Head and neck cancer treatment outcome prediction: a comparison between machine learning with conventional radiomics features and deep learning radiomics

Abstract

BackgroundRadiomics can provide in-depth characterization of cancers for treatment outcome prediction. Conventional radiomics rely on extraction of image features within a pre-defined image region of interest (ROI) which are typically fed to a classification algorithm for prediction of a clinical endpoint. Deep learning radiomics allows for a simpler workflow where images can be used directly as input to a convolutional neural network (CNN) with or without a pre-defined ROI.PurposeThe purpose of this study was to evaluate (i) conventional radiomics and (ii) deep learning radiomics for predicting overall survival (OS) and disease-free survival (DFS) for patients with head and neck squamous cell carcinoma (HNSCC) using pre-treatment 18F-fluorodeoxuglucose positron emission tomography (FDG PET) and computed tomography (CT) images.Materials and methodsFDG PET/CT images and clinical data of patients with HNSCC treated with radio(chemo)therapy at Oslo University Hospital (OUS; n = 139) and Maastricht University Medical Center (MAASTRO; n = 99) were collected retrospectively. OUS data was used for model training and initial evaluation. MAASTRO data was used for external testing to assess cross-institutional generalizability. Models trained on clinical and/or conventional radiomics features, with or without feature selection, were compared to CNNs trained on PET/CT images without or with the gross tumor volume (GTV) included. Model performance was measured using accuracy, area under the receiver operating characteristic curve (AUC), Matthew’s correlation coefficient (MCC), and the F1 score calculated for both classes separately.ResultsCNNs trained directly on images achieved the highest performance on external data for both endpoints. Adding both clinical and radiomics features to these image-based models increased performance further. Conventional radiomics including clinical data could achieve competitive performance. However, feature selection on clinical and radiomics data lead to overfitting and poor cross-institutional generalizability. CNNs without tumor and node contours achieved close to on-par performance with CNNs including contours.ConclusionHigh performance and cross-institutional generalizability can be achieved by combining clinical data, radiomics features and medical images together with deep learning models. However, deep learning models trained on images without contours can achieve competitive performance and could see potential use as an initial screening tool for high-risk patients.

Deep Texture Analysis—Enhancing CT Radiomics Features for Prediction of Head and Neck Cancer Treatment Outcomes: A Machine Learning Approach

“…Lastly, it is worth bringing attention to the fact that as an alternative to determining radiomics features and model building, another approach could be to utilize deep learning methods to build models directly from the images. Huynh et al investigated the effectiveness of predictive models using conventional radiomics features with deep learning models and found that CNNs trained on images achieved the highest performance and that adding radiomics and clinical features to these models could enhance the performance further [46]. When testing models with radiomics and clinical features, it was found that they were susceptible to overfitting and, in particular, poor cross-institutional generalizability perhaps due to small sample sizes and variability in data procuring [46].…”

“…Huynh et al investigated the effectiveness of predictive models using conventional radiomics features with deep learning models and found that CNNs trained on images achieved the highest performance and that adding radiomics and clinical features to these models could enhance the performance further [46]. When testing models with radiomics and clinical features, it was found that they were susceptible to overfitting and, in particular, poor cross-institutional generalizability perhaps due to small sample sizes and variability in data procuring [46]. However, although deep learning approaches yield attractive results, they can be considered as "black boxes" with limited transparency and interpretability.…”

Development and Validation of Prognostic Models Using Radiomic Features from Pre-Treatment Positron Emission Tomography (PET) Images in Head and Neck Squamous Cell Carcinoma (HNSCC) Patients

“…Positron emission tomography (PET) imaging is superior to other imaging modalities in identifying locoregional nodal involvement or distant metastasis [ 9 ]. PET radiomic features from primary tumors along with clinical variables may be useful in developing robust prognostic models, predicting and stratifying disease risks, and applying patient-specific treatment strategies [ 10 ].…”

Predicting Response to Exclusive Combined Radio-Chemotherapy in Naso-Oropharyngeal Cancer: The Role of Texture Analysis

“…Texture analysis is an application of radiomics based on the mathematical analysis of the spatial distribution of pixel values within a region of interest (ROI) of a radiological image, which allows for obtaining quantitative information on tissue heterogeneity not otherwise perceivable by the human eye [32,33]. The existence of radiomic features related to different tumor histotypes, different prognoses, and different risks in terms of recurrence or response to therapy makes its application in OPSCC and NPSCC tumors an interesting and promising field of research [34][35][36][37][38][39][40].…”