Hydranet: Data Augmentation for Regression Neural Networks

TL;DR

Hydranet introduces a novel data augmentation technique that recombines existing samples to improve regression neural networks, especially effective in medical imaging with limited labeled data, achieving higher accuracy than traditional methods.

Contribution

The paper presents a new data augmentation method for regression tasks that enhances performance with minimal labeled data, demonstrated on medical imaging applications.

Findings

Improved intraclass correlation coefficients on two medical imaging tasks.

Achieved higher accuracy with fewer training samples compared to traditional augmentation.

Effective in scenarios with scarce labeled data.

Abstract

Deep learning techniques are often criticized to heavily depend on a large quantity of labeled data. This problem is even more challenging in medical image analysis where the annotator expertise is often scarce. We propose a novel data-augmentation method to regularize neural network regressors that learn from a single global label per image. The principle of the method is to create new samples by recombining existing ones. We demonstrate the performance of our algorithm on two tasks: estimation of the number of enlarged perivascular spaces in the basal ganglia, and estimation of white matter hyperintensities volume. We show that the proposed method improves the performance over more basic data augmentation. The proposed method reached an intraclass correlation coefficient between ground truth and network predictions of 0.73 on the first task and 0.84 on the second task, only using between 25 and 30 scans with a single global label per scan for training. With the same number of training scans, more conventional data augmentation methods could only reach intraclass correlation coefficients of 0.68 on the first task, and 0.79 on the second task.