Learning Active Contour Models for Medical Image Segmentation

Image segmentation is an important step in medical image processing and has been widely studied and developed for refinement of clinical analysis and applications. New models based on deep learning have improved results but are restricted to pixel-wise fitting of the segmentation map. Our aim was to tackle this limitation by developing a new model based on deep learning which takes into account the area inside as well as outside the region of interest as well as the size of boundaries during learning. Specifically, we propose a new loss function which incorporates area and size information and integrates this into a dense deep learning model. We evaluated our approach on a dataset of more than 2,000 cardiac MRI scans. Our results show that the proposed loss function outperforms other mainstream loss function Cross-entropy on two common segmentation networks. Our loss function is robust while using different hyperparameter lambda.

1. Introduction

Image segmentation is a fundamental and challenging problem in computer vision, with the aim of partitioning an image in a meaningful way so that objects can be localized, distinguished and/or measured. In medical imaging, this is vital for further clinical analysis, diagnostics, treatment planning and measuring disease progression. High precision is typically required in bio-medical image segmentation [6], [24]. Recently, segmentation techniques based on deep convolutional neural networks (CNNs) have been developed for various medical imaging modalities, such as MRI, CT and X-ray, showing promising results and overcoming the limitations of conventional segmentation methods [17]. During the training process of a CNN model, its parameters are optimized through gradient descent approaches based on the errors measured by a loss function, which compares the prediction and ground truth images. Loss functions are critical for model optimization. In terms of classification problems, the L2 norm is also known as mean squared error (MSE) and cross-entropy (CE) are commonly used as loss functions [8], [33]. CE and the Dice coefficient (DC) have typically been used for segmentation problems [24], [12].