A study of CNN capacity applied to Left Venticle Segmentation in Cardiac MRI

TL;DR

This study investigates how CNN depth and dataset size influence left ventricle segmentation accuracy in cardiac MRI, revealing dataset size impacts performance more than network architecture.

Contribution

It introduces a framework for evaluating CNN depth and dataset size effects on cardiac MRI segmentation, comparing shallow and deep U-Net models across various dataset sizes.

Findings

Sample size has a greater impact on performance than network architecture.

In small datasets, hyper-parameters influence results more than network depth.

Performance differences between shallow and deep networks vary across U-Net families.

Abstract

CNN (Convolutional Neural Network) models have been successfully used for segmentation of the left ventricle (LV) in cardiac MRI (Magnetic Resonance Imaging), providing clinical measurements. In practice, two questions arise with deployment of CNNs: 1) when is it better to use a shallow model instead of a deeper one? 2) how the size of a dataset might change the network performance? We propose a framework to answer them, by experimenting with deep and shallow versions of three U-Net families, trained from scratch in six subsets varying from 100 to 10,000 images, different network sizes, learning rates and regularization values. 1620 models were evaluated using 5-fold cross-validation by loss and DICE. The results indicate that: sample size affects performance more than architecture or hyper-parameters; in small samples the performance is more sensitive to hyper-parameters than architecture; the performance difference between shallow and deeper networks is not the same across families.