Reducing the Model Variance of a Rectal Cancer Segmentation Network

Abstract

In preoperative imaging, the demarcation of rectal cancer with magnetic resonance images provides an important basis for cancer staging and treatment planning. Recently, deep learning has greatly improved the state-of-the-art method in automatic segmentation. However, limitations in data availability in the medical field can cause large variance and consequent overfitting to medical image segmentation networks. In this study, we propose methods to reduce the model variance of a rectal cancer segmentation network by adding a rectum segmentation task and performing data augmentation; the geometric correlation between the rectum and rectal cancer motivated the former approach. Moreover, we propose a method to perform a bias-variance analysis within an arbitrary region-of-interest (ROI) of a segmentation network, which we applied to assess the efficacy of our approaches in reducing model variance. As a result, adding a rectum segmentation task reduced the model variance of the rectal cancer segmentation network within tumor regions by a factor of 0.90; data augmentation further reduced the variance by a factor of 0.89. These approaches also reduced the training duration by a factor of 0.96 and a further factor of 0.78, respectively. Our approaches will improve the quality of rectal cancer staging by increasing the accuracy of its automatic demarcation and by providing rectum boundary information since rectal cancer staging requires the demarcation of both rectum and rectal cancer. Besides such clinical benefits, our method also enables segmentation networks to be assessed with bias-variance analysis within an arbitrary ROI, such as a cancerous region. INDEX TERMSBias-variance analysis, data augmentation, image segmentation, magnetic resonance imaging (MRI), multi-task learning, neural networks, rectal cancer segmentation, rectum segmentation.

A review of medical image data augmentation techniques for deep learning applications

“…If an image has one or more contours associated with it, the same transformation is applied to the contours. Geometric transformations are so common that they were utilised by 92 of the 93 basic augmentation studies 15–106 …”

“…Patches are generated from the under‐represented class to even the balance. 30 articles made use of cropping 15–17,19,23,32,33,35,41,43,48,50,58,59,66,68,69,71,74,78,80,82,84,85,90,97,98,102,104,105 …”

“…Gamma correction (3j), 107 linear contrast (3k) and histogram equalisation (3l) are common methods to adjust the contrast of an image. Twenty‐eight studies utilised intensity operations for data augmentation 18,20,21,28,34,38,40–42,45,46,55,62,67,70,75–77,79,84,87,88,94,102–104,106,108 …”

“…Some kernels are able to detect and intensify the edges of objects found within the image and have also been used for augmentation (3r). 16 studies reported use of filtering for data augmentation 20,22,24,33,41,49,57,63,72,91,96,98,99,104,106,108 …”

Automatic segmentation of rectal tumor on diffusion‐weighted images by deep learning with U‐Net

“…Recently, convolutional neural networks, particularly U‐shape networks (U‐Net), 11 have been successfully employed in the fully automatic segmentation of medical images. Most of the rectal or colorectal tumor segmentation use CT or T2WI due to the high resolution, high contrast, and high signal‐to‐noise ratio 12–23 . Segmentation on DWI or ADC is rarely reported.…”