Distant metastasis time to event analysis with CNNs in independent head and neck cancer cohorts

Abstract

Deep learning models based on medical images play an increasingly important role for cancer outcome prediction. The standard approach involves usage of convolutional neural networks (CNNs) to automatically extract relevant features from the patient’s image and perform a binary classification of the occurrence of a given clinical endpoint. In this work, a 2D-CNN and a 3D-CNN for the binary classification of distant metastasis (DM) occurrence in head and neck cancer patients were extended to perform time-to-event analysis. The newly built CNNs incorporate censoring information and output DM-free probability curves as a function of time for every patient. In total, 1037 patients were used to build and assess the performance of the time-to-event model. Training and validation was based on 294 patients also used in a previous benchmark classification study while for testing 743 patients from three independent cohorts were used. The best network could reproduce the good results from 3-fold cross validation [Harrell’s concordance indices (HCIs) of 0.78, 0.74 and 0.80] in two out of three testing cohorts (HCIs of 0.88, 0.67 and 0.77). Additionally, the capability of the models for patient stratification into high and low-risk groups was investigated, the CNNs being able to significantly stratify all three testing cohorts. Results suggest that image-based deep learning models show good reliability for DM time-to-event analysis and could be used for treatment personalisation.

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

“…The inclusion of GTV differed from the original research the CNN comes from. In that study [ 30 ], the single slice with the most extensive area of the tumor was used for the 2D-CNN experiments. We decided to include all slices of the GTV also to develop the deep learning model because ovarian masses often have cystic components and considering a single slice may lead to losing many relevant pieces of information.…”

“…As a second approach, we adopted the 2D-CNN built by Lombardo et al [ 30 ], which showed high performance in tumor classification. Image pre-processing was conducted in Python 3.7, while the 2D-CNN was implemented in Tensorflow 2.4.0.…”

Building Reliable Radiomic Models Using Image Perturbation

“…Distant metastasis-free survival, de ned as the interval from the rst day of treatment to the date of the event, was the clinical endpoint in this study to demonstrate the reliability assessment of the radiomic model 27 . Previous studies of binary classi cation models of HNC 25,28 have achieved good prediction results but were limited because the time-to-event was neglected during model development.…”

Zoo: Selecting Transcriptomic and Methylomic Biomarkers by Ensembling Animal-Inspired Swarm Intelligence Feature Selection Algorithms

“…A stratified three-fold cross-validation (S3FCV) strategy [59] was utilized to evaluate the classification performances. The random seed was set to 0.…”