Abstract:Organ segmentation from medical images is one of the most important pre-processing steps in computer-aided diagnosis, but it is a challenging task because of limited annotated data, low-contrast and non-homogenous textures. Compared with natural images, organs in the medical images have obvious anatomical prior knowledge (e.g., organ shape and position), which can be used to improve the segmentation accuracy. In this paper, we propose a novel segmentation framework which integrates the medical image anatomical prior through loss into the deep learning models. The proposed prior loss function is based on probabilistic atlas, which is called as deep atlas prior (DAP). It includes prior location and shape information of organs, which are important prior information for accurate organ segmentation. Further, we combine the proposed deep atlas prior loss with the conventional likelihood losses such as Dice loss and focal loss into an adaptive Bayesian loss in a Bayesian framework, which consists of a prior and a likelihood. The adaptive Bayesian loss dynamically adjusts the ratio of the DAP loss and the likelihood loss in the training epoch for better learning. The proposed loss function is universal and can be combined with a wide variety of existing deep segmentation models to further enhance their performance. We verify the significance of our proposed framework with some state-of-the-art models, including fully-supervised and semi-supervised segmentation models on a public dataset (ISBI LiTS 2017 Challenge) for liver segmentation and a private dataset for spleen segmentation.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the challenges of organ segmentation in medical images, especially the segmentation difficulties caused by factors such as limited labeled data, low contrast, and non - uniform texture. Unlike natural images, organs in medical images have obvious anatomical prior knowledge (such as organ shape and position), and this information can be used to improve the accuracy of segmentation. Therefore, the author proposes a new segmentation framework, which integrates the anatomical prior knowledge of medical images into the deep - learning model through the loss function. Specifically, the main contributions of the paper include: 1. **Proposing a new segmentation framework**: This framework integrates anatomical prior knowledge into the deep - learning model through the loss function to achieve more accurate organ segmentation. This method can be combined with a variety of existing deep - segmentation models, thereby further enhancing their performance. 2. **Introducing the Deep Atlas Prior (DAP) loss function based on the probability atlas**: The DAP loss function uses the position and shape information in the probability atlas as prior knowledge to help the network better learn the characteristics of organs. To balance the imbalance problem between positive and negative samples, an asymmetric DAP loss is also proposed. 3. **Designing the Bayesian loss function**: Combining the DAP loss with other traditional likelihood losses (such as Dice loss and focal loss) to form a Bayesian loss framework. In addition, an adaptive Bayesian loss is proposed, which can dynamically adjust the proportion of DAP loss and likelihood loss during the training process to achieve better learning results. 4. **Verifying the effectiveness of the method**: Experiments were carried out on public datasets (such as the ISBI LiTS 2017 Challenge) and private datasets to verify the effectiveness of the proposed method in liver and spleen segmentation tasks. The experimental results show that this method can significantly improve the accuracy of segmentation. In summary, this paper aims to solve the challenges faced by organ segmentation in medical images by combining anatomical prior knowledge and deep - learning technology, and proposes specific solutions and experimental verification.

Medical Image Segmentation With Deep Atlas Prior

Semi-supervised Segmentation of Liver Using Adversarial Learning with Deep Atlas Prior.

Segmentation of Cardiac Tissues and Organs for CCTA Images Based on a Deep Learning Model

Segmentation of liver and spleen based on computational anatomy models

Alternative Baselines for Low-Shot 3D Medical Image Segmentation-An Atlas Perspective

Weakly Supervised Segmentation by a Deep Geodesic Prior

Teaching AI the Anatomy Behind the Scan: Addressing Anatomical Flaws in Medical Image Segmentation with Learnable Prior

DCSegNet: Deep Learning Framework Based on Divide-and-Conquer Method for Liver Segmentation

Automatic Whole Heart Segmentation Based on Atlas Propagation with A Priori Anatomical Information

Learning what and where to segment: A new perspective on medical image few-shot segmentation

MTL-ABS3Net: Atlas-Based Semi-Supervised Organ Segmentation Network with Multi-Task Learning for Medical Images

RAP-Net: Coarse-to-Fine Multi-Organ Segmentation with Single Random Anatomical Prior

Hybrid Intelligent-Annotation Organ Segmentation on Medical Datasets

High-level Prior-based Loss Functions for Medical Image Segmentation: A Survey

A deep learning framework for pancreas segmentation with multi-atlas registration and 3D level-set

Adaptive Region-Specific Loss for Improved Medical Image Segmentation

Weakly-Supervised 3D Medical Image Segmentation using Geometric Prior and Contrastive Similarity

Abdominal multi-organ segmentation with cascaded convolutional and adversarial deep networks

Multi-atlas image segmentation for the low-resolution medical images

Organ Segmentation Using Atlas Registration and Fuzzy Connectedness

Comparative clinical evaluation of atlas and deep-learning-based auto-segmentation of organ structures in liver cancer