Multi-Model Medical Image Segmentation Using Multi-Stage Generative Adversarial Networks

Abstract

Image segmentation is a challenging problem in medical applications. Medical imaging has become an integral part of machine learning research, as it enables inspecting interior human body with no surgical intervention. Much research has been conducted to study brain segmentation. However, prior studies usually employ one-stage models to segment brain tissues, which could lead to a significant information loss. In this paper, we propose a multi-stage Generative Adversarial Network (GAN ) model to resolve existing issues of one-stage models. To do this, we apply a coarse-to-fine method to improve brain segmentation using a multi-stage GAN . In the first stage, our model generates a coarse outline for both the background and brain tissues. Then, in the second stage, the model generates a refine outline for the white matter (W M ), gray matter (GM ), and cerebrospinal fluid (CSF ). We perform a fusion of the coarse and refine outlines to achieve high results. Despite using very limited data, we obtain an improved Dice Coefficient (DC) accuracy of up to 5% compared to one-stage models. We conclude that our model is more efficient and accurate in practice for brain segmentation of both infants and adults. In addition, we observe that our multi-stage model is 2.69−13.93 minutes faster than prior models. Moreover, our multi-stage model achieves higher performance with only a few-shot learning, in which only limited labeled data is available. Therefore, for medical images, our solution is applicable to a wide range of image segmentation applications for which convolution neural networks and one-stage methods have failed. This helps to advance the process of analyzing brain images, thus providing many advantages to the healthcare system, especially in critical health situations where urgent intervention is needed.

Transfer Learning using Generative Adversarial Networks for MRI Brain Image Segmentation

“…We train and test our GAN model on two datasets of different ages: adults and infants. Our evaluation results show that the proposed model only outperforms two baselines (Standard GAN and 3D,FCN+MIL+G+K [15]) on the three tissues, but also outperforms multi-stage [24] on two tissues. Adopted the transfer learning approach using GT layer improved the results in GM and W M d on the MIC-CAIiSEG dataset and W M on the MRBrains dataset compared with Multi-stage [24].…”

“…Their model was evaluated on the (BRAT S) dataset (2012 and 2018) and showed an improvement in segmentation results. Khaled et al proposed two brain tissues segmentation models, one using FCN+MIL+G+K [15] and another using a multi-stage GAN model [24]. They evaluated their models on two infants and adults brain images and obtained improved segmentation results, expressed by dice coefficients of up to 94% for segmenting GM and WM.…”

Two‐stage coarse‐to‐fine method for pathological images in medical decision‐making systems

“…The network uses a multi-scale self-transformer module to segment the ulcer region, capturing global, long-range pixel dependencies using different levels of multi-scale features and making full use of unlabelled samples. Khaled et al [41] proposed a multi-stage GAN to segment brain tissue to address the significant information loss caused by the one-stage segmentation model. The network applies a coarse-to-fine approach, generating coarse contours for the background and brain tissue However, the above methods still face challenges when faced with the segmentation of pathological images.…”

Two Independent Teachers are Better Role Model

“…Deep learning is on the horizon with the promise to enhance the segmentation of medical images due to the tedious and expensive annotating images [1,2]. Numerous studies have been conducted on various deep learning models that have accomplished a wide variety of tasks [3,4], including brain segmentation, tumor segmentation, etc. Many of these tasks have stringent quality of result requirements.…”