Despite the unprecedented performance of deep neural networks (DNNs) in computer vision, their clinical application in the diagnosis and prognosis of cancer using medical imaging has been limited. One of the critical challenges for integrating diagnostic DNNs into radiological and oncological applications is their lack of interpretability, preventing clinicians from understanding the model predictions. Therefore, we studied and propose the integration of expert-derived radiomics and DNN-predicted biomarkers in interpretable classifiers, which we refer to as ConRad, for computerized tomography (CT) scans of lung cancer. Importantly, the tumor biomarkers can be predicted from a concept bottleneck model (CBM) such that once trained, our ConRad models do not require labor-intensive and time-consuming biomarkers. In our evaluation and practical application, the only input to ConRad is a segmented CT scan. The proposed model was compared to convolutional neural networks (CNNs) which act as a black box classifier. We further investigated and evaluated all combinations of radiomics, predicted biomarkers and CNN features in five different classifiers. We found the ConRad models using nonlinear SVM and the logistic regression with the Lasso outperformed the others in five-fold cross-validation, with the interpretability of ConRad being its primary advantage. The Lasso is used for feature selection, which substantially reduces the number of nonzero weights while increasing the accuracy. Overall, the proposed ConRad model combines CBM-derived biomarkers and radiomics features in an interpretable ML model which demonstrates excellent performance for lung nodule malignancy classification.
Keywords: artificial intelligence; concept bottleneck; deep learning; explainability; feature engineering; feature selection; interpretability; radiomics; tumor biomarkers.