Automatic detection of optical nerve lesions using a 3D convolutional neural network

Code for the paper "Automatic detection of optical nerve lesions using a 3D convolutional neural network", published in Neuroimage:Clinical. DOI: https://doi.org/10.1016/j.nicl.2022.103187

Abstract

Optic neuritis (ON) is one of the first manifestations of multiple sclerosis, a disabling disease with rising prevalence. Detecting optic nerve lesions could be a relevant diagnostic marker in patients with multiple sclerosis. We aim to create an automated, interpretable method for optic nerve lesion detection from MRI scans. We present a 3D convolutional neural network (CNN) model that learns to detect optic nerve lesions based on T2-weighted fat-saturated MRI scans. We validated our system on two different datasets (N=107 and 62) and interpreted the behaviour of the model using saliency maps. The model showed good performance (68.11% balanced accuracy) that generalizes to unseen data (64.11%). The developed network focuses its attention to the areas that correspond to lesions in the optic nerve. The method shows robustness and, when using only a single imaging sequence, its performance is not far from diagnosis by trained radiologists with the same constraint. Given its speed and performance, the developed methodology could serve as a first step to develop methods that could be translated into a clinical setting.

Graphical Abstract:

Repository organization:

• LesionNerviOpticDataAug/: Code in bash and MATLAB containing the script to segment and extract patches around the optic nerve for a FAT-SAT scan.
• old/: Old, legacy code, preserved this way to suboptimal github organization. Contains old versions of the code which can be useful to directly compare to the current one.
• cm_auc_creation.ipynb: Jupyter script to generate figures for the paper from the results.
• CNN_hyperparamsearch_TRIO.py: Script for doing an hyperparameter search to select best hyperparameters with the training dataset (D1).
• CNN_clasification_PRISMA_BONA.py: Script for testing on the separate dataset (D2). Shares a lot of code with CNN_hyperparamsearch_TRIO.py
• SVM_val.py and SVM_test.py: Scripts to train and test models for the SVM and RF algorithms described in the paper.
• util_functions.py: contains functions used across the papers.

Credits

• Marcos Frías, for the implementation and design of the experimental procedure, neural network architecture and evaluation.
• Aran Garcia-Vidal, for the optic nerve lesion crop script and graphical interface.