Evaluation of transfer learning in deep convolutional neural network models for cardiac short axis slice classification

Namgyu Ho,Yoon-Chul Kim

doi:10.1038/s41598-021-81525-9

Abstract

In computer-aided analysis of cardiac MRI data, segmentations of the left ventricle (LV) and myocardium are performed to quantify LV ejection fraction and LV mass, and they are performed after the identification of a short axis slice coverage, where automatic classification of the slice range of interest is preferable. Standard cardiac image post-processing guidelines indicate the importance of the correct identification of a short axis slice range for accurate quantification. We investigated the feasibility of applying transfer learning of deep convolutional neural networks (CNNs) as a means to automatically classify the short axis slice range, as transfer learning is well suited to medical image data where labeled data is scarce and expensive to obtain. The short axis slice images were classified into out-of-apical, apical-to-basal, and out-of-basal, on the basis of short axis slice location in the LV. We developed a custom user interface to conveniently label image slices into one of the three categories for the generation of training data and evaluated the performance of transfer learning in nine popular deep CNNs. Evaluation with unseen test data indicated that among the CNNs the fine-tuned VGG16 produced the highest values in all evaluation categories considered and appeared to be the most appropriate choice for the cardiac slice range classification.

Highlights

In computer-aided analysis of cardiac magnetic resonance imaging (MRI) data, segmentations of the left ventricle (LV) and myocardium are performed to quantify LV ejection fraction and LV mass, and they are performed after the identification of a short axis slice coverage, where automatic classification of the slice range of interest is preferable
In the field of cardiac magnetic resonance imaging (MRI), deep convolutional neural networks (CNNs) are applied to the left ventricle (LV), right ventricle (RV), and myocardial segmentation for automatic quantification of ejection fraction and myocardial m ass[2,3,4]
The F1 score, micro-averaged AUC, accuracy, and soft accuracy of deep CNN models were compared, and the fine-tuned VGG16 resulted in the highest values in all six evaluation categories (Table 2)

Summary

Introduction

In computer-aided analysis of cardiac MRI data, segmentations of the left ventricle (LV) and myocardium are performed to quantify LV ejection fraction and LV mass, and they are performed after the identification of a short axis slice coverage, where automatic classification of the slice range of interest is preferable. We investigated the feasibility of applying transfer learning of deep convolutional neural networks (CNNs) as a means to automatically classify the short axis slice range, as transfer learning is well suited to medical image data where labeled data is scarce and expensive to obtain. Margeta et al demonstrated the use of fine-tuned CNNs for cardiac MRI scan plane r ecognition[8] Their method focused on classifying cardiac MR images into five categories (i.e., 2-chamber, 3-chamber, 4-chamber, LV outflow tract, and short axis). Mormont et al applied transfer learning to digital p athology[20] They extracted features from the convolutional layers of pre-trained networks and applied various machine learning techniques for classification. They applied fine-tuning to top-performing base networks to maximize performance. Other novel designs such as “dense blocks”[28] have achieved comparable results in various datasets with a reduced number of parameters

Methods

Results

Conclusion