA) and calibration curves were performed to test the application of the models.

Results

One clinical data (radiotherapy) and 10 radiomic features were identified and applied for the predictive model. The radiomics integrated with clinical data could achieve excellent predictive performance, with AUC values of 0.93 (95% CI 0.87-0.99) and 0.85 (95% CI 0.69-1.00) in the training group and the validation group, respectively. DCA and calibration curves demonstrated a good clinical feasibility and utility of this model.

Conclusion

Enhanced CT image-based radiomics could predict the response of ESCC patients to NIT with high accuracy and robustness. The developed predictive model offers a valuable tool for assessing treatment efficacy prior to initiating therapy, thus providing individualized treatment regimens for patients.

New Perspectives for Estimating Body Composition From Computed Tomography: Clothing Associated Artifacts

As the value of clinical imaging is expanded through retrospective analyses, it is imperative that all efforts are made to optimize validity. Such considerations for retrospective designs should prioritize factors like naturalistic conditions for observations and measurement replicability, while avoiding sample biases and reliance on strict clinical timelines. Valid methodological approaches are immanent for successful translation from retrospective observational designs into prospective pragmatic research with actionable potential. In particular, thousands of studies have sought to associate clinical outcomes to measures of body composition across diverse patient groups. Post-hoc use of computed tomography (CT) to quantify adiposity and lean tissue characteristics has most frequently involved just a single slice at the level of the third lumbar vertebrae (L3). Abundant in statistical significance and inconsistencies alike, such methods have yet to be implemented or deemed valuable for making real-world clinical decisions. We present herein a concerning perspective, for both magnitude and prevalence, of a widely overlooked source of data variability for this methodology: the hinderance of pants and other tightly fit clothing.

Using Vision Transformer for high robustness and generalization in predicting EGFR mutation status in lung adenocarcinoma

BACKGROUND

Lung adenocarcinoma is a common cause of cancer-related deaths worldwide, and accurate EGFR genotyping is crucial for optimal treatment outcomes. Conventional methods for identifying the EGFR genotype have several limitations. Therefore, we proposed a deep learning model using non-invasive CT images to predict EGFR mutation status with robustness and generalizability.

METHODS

A total of 525 patients were enrolled at the local hospital to serve as the internal data set for model training and validation. In addition, a cohort of 30 patients from the publicly available Cancer Imaging Archive Data Set was selected for external testing. All patients underwent plain chest CT, and their EGFR mutation status labels were categorized as either mutant or wild type. The CT images were analyzed using a self-attention-based ViT-B/16 model to predict the EGFR mutation status, and the model's performance was evaluated. To produce an attention map indicating the suspicious locations of EGFR mutations, Grad-CAM was utilized.

RESULTS

The ViT deep learning model achieved impressive results, with an accuracy of 0.848, an AUC of 0.868, a sensitivity of 0.924, and a specificity of 0.718 on the validation cohort. Furthermore, in the external test cohort, the model achieved comparable performances, with an accuracy of 0.833, an AUC of 0.885, a sensitivity of 0.900, and a specificity of 0.800.

CONCLUSIONS

The ViT model demonstrates a high level of accuracy in predicting the EGFR mutation status of lung adenocarcinoma patients. Moreover, with the aid of attention maps, the model can assist clinicians in making informed clinical decisions.

SAROS: A dataset for whole-body region and organ segmentation in CT imaging

The Sparsely Annotated Region and Organ Segmentation (SAROS) dataset was created using data from The Cancer Imaging Archive (TCIA) to provide a large open-access CT dataset with high-quality annotations of body landmarks. In-house segmentation models were employed to generate annotation proposals on randomly selected cases from TCIA. The dataset includes 13 semantic body region labels (abdominal/thoracic cavity, bones, brain, breast implant, mediastinum, muscle, parotid/submandibular/thyroid glands, pericardium, spinal cord, subcutaneous tissue) and six body part labels (left/right arm/leg, head, torso). Case selection was based on the DICOM series description, gender, and imaging protocol, resulting in 882 patients (438 female) for a total of 900 CTs. Manual review and correction of proposals were conducted in a continuous quality control cycle. Only every fifth axial slice was annotated, yielding 20150 annotated slices from 28 data collections. For the reproducibility on downstream tasks, five cross-validation folds and a test set were pre-defined. The SAROS dataset serves as an open-access resource for training and evaluating novel segmentation models, covering various scanner vendors and diseases.

MRI-based radiomics for predicting histology in malignant salivary gland tumors: methodology and "proof of principle"

Defining the exact histological features of salivary gland malignancies before treatment remains an unsolved problem that compromises the ability to tailor further therapeutic steps individually. Radiomics, a new methodology to extract quantitative information from medical images, could contribute to characterizing the individual cancer phenotype already before treatment in a fast and non-invasive way. Consequently, the standardization and implementation of radiomic analysis in the clinical routine work to predict histology of salivary gland cancer (SGC) could also provide improvements in clinical decision-making. In this study, we aimed to investigate the potential of radiomic features as imaging biomarker to distinguish between high grade and low-grade salivary gland malignancies. We have also investigated the effect of image and feature level harmonization on the performance of radiomic models. For this study, our dual center cohort consisted of 126 patients, with histologically proven SGC, who underwent curative-intent treatment in two tertiary oncology centers. We extracted and analyzed the radiomics features of 120 pre-therapeutic MRI images with gadolinium (T1 sequences), and correlated those with the definitive post-operative histology. In our study the best radiomic model achieved average AUC of 0.66 and balanced accuracy of 0.63. According to the results, there is significant difference between the performance of models based on MRI intensity normalized images + harmonized features and other models (p value < 0.05) which indicates that in case of dealing with heterogeneous dataset, applying the harmonization methods is beneficial. Among radiomic features minimum intensity from first order, and gray level-variance from texture category were frequently selected during multivariate analysis which indicate the potential of these features as being used as imaging biomarker. The present bicentric study presents for the first time the feasibility of implementing MR-based, handcrafted radiomics, based on T1 contrast-enhanced sequences and the ComBat harmonization method in an effort to predict the formal grading of salivary gland carcinoma with satisfactory performance.

Artificial intelligence in liver imaging: methods and applications

Liver disease is regarded as one of the major health threats to humans. Radiographic assessments hold promise in terms of addressing the current demands for precisely diagnosing and treating liver diseases, and artificial intelligence (AI), which excels at automatically making quantitative assessments of complex medical image characteristics, has made great strides regarding the qualitative interpretation of medical imaging by clinicians. Here, we review the current state of medical-imaging-based AI methodologies and their applications concerning the management of liver diseases. We summarize the representative AI methodologies in liver imaging with focusing on deep learning, and illustrate their promising clinical applications across the spectrum of precise liver disease detection, diagnosis and treatment. We also address the current challenges and future perspectives of AI in liver imaging, with an emphasis on feature interpretability, multimodal data integration and multicenter study. Taken together, it is revealed that AI methodologies, together with the large volume of available medical image data, might impact the future of liver disease care.

A comparison of machine learning methods for radiomics modeling in prediction of occult lymph node metastasis in clinical stage IA lung adenocarcinoma patients

Background

Accurate prediction of occult lymph node metastasis (ONM) is an important basis for determining whether lymph node (LN) dissection is necessary in clinical stage IA lung adenocarcinoma patients. The aim of this study is to determine the best machine learning algorithm for radiomics modeling and to compare the performances of the radiomics model, the clinical-radilogical model and the combined model incorporate both radiomics features and clinical-radilogical features in preoperatively predicting ONM in clinical stage IA lung adenocarcinoma patients.

Methods

Patients with clinical stage IA lung adenocarcinoma undergoing curative surgery from one institution were retrospectively recruited and assigned to training and test cohorts. Radiomics features were extracted from the preoperative computed tomography (CT) images of the primary tumor. Seven machine learning algorithms were used to construct radiomics models, and the model with the best performance, evaluated using the area under the curve (AUC), was selected. Univariate and multivariate logistic regression analyses were performed on the clinical-radiological features to identify statistically significant features and to develop a clinical model. The optimal radiomics and clinical models were integrated to build a combined model, and its predictive performance was assessed using receiver operating characteristic curves, Brier score, and decision curve analysis (DCA).

Results

This study included 258 patients who underwent resection (training cohort, n=182; test cohort, n=76). Six radiomics features were identified. Among the seven machine learning algorithms, extreme gradient boosting (XGB) demonstrated the highest performance for radiomics modeling, with an AUC of 0.917. The combined model improved the AUC to 0.933 and achieved a Brier score of 0.092. DCA revealed that the combined model had optimal clinical efficacy.

Conclusions

The superior performance of the combined model, based on XGB algorithm in predicting ONM in patients with clinical stage IA lung adenocarcinoma, might aid surgeons in deciding whether to conduct mediastinal LN dissection and contribute to improve patients' prognosis.

Novel model integrating computed tomography-based image markers with genetic markers for discriminating radiation pneumonitis in patients with unresectable stage III non-small cell lung cancer receiving radiotherapy: a retrospective multi-center radiogenomics study

BACKGROUND

Chemoradiotherapy is a critical treatment for patients with locally advanced and unresectable non-small cell lung cancer (NSCLC), and it is essential to identify high-risk patients as early as possible owing to the high incidence of radiation pneumonitis (RP). Increasing attention is being paid to the effects of endogenous factors for RP. This study aimed to investigate the value of computed tomography (CT)-based radiomics combined with genomics in analyzing the risk of grade ≥ 2 RP in unresectable stage III NSCLC.

METHODS

In this retrospective multi-center observational study, 100 patients with unresectable stage III NSCLC who were treated with chemoradiotherapy were analyzed. Radiomics features of the entire lung were extracted from pre-radiotherapy CT images. The least absolute shrinkage and selection operator algorithm was used for optimal feature selection to calculate the Rad-score for predicting grade ≥ 2 RP. Genomic DNA was extracted from formalin-fixed paraffin-embedded pretreatment biopsy tissues. Univariate and multivariate logistic regression analyses were performed to identify predictors of RP for model development. The area under the receiver operating characteristic curve was used to evaluate the predictive capacity of the model. Statistical comparisons of the area under the curve values between different models were performed using the DeLong test. Calibration and decision curves were used to demonstrate discriminatory and clinical benefit ratios, respectively.

RESULTS

The Rad-score was constructed from nine radiomic features to predict grade ≥ 2 RP. Multivariate analysis demonstrated that histology, Rad-score, and XRCC1 (rs25487) allele mutation were independent high-risk factors correlated with RP. The area under the curve of the integrated model combining clinical factors, radiomics, and genomics was significantly higher than that of any single model (0.827 versus 0.594, 0.738, or 0.641). Calibration and decision curve analyses confirmed the satisfactory clinical feasibility and utility of the nomogram.

CONCLUSION

Histology, Rad-score, and XRCC1 (rs25487) allele mutation could predict grade ≥ 2 RP in patients with locally advanced unresectable NSCLC after chemoradiotherapy, and the integrated model combining clinical factors, radiomics, and genomics demonstrated the best predictive efficacy.

Convolutional neural network-based program to predict lymph node metastasis of non-small cell lung cancer using 18F-FDG PET

PURPOSE

To develop a convolutional neural network (CNN)-based program to analyze maximum intensity projection (MIP) images of 2-deoxy-2-[F-18]fluoro-D-glucose (FDG) positron emission tomography (PET) scans, aimed at predicting lymph node metastasis of non-small cell lung cancer (NSCLC), and to evaluate its effectiveness in providing diagnostic assistance to radiologists.

METHODS

We obtained PET images of NSCLC from public datasets, including those of 435 patients with available N-stage information, which were divided into a training set (n = 304) and a test set (n = 131). We generated 36 maximum intensity projection (MIP) images for each patient. A residual network (ResNet-50)-based CNN was trained using the MIP images of the training set to predict lymph node metastasis. Lymph node metastasis in the test set was predicted by the trained CNN as well as by seven radiologists twice: first without and second with CNN assistance. Diagnostic performance metrics, including accuracy and prediction error (the difference between the truth and the predictions), were calculated, and reading times were recorded.

RESULTS

In the test set, 67 (51%) patients exhibited lymph node metastases and the CNN yielded 0.748 predictive accuracy. With the assistance of the CNN, the prediction error was significantly reduced for six of the seven radiologists although the accuracy did not change significantly. The prediction time was significantly reduced for five of the seven radiologists with the median reduction ratio 38.0%.

CONCLUSION

The CNN-based program could potentially assist radiologists in predicting lymph node metastasis by increasing diagnostic confidence and reducing reading time without affecting diagnostic accuracy, at least in the limited situations using MIP images.

Radiological artificial intelligence - predicting personalized immunotherapy outcomes in lung cancer

Personalized medicine has revolutionized approaches to treatment in the field of lung cancer by enabling therapies to be specific to each patient. However, physicians encounter an immense number of challenges in providing the optimal treatment regimen for the individual given the sheer complexity of clinical aspects such as tumor molecular profile, tumor microenvironment, expected adverse events, acquired or inherent resistance mechanisms, the development of brain metastases, the limited availability of biomarkers and the choice of combination therapy. The integration of innovative next-generation technologies such as deep learning-a subset of machine learning-and radiomics has the potential to transform the field by supporting clinical decision making in cancer treatment and the delivery of precision therapies while integrating numerous clinical considerations. In this review, we present a brief explanation of the available technologies, the benefits of using these technologies in predicting immunotherapy response in lung cancer, and the expected future challenges in the context of precision medicine.

Weakly supervised segmentation models as explainable radiological classifiers for lung tumour detection on CT images

PURPOSE

Interpretability is essential for reliable convolutional neural network (CNN) image classifiers in radiological applications. We describe a weakly supervised segmentation model that learns to delineate the target object, trained with only image-level labels ("image contains object" or "image does not contain object"), presenting a different approach towards explainable object detectors for radiological imaging tasks.

METHODS

A weakly supervised Unet architecture (WSUnet) was trained to learn lung tumour segmentation from image-level labelled data. WSUnet generates voxel probability maps with a Unet and then constructs an image-level prediction by global max-pooling, thereby facilitating image-level training. WSUnet's voxel-level predictions were compared to traditional model interpretation techniques (class activation mapping, integrated gradients and occlusion sensitivity) in CT data from three institutions (training/validation: n = 412; testing: n = 142). Methods were compared using voxel-level discrimination metrics and clinical value was assessed with a clinician preference survey on data from external institutions.

RESULTS

Despite the absence of voxel-level labels in training, WSUnet's voxel-level predictions localised tumours precisely in both validation (precision: 0.77, 95% CI: [0.76-0.80]; dice: 0.43, 95% CI: [0.39-0.46]), and external testing (precision: 0.78, 95% CI: [0.76-0.81]; dice: 0.33, 95% CI: [0.32-0.35]). WSUnet's voxel-level discrimination outperformed the best comparator in validation (area under precision recall curve (AUPR): 0.55, 95% CI: [0.49-0.56] vs. 0.23, 95% CI: [0.21-0.25]) and testing (AUPR: 0.40, 95% CI: [0.38-0.41] vs. 0.36, 95% CI: [0.34-0.37]). Clinicians preferred WSUnet predictions in most instances (clinician preference rate: 0.72 95% CI: [0.68-0.77]).

CONCLUSION

Weakly supervised segmentation is a viable approach by which explainable object detection models may be developed for medical imaging.

CRITICAL RELEVANCE STATEMENT

WSUnet learns to segment images at voxel level, training only with image-level labels. A Unet backbone first generates a voxel-level probability map and then extracts the maximum voxel prediction as the image-level prediction. Thus, training uses only image-level annotations, reducing human workload. WSUnet's voxel-level predictions provide a causally verifiable explanation for its image-level prediction, improving interpretability.

KEY POINTS

• Explainability and interpretability are essential for reliable medical image classifiers. • This study applies weakly supervised segmentation to generate explainable image classifiers. • The weakly supervised Unet inherently explains its image-level predictions at voxel level.

A Radiomics Approach to Identify Immunologically Active Tumor in Patients with Head and Neck Squamous Cell Carcinomas

BACKGROUND

We recently developed a gene-expression-based HOT score to identify the hot/cold phenotype of head and neck squamous cell carcinomas (HNSCCs), which is associated with the response to immunotherapy. Our goal was to determine whether radiomic profiling from computed tomography (CT) scans can distinguish hot and cold HNSCC.

METHOD

We included 113 patients from The Cancer Genome Atlas (TCGA) and 20 patients from the Groupe Hospitalier Pitié-Salpêtrière (GHPS) with HNSCC, all with available pre-treatment CT scans. The hot/cold phenotype was computed for all patients using the HOT score. The IBEX software (version 4.11.9, accessed on 30 march 2020) was used to extract radiomic features from the delineated tumor region in both datasets, and the intraclass correlation coefficient (ICC) was computed to select robust features. Machine learning classifier models were trained and tested in the TCGA dataset and validated using the area under the receiver operator characteristic curve (AUC) in the GHPS cohort.

RESULTS

A total of 144 radiomic features with an ICC >0.9 was selected. An XGBoost model including these selected features showed the best performance prediction of the hot/cold phenotype with AUC = 0.86 in the GHPS validation dataset.

CONCLUSIONS AND RELEVANCE

We identified a relevant radiomic model to capture the overall hot/cold phenotype of HNSCC. This non-invasive approach could help with the identification of patients with HNSCC who may benefit from immunotherapy.

Deep learning, radiomics and radiogenomics applications in the digital breast tomosynthesis: a systematic review

BACKGROUND

Recent advancements in computing power and state-of-the-art algorithms have helped in more accessible and accurate diagnosis of numerous diseases. In addition, the development of de novo areas in imaging science, such as radiomics and radiogenomics, have been adding more to personalize healthcare to stratify patients better. These techniques associate imaging phenotypes with the related disease genes. Various imaging modalities have been used for years to diagnose breast cancer. Nonetheless, digital breast tomosynthesis (DBT), a state-of-the-art technique, has produced promising results comparatively. DBT, a 3D mammography, is replacing conventional 2D mammography rapidly. This technological advancement is key to AI algorithms for accurately interpreting medical images.

OBJECTIVE AND METHODS

This paper presents a comprehensive review of deep learning (DL), radiomics and radiogenomics in breast image analysis. This review focuses on DBT, its extracted synthetic mammography (SM), and full-field digital mammography (FFDM). Furthermore, this survey provides systematic knowledge about DL, radiomics, and radiogenomics for beginners and advanced-level researchers.

RESULTS

A total of 500 articles were identified, with 30 studies included as the set criteria. Parallel benchmarking of radiomics, radiogenomics, and DL models applied to the DBT images could allow clinicians and researchers alike to have greater awareness as they consider clinical deployment or development of new models. This review provides a comprehensive guide to understanding the current state of early breast cancer detection using DBT images.

CONCLUSION

Using this survey, investigators with various backgrounds can easily seek interdisciplinary science and new DL, radiomics, and radiogenomics directions towards DBT.

Integrative radiomics and transcriptomics analyses reveal subtype characterization of non-small cell lung cancer

OBJECTIVES

To assess whether integrative radiomics and transcriptomics analyses could provide novel insights for radiomic features' molecular annotation and effective risk stratification in non-small cell lung cancer (NSCLC).

METHODS

A total of 627 NSCLC patients from three datasets were included. Radiomics features were extracted from segmented 3-dimensional tumour volumes and were z-score normalized for further analysis. In transcriptomics level, 186 pathways and 28 types of immune cells were assessed by using the Gene Set Variation Analysis (GSVA) algorithm. NSCLC patients were categorized into subgroups based on their radiomic features and pathways enrichment scores using consensus clustering. Subgroup-specific radiomics features were used to validate clustering performance and prognostic value. Kaplan-Meier survival analysis with the log-rank test and univariable and multivariable Cox analyses were conducted to explore survival differences among the subgroups.

RESULTS

Three radiotranscriptomics subtypes (RTSs) were identified based on the radiomics and pathways enrichment profiles. The three RTSs were characterized as having specific molecular hallmarks: RTS1 (proliferation subtype), RTS2 (metabolism subtype), and RTS3 (immune activation subtype). RTS3 showed increased infiltration of most immune cells. The RTS stratification strategy was validated in a validation cohort and showed significant prognostic value. Survival analysis demonstrated that the RTS strategy could stratify NSCLC patients according to prognosis (p = 0.009), and the RTS strategy remained an independent prognostic indicator after adjusting for other clinical parameters.

CONCLUSIONS

This radiotranscriptomics study provides a stratification strategy for NSCLC that could provide information for radiomics feature molecular annotation and prognostic prediction.

KEY POINTS

• Radiotranscriptomics subtypes (RTSs) could be used to stratify molecularly heterogeneous patients. • RTSs showed relationships between molecular phenotypes and radiomics features. • The RTS algorithm could be used to identify patients with poor prognosis.

Machine learning with multimodal data for COVID-19

In response to the unprecedented global healthcare crisis of the COVID-19 pandemic, the scientific community has joined forces to tackle the challenges and prepare for future pandemics. Multiple modalities of data have been investigated to understand the nature of COVID-19. In this paper, MIDRC investigators present an overview of the state-of-the-art development of multimodal machine learning for COVID-19 and model assessment considerations for future studies. We begin with a discussion of the lessons learned from radiogenomic studies for cancer diagnosis. We then summarize the multi-modality COVID-19 data investigated in the literature including symptoms and other clinical data, laboratory tests, imaging, pathology, physiology, and other omics data. Publicly available multimodal COVID-19 data provided by MIDRC and other sources are summarized. After an overview of machine learning developments using multimodal data for COVID-19, we present our perspectives on the future development of multimodal machine learning models for COVID-19.

Performance of alternative manual and automated deep learning segmentation techniques for the prediction of benign and malignant lung nodules

Purpose

We aim to evaluate the performance of radiomic biopsy (RB), best-fit bounding box (BB), and a deep-learning-based segmentation method called no-new-U-Net (nnU-Net), compared to the standard full manual (FM) segmentation method for predicting benign and malignant lung nodules using a computed tomography (CT) radiomic machine learning model.

Materials and Methods

A total of 188 CT scans of lung nodules from 2 institutions were used for our study. One radiologist identified and delineated all 188 lung nodules, whereas a second radiologist segmented a subset (n = 20 ) of these nodules. Both radiologists employed FM and RB segmentation methods. BB segmentations were generated computationally from the FM segmentations. The nnU-Net, a deep-learning-based segmentation method, performed automatic nodule detection and segmentation. The time radiologists took to perform segmentations was recorded. Radiomic features were extracted from each segmentation method, and models to predict benign and malignant lung nodules were developed. The Kruskal-Wallis and DeLong tests were used to compare segmentation times and areas under the curve (AUC), respectively.

Results

For the delineation of the FM, RB, and BB segmentations, the two radiologists required a median time (IQR) of 113 (54 to 251.5), 21 (9.25 to 38), and 16 (12 to 64.25) s, respectively (p = 0.04 ). In dataset 1, the mean AUC (95% CI) of the FM, RB, BB, and nnU-Net model were 0.964 (0.96 to 0.968), 0.985 (0.983 to 0.987), 0.961 (0.956 to 0.965), and 0.878 (0.869 to 0.888). In dataset 2, the mean AUC (95% CI) of the FM, RB, BB, and nnU-Net model were 0.717 (0.705 to 0.729), 0.919 (0.913 to 0.924), 0.699 (0.687 to 0.711), and 0.644 (0.632 to 0.657).

Conclusion

Radiomic biopsy-based models outperformed FM and BB models in prediction of benign and malignant lung nodules in two independent datasets while deep-learning segmentation-based models performed similarly to FM and BB. RB could be a more efficient segmentation method, but further validation is needed.

Three dimensional computed tomography texture analysis of pulmonary lesions: Does radiomics allow differentiation between carcinoma, neuroendocrine tumor and organizing pneumonia?

PURPOSE

To investigate whether CT texture analysis allows differentiation between adenocarcinomas, squamous cell carcinomas, carcinoids, small cell lung cancers and organizing pneumonia and between carcinomas and neuroendocrine tumors.

METHOD

This retrospective study included patients 133 patients (30 patients with organizing pneumonia, 30 patients with adenocarcinoma, 30 patients with squamous cell carcinoma, 23 patients with small cell lung cancer, 20 patients with carcinoid), who underwent CT-guided biopsy of the lung and had a corresponding histopathologic diagnosis. Pulmonary lesions were segmented in consensus by two radiologists with and without a threshold of -50HU in three dimensions. Groupwise comparisons were performed to assess for differences between all five above-listed entities and between carcinomas and neuroendocrine tumors.

RESULTS

Pairwise comparisons of the five entities revealed 53 statistically significant texture features when using no HU-threshold and 6 statistically significant features with a threshold of -50HU. The largest AUC (0.818 [95%CI 0.706-0.930]) was found for the feature wavelet-HHH_glszm_SmallAreaEmphasis for discrimination of carcinoid from the other entities when using no HU-threshold. In differentiating neuroendocrine tumors from carcinomas, 173 parameters proved statistically significant when using no HU threshold versus 52 parameters when using a -50HU-threshold. The largest AUC (0.810 [95%CI 0.728-0,893]) was found for the parameter original_glcm_Correlation for discrimination of neuroendocrine tumors from carcinomas when using no HU-threshold.

CONCLUSIONS

CT texture analysis revealed features that differed significantly between malignant pulmonary lesions and organizing pneumonia and between carcinomas and neuroendocrine tumors of the lung. Applying a HU-threshold for segmentation substantially influenced the results of texture analysis.

Radiomics for characterization of the glioma immune microenvironment

Increasing evidence suggests that besides mutational and molecular alterations, the immune component of the tumor microenvironment also substantially impacts tumor behavior and complicates treatment response, particularly to immunotherapies. Although the standard method for characterizing tumor immune profile is through performing integrated genomic analysis on tissue biopsies, the dynamic change in the immune composition of the tumor microenvironment makes this approach not feasible, especially for brain tumors. Radiomics is a rapidly growing field that uses advanced imaging techniques and computational algorithms to extract numerous quantitative features from medical images. Recent advances in machine learning methods are facilitating biological validation of radiomic signatures and allowing them to "mine" for a variety of significant correlates, including genetic, immunologic, and histologic data. Radiomics has the potential to be used as a non-invasive approach to predict the presence and density of immune cells within the microenvironment, as well as to assess the expression of immune-related genes and pathways. This information can be essential for patient stratification, informing treatment decisions and predicting patients' response to immunotherapies. This is particularly important for tumors with difficult surgical access such as gliomas. In this review, we provide an overview of the glioma microenvironment, describe novel approaches for clustering patients based on their tumor immune profile, and discuss the latest progress on utilization of radiomics for immune profiling of glioma based on current literature.

Radiomics in colorectal cancer patients

The main therapeutic options for colorectal cancer are surgical resection and adjuvant chemotherapy in non-metastatic disease. However, the evaluation of the overall adjuvant chemotherapy benefit in patients with a high risk of recurrence is challenging. Radiological images can represent a source of data that can be analyzed by using automated computer-based techniques, working on numerical information coded within Digital Imaging and Communications in Medicine files: This image numerical analysis has been named "radiomics". Radiomics allows the extraction of quantitative features from radiological images, mainly invisible to the naked eye, that can be further analyzed by artificial intelligence algorithms. Radiomics is expanding in oncology to either understand tumor biology or for the development of imaging biomarkers for diagnosis, staging, and prognosis, prediction of treatment response and diseases monitoring and surveillance. Several efforts have been made to develop radiomics signatures for colorectal cancer patient using computed tomography (CT) images with different aims: The preoperative prediction of lymph node metastasis, detecting BRAF and RAS gene mutations. Moreover, the use of delta-radiomics allows the analysis of variations of the radiomics parameters extracted from CT scans performed at different timepoints. Most published studies concerning radiomics and magnetic resonance imaging (MRI) mainly focused on the response of advanced tumors that underwent neoadjuvant therapy. Nodes status is the main determinant of adjuvant chemotherapy. Therefore, several radiomics model based on MRI, especially on T2-weighted images and ADC maps, for the preoperative prediction of nodes metastasis in rectal cancer has been developed. Current studies mostly focused on the applications of radiomics in positron emission tomography/CT for the prediction of survival after curative surgical resection and assessment of response following neoadjuvant chemoradiotherapy. Since colorectal liver metastases develop in about 25% of patients with colorectal carcinoma, the main diagnostic tasks of radiomics should be the detection of synchronous and metachronous lesions. Radiomics could be an additional tool in clinical setting, especially in identifying patients with high-risk disease. Nevertheless, radiomics has numerous shortcomings that make daily use extremely difficult. Further studies are needed to assess performance of radiomics in stratifying patients with high-risk disease.

CT texture analysis reliability in pulmonary lesions: the influence of 3D vs. 2D lesion segmentation and volume definition by a Hounsfield-unit threshold

OBJECTIVE

Reproducibility problems are a known limitation of radiomics. The segmentation of the target lesion plays a critical role in texture analysis variability. This study's aim was to compare the interobserver reliability of manual 2D vs. 3D lung lesion segmentation with and without pre-definition of the volume using a threshold of - 50 HU.

METHODS

Seventy-five patients with histopathologically proven lung lesions (15 patients each with adenocarcinoma, squamous cell carcinoma, small cell lung cancer, carcinoid, and organizing pneumonia) who underwent an unenhanced CT scan of the chest were included. Three radiologists independently segmented each lesion manually in 3D and 2D with and without pre-segmentation volume definition by a HU threshold, and shape parameters and original, Laplacian of Gaussian-filtered, and wavelet-based texture features were derived. To assess interobserver reliability and identify the most robust texture features, intraclass correlation coefficients (ICCs) for different segmentation settings were calculated.

RESULTS

Shape parameters had high reliability (64-79% had excellent and good ICCs). Texture features had weak reliability levels, with the highest ICCs (38% excellent or good) found for original features in 3D segmentation without the use of a HU threshold. A small proportion (4.3-11.5%) of texture features had excellent or good ICC values at all segmentation settings.

CONCLUSION

Interobserver reliability of texture features from CT scans of a heterogeneous collection of manually segmented lung lesions was low with a small proportion of features demonstrating high reliability independent of the segmentation settings. These results indicate a limited applicability of texture analysis and the need to define robust texture features in patients with lung lesions.

KEY POINTS

• Our study showed a low reproducibility of texture features when 3 radiologists independently segmented lung lesions in CT images, which highlights a serious limitation of texture analysis. • Interobserver reliability of texture features was low regardless of whether the lesion was segmented in 2D and 3D with or without a HU threshold. • In contrast to texture features, shape parameters showed a high interobserver reliability when lesions were segmented in 2D vs. 3D with and without a HU threshold of - 50.

Residual neural network with mixed loss based on batch training technique for identification of EGFR mutation status in lung cancer

Epidermal growth factor receptor (EGFR) is the key to targeted therapy with tyrosine kinase inhibitors in lung cancer. Traditional identification of EGFR mutation status requires biopsy and sequence testing, which may not be suitable for certain groups who cannot perform biopsy. In this paper, using easily accessible and non-invasive CT images, the residual neural network (ResNet) with mixed loss based on batch training technique is proposed for identification of EGFR mutation status in lung cancer. In this model, the ResNet is regarded as the baseline for feature extraction to avoid the gradient disappearance. Besides, a new mixed loss based on the batch similarity and the cross entropy is proposed to guide the network to better learn the model parameters. The proposed mixed loss utilizes the similarity among batch samples to evaluate the distribution of training data, which can reduce the similarity of different classes and the difference of the same classes. In the experiments, VGG16Net, DenseNet, ResNet18, ResNet34 and ResNet50 models with the mixed loss are trained on the public CT dataset with 155 patients including EGFR mutation status from TCIA. The trained networks are employed to the collected preoperative CT dataset with 56 patients from the cooperative hospital for validating the efficiency of the proposed models. Experimental results show that the proposed models are more appropriate and effective on the lung cancer dataset for identifying the EGFR mutation status. In these models, the ResNet34 with mixed loss is optimal (accuracy = 81.58%, AUC = 0.8861, sensitivity = 80.02%, specificity = 82.90%).

Data-Driven Radiogenomic Approach for Deciphering Molecular Mechanisms Underlying Imaging Phenotypes in Lung Adenocarcinoma: A Pilot Study

The heterogeneity of lung tumor nodules is reflected in their phenotypic characteristics in radiological images. The radiogenomics field employs quantitative image features combined with transcriptome expression levels to understand tumor heterogeneity molecularly. Due to the different data acquisition techniques for imaging traits and genomic data, establishing meaningful connections poses a challenge. We analyzed 86 image features describing tumor characteristics (such as shape and texture) with the underlying transcriptome and post-transcriptome profiles of 22 lung cancer patients (median age 67.5 years, from 42 to 80 years) to unravel the molecular mechanisms behind tumor phenotypes. As a result, we were able to construct a radiogenomic association map (RAM) linking tumor morphology, shape, texture, and size with gene and miRNA signatures, as well as biological correlates of GO terms and pathways. These indicated possible dependencies between gene and miRNA expression and the evaluated image phenotypes. In particular, the gene ontology processes "regulation of signaling" and "cellular response to organic substance" were shown to be reflected in CT image phenotypes, exhibiting a distinct radiomic signature. Moreover, the gene regulatory networks involving the TFs TAL1, EZH2, and TGFBR2 could reflect how the texture of lung tumors is potentially formed. The combined visualization of transcriptomic and image features suggests that radiogenomic approaches could identify potential image biomarkers for underlying genetic variation, allowing a broader view of the heterogeneity of the tumors. Finally, the proposed methodology could also be adapted to other cancer types to expand our knowledge of the mechanistic interpretability of tumor phenotypes.

Decentralized collaborative multi-institutional PET attenuation and scatter correction using federated deep learning

PURPOSE

Attenuation correction and scatter compensation (AC/SC) are two main steps toward quantitative PET imaging, which remain challenging in PET-only and PET/MRI systems. These can be effectively tackled via deep learning (DL) methods. However, trustworthy, and generalizable DL models commonly require well-curated, heterogeneous, and large datasets from multiple clinical centers. At the same time, owing to legal/ethical issues and privacy concerns, forming a large collective, centralized dataset poses significant challenges. In this work, we aimed to develop a DL-based model in a multicenter setting without direct sharing of data using federated learning (FL) for AC/SC of PET images.

METHODS

Non-attenuation/scatter corrected and CT-based attenuation/scatter corrected (CT-ASC) 18F-FDG PET images of 300 patients were enrolled in this study. The dataset consisted of 6 different centers, each with 50 patients, with scanner, image acquisition, and reconstruction protocols varying across the centers. CT-based ASC PET images served as the standard reference. All images were reviewed to include high-quality and artifact-free PET images. Both corrected and uncorrected PET images were converted to standardized uptake values (SUVs). We used a modified nested U-Net utilizing residual U-block in a U-shape architecture. We evaluated two FL models, namely sequential (FL-SQ) and parallel (FL-PL) and compared their performance with the baseline centralized (CZ) learning model wherein the data were pooled to one server, as well as center-based (CB) models where for each center the model was built and evaluated separately. Data from each center were divided to contribute to training (30 patients), validation (10 patients), and test sets (10 patients). Final evaluations and reports were performed on 60 patients (10 patients from each center).

RESULTS

In terms of percent SUV absolute relative error (ARE%), both FL-SQ (CI:12.21-14.81%) and FL-PL (CI:11.82-13.84%) models demonstrated excellent agreement with the centralized framework (CI:10.32-12.00%), while FL-based algorithms improved model performance by over 11% compared to CB training strategy (CI: 22.34-26.10%). Furthermore, the Mann-Whitney test between different strategies revealed no significant differences between CZ and FL-based algorithms (p-value > 0.05) in center-categorized mode. At the same time, a significant difference was observed between the different training approaches on the overall dataset (p-value < 0.05). In addition, voxel-wise comparison, with respect to reference CT-ASC, exhibited similar performance for images predicted by CZ (R2 = 0.94), FL-SQ (R2 = 0.93), and FL-PL (R2 = 0.92), while CB model achieved a far lower coefficient of determination (R2 = 0.74). Despite the strong correlations between CZ and FL-based methods compared to reference CT-ASC, a slight underestimation of predicted voxel values was observed.

CONCLUSION

Deep learning-based models provide promising results toward quantitative PET image reconstruction. Specifically, we developed two FL models and compared their performance with center-based and centralized models. The proposed FL-based models achieved higher performance compared to center-based models, comparable with centralized models. Our work provided strong empirical evidence that the FL framework can fully benefit from the generalizability and robustness of DL models used for AC/SC in PET, while obviating the need for the direct sharing of datasets between clinical imaging centers.

Construction of a radiogenomic association map of pancreatic ductal adenocarcinoma

BACKGROUND

Pancreatic adenocarcinoma (PDAC) persists as a malignancy with high morbidity and mortality that can benefit from new means to characterize and detect these tumors, such as radiogenomics. In order to address this gap in the literature, constructed a transcriptomic-CT radiogenomic (RG) map for PDAC.

METHODS

In this Institutional Review Board approved study, a cohort of subjects (n = 50) with gene expression profile data paired with histopathologically confirmed resectable or borderline resectable PDAC were identified. Studies with pre-operative contrast-enhanced CT images were independently assessed for a set of 88 predefined imaging features. Microarray gene expression profiling was then carried out on the histopathologically confirmed pancreatic adenocarcinomas and gene networks were constructed using Weighted Gene Correlation Network Analysis (WCGNA) (n = 37). Data were analyzed with bioinformatics analyses, multivariate regression-based methods, and Kaplan-Meier survival analyses.

RESULTS

Survival analyses identified multiple features of interest that were significantly associated with overall survival, including Tumor Height (P = 0.014), Tumor Contour (P = 0.033), Tumor-stroma Interface (P = 0.014), and the Tumor Enhancement Ratio (P = 0.047). Gene networks for these imaging features were then constructed using WCGNA and further annotated according to the Gene Ontology (GO) annotation framework for a biologically coherent interpretation of the imaging trait-associated gene networks, ultimately resulting in a PDAC RG CT-transcriptome map composed of 3 stage-independent imaging traits enriched in metabolic processes, telomerase activity, and podosome assembly (P < 0.05).

CONCLUSIONS

A CT-transcriptomic RG map for PDAC composed of semantic and quantitative traits with associated biology processes predictive of overall survival, was constructed, that serves as a reference for further mechanistic studies for non-invasive phenotyping of pancreatic tumors.

Synergies of Radiomics and Transcriptomics in Lung Cancer Diagnosis: A Pilot Study

Radiotranscriptomics is an emerging field that aims to investigate the relationships between the radiomic features extracted from medical images and gene expression profiles that contribute in the diagnosis, treatment planning, and prognosis of cancer. This study proposes a methodological framework for the investigation of these associations with application on non-small-cell lung cancer (NSCLC). Six publicly available NSCLC datasets with transcriptomics data were used to derive and validate a transcriptomic signature for its ability to differentiate between cancer and non-malignant lung tissue. A publicly available dataset of 24 NSCLC-diagnosed patients, with both transcriptomic and imaging data, was used for the joint radiotranscriptomic analysis. For each patient, 749 Computed Tomography (CT) radiomic features were extracted and the corresponding transcriptomics data were provided through DNA microarrays. The radiomic features were clustered using the iterative K-means algorithm resulting in 77 homogeneous clusters, represented by meta-radiomic features. The most significant differentially expressed genes (DEGs) were selected by performing Significance Analysis of Microarrays (SAM) and 2-fold change. The interactions among the CT imaging features and the selected DEGs were investigated using SAM and a Spearman rank correlation test with a False Discovery Rate (FDR) of 5%, leading to the extraction of 73 DEGs significantly correlated with radiomic features. These genes were used to produce predictive models of the meta-radiomics features, defined as p-metaomics features, by performing Lasso regression. Of the 77 meta-radiomic features, 51 can be modeled in terms of the transcriptomic signature. These significant radiotranscriptomics relationships form a reliable basis to biologically justify the radiomics features extracted from anatomic imaging modalities. Thus, the biological value of these radiomic features was justified via enrichment analysis on their transcriptomics-based regression models, revealing closely associated biological processes and pathways. Overall, the proposed methodological framework provides joint radiotranscriptomics markers and models to support the connection and complementarities between the transcriptome and the phenotype in cancer, as demonstrated in the case of NSCLC.

Low muscle mass in lung cancer is associated with an inflammatory and immunosuppressive tumor microenvironment

BACKGROUND

Computed tomographies (CT) are useful for identifying muscle loss in non-small lung cancer (NSCLC) cachectic patients. However, we lack consensus on the best cutoff point for pectoralis muscle loss. We aimed to characterize NSCLC patients based on muscularity, clinical data, and the transcriptional profile from the tumor microenvironment to build a cachexia classification model.

METHODS

We used machine learning to generate a muscle loss prediction model, and the tumor's cellular and transcriptional profile was characterized in patients with low muscularity. First, we measured the pectoralis muscle area (PMA) of 211 treatment-naive NSCLC patients using CT available in The Cancer Imaging Archive. The cutoffs were established using machine learning algorithms (CART and Cutoff Finder) on PMA, clinical, and survival data. We evaluated the prediction model in a validation set (36 NSCLC). Tumor RNA-Seq (GSE103584) was used to profile the transcriptome and cellular composition based on digital cytometry.

RESULTS

CART demonstrated that a lower PMA was associated with a high risk of death (HR = 1.99). Cutoff Finder selected PMA cutoffs separating low-muscularity (LM) patients based on the risk of death (P-value = 0.003; discovery set). The cutoff presented 84% of success in classifying low muscle mass. The high risk of LM patients was also found in the validation set. Tumor RNA-Seq revealed 90 upregulated secretory genes in LM that potentially interact with muscle cell receptors. The LM upregulated genes enriched inflammatory biological processes. Digital cytometry revealed that LM patients presented high proportions of cytotoxic and exhausted CD8+ T cells.

CONCLUSIONS

Our prediction model identified cutoffs that distinguished patients with lower PMA and survival with an inflammatory and immunosuppressive TME enriched with inflammatory factors and CD8+ T cells.

Introduction to radiomics for a clinical audience

Radiomics is a rapidly developing field of research focused on the extraction of quantitative features from medical images, thus converting these digital images into minable, high-dimensional data, which offer unique biological information that can enhance our understanding of disease processes and provide clinical decision support. To date, most radiomics research has been focused on oncological applications; however, it is increasingly being used in a raft of other diseases. This review gives an overview of radiomics for a clinical audience, including the radiomics pipeline and the common pitfalls associated with each stage. Key studies in oncology are presented with a focus on both those that use radiomics analysis alone and those that integrate its use with other multimodal data streams. Importantly, clinical applications outside oncology are also presented. Finally, we conclude by offering a vision for radiomics research in the future, including how it might impact our practice as radiologists.

A Comprehensive Landscape of Imaging Feature-Associated RNA Expression Profiles in Human Breast Tissue

The expression abundance of transcripts in nondiseased breast tissue varies among individuals. The association study of genotypes and imaging phenotypes may help us to understand this individual variation. Since existing reports mainly focus on tumors or lesion areas, the heterogeneity of pathological image features and their correlations with RNA expression profiles for nondiseased tissue are not clear. The aim of this study is to discover the association between the nucleus features and the transcriptome-wide RNAs. We analyzed both microscopic histology images and RNA-sequencing data of 456 breast tissues from the Genotype-Tissue Expression (GTEx) project and constructed an automatic computational framework. We classified all samples into four clusters based on their nucleus morphological features and discovered feature-specific gene sets. The biological pathway analysis was performed on each gene set. The proposed framework evaluates the morphological characteristics of the cell nucleus quantitatively and identifies the associated genes. We found image features that capture population variation in breast tissue associated with RNA expressions, suggesting that the variation in expression pattern affects population variation in the morphological traits of breast tissue. This study provides a comprehensive transcriptome-wide view of imaging-feature-specific RNA expression for healthy breast tissue. Such a framework could also be used for understanding the connection between RNA expression and morphology in other tissues and organs. Pathway analysis indicated that the gene sets we identified were involved in specific biological processes, such as immune processes.

Geometry-invariant abnormality detection

Cancer is a highly heterogeneous condition best visualised in positron emission tomography. Due to this heterogeneity, a general-purpose cancer detection model can be built using unsupervised learning anomaly detection models. While prior work in this field has showcased the efficacy of abnormality detection methods (e.g. Transformer-based), these have shown significant vulnerabilities to differences in data geometry. Changes in image resolution or observed field of view can result in inaccurate predictions, even with significant data pre-processing and augmentation. We propose a new spatial conditioning mechanism that enables models to adapt and learn from varying data geometries, and apply it to a state-of-the-art Vector-Quantized Variational Autoencoder + Transformer abnormality detection model. We showcase that this spatial conditioning mechanism statistically-significantly improves model performance on whole-body data compared to the same model without conditioning, while allowing the model to perform inference at varying data geometries.

Comprehensive characterization of genomic and radiologic features reveals distinct driver patterns of RTK/RAS pathway in ground-glass opacity pulmonary nodules

Ground-glass opacity (GGO)-associated pulmonary nodules have been known as a radiologic feature of early-stage lung cancers and exhibit an indolent biological behavior. However, the correlation between driver genes and radiologic features as well as the immune microenvironment remains poorly understood. We performed a custom 1021-gene panel sequencing of 334 resected pulmonary nodules presenting as GGO from 262 Chinese patients. A total of 130 multiple pulmonary nodules were sampled from 58 patients. Clinical-pathologic and radiologic parameters of these pulmonary nodules were collected. Immunohistochemistry (IHC) and multiplex immunofluorescent staining (mIF) were applied to analyze proliferation and immune cell markers of GGO-associated pulmonary nodules. Compared with pure GGO nodules, mixed GGO nodules were enriched for invasive adenocarcinoma (IAC) (182/216 vs 73/118, P &lt; .001). Eighty-eight percent (294/334) of GGO-associated nodules carried at least one mutation in EGFR/ERBB2/BRAF/KRAS/MAP2K1 of the RTK/RAS signaling pathway, and the alterations in these driver genes were mutually exclusive. The analysis of multifocal pulmonary nodules from the same patient revealed evidence of functional convergence on RTK/RAS pathways. Nodules with ERBB2/BRAF/MAP2K1 mutations tended to be more indolent than those with EGFR and KRAS mutations. IHC and mIF staining showed that KRAS-mutant GGO nodules displayed higher infiltration of CD4+ T cell and CD8+ T cell as well as stronger proliferation and immune inhibitory signals. Our study demonstrates a driver landscape of radiologically detectable GGO-associated pulmonary nodules in Chinese patients and supports that different driver patterns in RTK/RAS pathway are corresponding to different radiologic features.

Artificial intelligence in gastric cancer: applications and challenges

Gastric cancer (GC) is one of the most common malignant tumors with high mortality. Accurate diagnosis and treatment decisions for GC rely heavily on human experts' careful judgments on medical images. However, the improvement of the accuracy is hindered by imaging conditions, limited experience, objective criteria, and inter-observer discrepancies. Recently, the developments of machine learning, especially deep-learning algorithms, have been facilitating computers to extract more information from data automatically. Researchers are exploring the far-reaching applications of artificial intelligence (AI) in various clinical practices, including GC. Herein, we aim to provide a broad framework to summarize current research on AI in GC. In the screening of GC, AI can identify precancerous diseases and assist in early cancer detection with endoscopic examination and pathological confirmation. In the diagnosis of GC, AI can support tumor-node-metastasis (TNM) staging and subtype classification. For treatment decisions, AI can help with surgical margin determination and prognosis prediction. Meanwhile, current approaches are challenged by data scarcity and poor interpretability. To tackle these problems, more regulated data, unified processing procedures, and advanced algorithms are urgently needed to build more accurate and robust AI models for GC.

Incremental benefits of size-zone matrix-based radiomics features for the prognosis of lung adenocarcinoma: advantage of spatial partitioning on tumor evaluation

OBJECTIVES

Prognostic models of lung adenocarcinoma (ADC) can be built using radiomics features from various categories. The size-zone matrix (SZM) features have a strong biological basis related to tumor partitioning, but their incremental benefits have not been fully explored. In our study, we aimed to evaluate the incremental benefits of SZM features for the prognosis of lung ADC.

METHODS

A total of 298 patients were included and their pretreatment computed tomography images were analyzed in fivefold cross-validation. We built a risk model of overall survival using SZM features and compared it with a conventional radiomics risk model and a clinical variable-based risk model. We also compared it with other models incorporating various combinations of SZM features, other radiomics features, and clinical variables. A total of seven risk models were compared and evaluated using the hazard ratio (HR) on the left-out test fold.

RESULTS

As a baseline, the clinical variable risk model showed an HR of 2.739. Combining the radiomics signature with SZM feature was better (HR 4.034) than using radiomics signature alone (HR 3.439). Combining radiomics signature, SZM feature, and clinical variable (HR 6.524) fared better than just combining radiomics signature and clinical variables (HR 4.202). These results confirmed the added benefits of SZM features for prognosis in lung ADC.

CONCLUSION

Combining SZM feature with the radiomics signature was better than using the radiomics signature alone and the benefits of SZM features were maintained when clinical variables were added confirming the incremental benefits of SZM features for lung ADC prognosis.

KEY POINTS

• Size-zone matrix (SZM) features provide incremental benefits for the prognosis of lung adenocarcinoma. • Combining the radiomics signature with SZM features performed better than using a radiomics signature alone.

Artificial intelligence and machine learning in cancer imaging

An increasing array of tools is being developed using artificial intelligence (AI) and machine learning (ML) for cancer imaging. The development of an optimal tool requires multidisciplinary engagement to ensure that the appropriate use case is met, as well as to undertake robust development and testing prior to its adoption into healthcare systems. This multidisciplinary review highlights key developments in the field. We discuss the challenges and opportunities of AI and ML in cancer imaging; considerations for the development of algorithms into tools that can be widely used and disseminated; and the development of the ecosystem needed to promote growth of AI and ML in cancer imaging.

Investigating the Lung Adenocarcinoma Stem Cell Biomarker Expressions Using Machine Learning Approaches

The objective of the study is to look at the activation of stem cell-related markers in lung adenocarcinoma. Utilizing an unsupervised machine learning approach centered on the mRNA expression of pluripotent stem cells as well as its subsequent developed progeny, the mRNA stemness index of further around 500 LUAD patients from The Cancer Genome Atlas dataset was generated. In LUADs, mRNAsi had first been investigated using differential variations, survivability analyses, medical phases, and sexuality. A computational approach is used for identifying cell clusters utilizing fuzzy clustering. There at transcriptional as well as protein stages, the interactions between the genetic markers were investigated. The functionality and processes of the important genes were annotated using expression values. The degree of gene expression related to the clinical symptoms and the likelihood of surviving have also been confirmed. In cancer patients, the mRNAsi genes were highly elevated. In particular, the mRNAsi score rises with advanced trials and varies markedly by sexuality. Within several years, reduced mRNAsi categories will have superior overall survivability in large LUADs. Individuals with chronic LUAD had greater mRNAsi and had reduced average survivability. The important genes and the distinguished categories have been chosen based on their mRNAsi connections. Some of the major genes related to cell proliferating Gene Ontology concepts were found enriched out from the cell cycle Kyoto Encyclopedia of Genes and Genomes (KEGG) process. Specific genes were found to be linked to CSC features. Their activation grew in lockstep with the progression of LUAD's pathology, so these markers appeared amplified in pan-cancers. These important markers were discovered to have substantial connections as a group, suggesting that they could be exploited as drug applications in the therapy of LUAD by suppressing stemness traits.

Noninvasive Determination of the IDH Status of Gliomas Using MRI and MRI-Based Radiomics: Impact on Diagnosis and Prognosis

Gliomas are the most common primary malignant brain tumors in adults. The fifth edition of the WHO Classification of Tumors of the Central Nervous System, published in 2021, provided molecular and practical approaches to CNS tumor taxonomy. Currently, molecular features are essential for differentiating the histological subtypes of gliomas, and recent studies have emphasized the importance of isocitrate dehydrogenase (IDH) mutations in stratifying biologically distinct subgroups of gliomas. IDH plays a significant role in gliomagenesis, and the association of IDH status with prognosis is very clear. Recently, there has been much progress in conventional MR imaging (cMRI), advanced MR imaging (aMRI), and radiomics, which are widely used in the study of gliomas. These advances have resulted in an improved correlation between MR signs and IDH mutation status, which will complement the prediction of the IDH phenotype. Although imaging cannot currently substitute for genetic tests, imaging findings have shown promising signs of diagnosing glioma subtypes and evaluating the efficacy and prognosis of individualized molecular targeted therapy. This review focuses on the correlation between MRI and MRI-based radiomics and IDH gene-phenotype prediction, discussing the value and application of these techniques in the diagnosis and evaluation of the prognosis of gliomas.

Integrated imaging and molecular analysis to decipher tumor microenvironment in the era of immunotherapy

Radiological imaging is an integral component of cancer care, including diagnosis, staging, and treatment response monitoring. It contains rich information about tumor phenotypes that are governed not only by cancer cellintrinsic biological processes but also by the tumor microenvironment, such as the composition and function of tumor-infiltrating immune cells. By analyzing the radiological scans using a quantitative radiomics approach, robust relations between specific imaging and molecular phenotypes can be established. Indeed, a number of studies have demonstrated the feasibility of radiogenomics for predicting intrinsic molecular subtypes and gene expression signatures in breast cancer based on MRI. In parallel, promising results have been shown for inferring the amount of tumor-infiltrating lymphocytes, a key factor for the efficacy of cancer immunotherapy, from standard-of-care radiological images. Compared with the biopsy-based approach, radiogenomics offers a unique avenue to profile the molecular makeup of the tumor and immune microenvironment as well as its evolution in a noninvasive and holistic manner through longitudinal imaging scans. Here, we provide a systematic review of the state of the art radiogenomics studies in the era of immunotherapy and discuss emerging paradigms and opportunities in AI and deep learning approaches. These technical advances are expected to transform the radiogenomics field, leading to the discovery of reliable imaging biomarkers. This will pave the way for their clinical translation to guide precision cancer therapy.

Computed tomography-based radiomics quantification predicts epidermal growth factor receptor mutation status and efficacy of first-line targeted therapy in lung adenocarcinoma

Background

Biomarkers that predict the efficacy of first-line tyrosine kinase inhibitors (TKIs) are pivotal in epidermal growth factor receptor (EGFR) mutant advanced lung adenocarcinoma. Imaging-based biomarkers have attracted much attention in anticancer therapy. This study aims to use the machine learning method to distinguish EGFR mutation status and further explores the predictive role of EGFR mutation-related radiomics features in response to first-line TKIs.

Methods

We retrospectively analyzed pretreatment CT images and clinical information from a cohort of lung adenocarcinomas. We entered the top-ranked features into a support vector machine (SVM) classifier to establish a radiomics signature that predicted EGFR mutation status. Furthermore, we identified the best response-related features based on EGFR mutant-related features in first-line TKI therapy patients. Then we test and validate the predictive effect of the best response-related features for progression-free survival (PFS).

Results

Six hundred ninety-two patients were enrolled in building radiomics signatures. The 13 top-ranked features were input into an SVM classifier to establish the radiomics signature of the training cohort (n = 514), and the predictive score of the radiomics signature was assessed on an independent validation group with 178 patients and obtained an area under the curve (AUC) of 74.13%, an F1 score of 68.29%, a specificity of 79.55%, an accuracy of 70.79%, and a sensitivity of 62.22%. More importantly, the skewness-Low (≤0.882) or 10th percentile-Low group (≤21.132) had a superior partial response (PR) rate than the skewness-High or 10th percentile-High group (p < 0.01). Higher skewness (hazard ratio (HR) = 1.722, p = 0.001) was also found to be significantly associated with worse PFS.

Conclusions

The radiomics signature can be used to predict EGFR mutation status. Skewness may contribute to the stratification of disease progression in lung cancer patients treated with first-line TKIs.