Effects of Image Size on Deep Learning

TL;DR

This study investigates how image size affects deep learning performance in MRI analysis, introducing new interpolation strategies and demonstrating larger images yield results closer to manual quantification.

Contribution

The paper presents a novel interpolation mask handling strategy and evaluates the impact of image size on MI quantification accuracy in deep learning models.

Findings

Larger LGE MRI images improve MI quantification accuracy.

A new interpolation mask strategy enhances ground truth segmentation.

Quantification results are closer to manual assessments with bigger images.

Abstract

In this work, the best size for late gadolinium enhancement (LGE) magnetic resonance imaging (MRI) images in the training dataset was determined to optimize deep learning training outcomes. Non-extra pixel and extra pixel interpolation algorithms were used to determine the new size of the LGE-MRI images. A novel strategy was introduced to handle interpolation masks and remove extra class labels in interpolated ground truth (GT) segmentation masks. The expectation maximization, weighted intensity, a priori information (EWA) algorithm was used for quantification of myocardial infarction (MI) in automatically segmented LGE-MRI images. Arbitrary threshold, comparison of the sums, and sums of differences are methods used to estimate the relationship between semi-automatic or manual and fully automated quantification of myocardial infarction (MI) results. The relationship between semi-automatic and fully automated quantification of MI results was found to be closer in the case of bigger LGE MRI images (55.5% closer to manual results) than in the case of smaller LGE MRI images (22.2% closer to manual results).