SUMMARY: Radiomics describes a broad set of computational methods that extract quantitative features from radiographic images. The resulting features can be used to inform imaging diagnosis, prognosis, and therapy response in oncology. However, major challenges remain for methodologic developments to optimize feature extraction and provide rapid information flow in clinical settings. Equally important, to be clinically useful, predictive radiomic properties must be clearly linked to meaningful biologic characteristics and qualitative imaging properties familiar to radiologists. Here we use a cross-disciplinary approach to highlight studies in radiomics. We review brain tumor radiologic studies (eg, imaging interpretation) through computational models (eg, computer vision and machine learning) that provide novel clinical insights. We outline current quantitative image feature extraction and prediction strategies with different levels of available clinical classes for supporting clinical decision-making. We further discuss machine-learning challenges and data opportunities to advance radiomic studies.

Deep learning training typically starts with a random sampling initialization approach to set the weights of trainable layers. Therefore, different and/or uncontrolled weight initialization prevents learning the same model multiple times. Consequently, such models yield different results during testing. However, even with the exact same initialization for the weights, a lack of repeatability, replicability, and reproducibility may still be observed during deep learning for many reasons such as software versions, implementation variations, and hardware differences. In this paper, we study repeatability when training deep learning models for segmentation and classification tasks using U-Net and LeNet-5 architectures in two development environments Pytorch and Keras (with TensorFlow backend). We show that even with the available control of randomization in Keras and TensorFlow, there are uncontrolled randomizations. We also show repeatable results for the same deep learning architectures using the Pytorch deep learning library. Finally, we discuss variations in the implementation of the weight initialization algorithm across deep learning libraries as a source of uncontrolled error in deep learning results.

Lung cancer has a high incidence and mortality rate. Early detection and diagnosis of lung cancers is best achieved with low-dose computed tomography (CT). Classical radiomics features extracted from lung CT images have been shown as able to predict cancer incidence and prognosis. With the advancement of deep learning and convolutional neural networks (CNNs), deep features can be identified to analyze lung CTs for prognosis prediction and diagnosis. Due to a limited number of available images in the medical field, the transfer learning concept can be helpful. Using subsets of participants from the National Lung Screening Trial (NLST), we utilized a transfer learning approach to differentiate lung cancer nodules versus positive controls. We experimented with three different pretrained CNNs for extracting deep features and used five different classifiers. Experiments were also conducted with deep features from different color channels of a pretrained CNN. Selected deep features were combined with radiomics features. A CNN was designed and trained. Combinations of features from pretrained, CNNs trained on NLST data, and classical radiomics were used to build classifiers. The best accuracy (76.79%) was obtained using feature combinations. An area under the receiver operating characteristic curve of 0.87 was obtained using a CNN trained on an augmented NLST data cohort.

Nomogram models consisting of radiomic signatures and lymph node status obtained from pretreatment PET/CT images and T stage have the potential to identify patients in whom standard chemotherapy is... Purpose To determine if quantitative features extracted from pretherapy fluorine 18 fluorodeoxyglucose (18F-FDG) PET/CT estimate prognosis in patients with locally advanced cervical cancer treated ...

We propose an approach for characterizing structural heterogeneity of lung cancer nodules using Computed Tomography Texture Analysis (CTTA). Measures of heterogeneity were used to test the hypothesis that heterogeneity can be used as predictor of nodule malignancy and patient survival. To do this, we use the National Lung Screening Trial (NLST) dataset to determine if heterogeneity can represent differences between nodules in lung cancer and nodules in non-lung cancer patients. 253 participants are in the training set and 207 participants in the test set. To discriminate cancerous from non-cancerous nodules at the time of diagnosis, a combination of heterogeneity and radiomic features were evaluated to produce the best area under receiver operating characteristic curve (AUROC) of 0.85 and accuracy 81.64%. Second, we tested the hypothesis that heterogeneity can predict patient survival. We analyzed 40 patients diagnosed with lung adenocarcinoma (20 short-term and 20 long-term survival patients) using a leave-one-out cross validation approach for performance evaluation. A combination of heterogeneity features and radiomic features produce an AUROC of 0.9 and an accuracy of 85% to discriminate long- and short-term survivors.

A number of recent papers have shown experimental evidence that suggests it is possible to build highly accurate deep neural network models to detect COVID-19 from chest X-ray images. In this paper, we show that good generalization to unseen sources has not been achieved. Experiments with richer data sets than have previously been used show models have high accuracy on seen sources, but poor accuracy on unseen sources. The reason for the disparity is that the convolutional neural network model, which learns features, can focus on differences in X-ray machines or in positioning within the machines, for example. Any feature that a person would clearly rule out is called a confounding feature. Some of the models were trained on COVID-19 image data taken from publications, which may be different than raw images. Some data sets were of pediatric cases with pneumonia where COVID-19 chest X-rays are almost exclusively from adults, so lung size becomes a spurious feature that can be exploited. In this work, we have eliminated many confounding features by working with as close to raw data as possible. Still, deep learned models may leverage source specific confounders to differentiate COVID-19 from pneumonia preventing generalizing to new data sources (i.e. external sites). Our models have achieved an AUC of 1.00 on seen data sources but in the worst case only scored an AUC of 0.38 on unseen ones. This indicates that such models need further assessment/development before they can be broadly clinically deployed. An example of fine-tuning to improve performance at a new site is given.