A review of deep learning based methods for medical image multi-organ segmentation

Abstract

Deep learning has revolutionized image processing and achieved the-state-of-art performance in many medical image segmentation tasks. Many deep learning-based methods have been published to segment different parts of the body for different medical applications. It is necessary to summarize the current state of development for deep learning in the field of medical image segmentation. In this paper, we aim to provide a comprehensive review with a focus on multi-organ image segmentation, which is crucial for radiotherapy where the tumor and organs-at-risk need to be contoured for treatment planning. We grouped the surveyed methods into two broad categories which are 'pixel-wise classification' and 'end-to-end segmentation'. Each category was divided into subgroups according to their network design. For each type, we listed the surveyed works, highlighted important contributions and identified specific challenges. Following the detailed review, we discussed the achievements, shortcomings and future potentials of each category. To enable direct comparison, we listed the performance of the surveyed works that used thoracic and head-and-neck benchmark datasets.

Real-world analysis of manual editing of deep learning contouring in the thorax region

“…Deep learning-based (DL) automatic contouring for radiotherapy has evolved over the past years leading to clinical implementation at many institutes and for many treatment sites. Significant time saving compared to both manual contouring and atlas-based contouring methods is achieved [1] , [2] , [3] , [4] , [5] , [6] , [7] , [8] , [9] , [10] . Even though the majority of current clinical tools still use atlas-based auto-contouring, DL-contouring is being offered more often recently [5] , [9] .…”

Segmentation of Melanocytic Lesion Images Using Gamma Correction with Clustering of Keypoint Descriptors

“…It should also be noted that while deep learning methods have shown good promise in object classification challenges because of their learning ability using feature sets, recent literature reports have suggested that their accuracies in the domain of medical image segmentation need further improvement [75]. Deep learning segmentation methods have been reported to also lack pixel-level accuracy without the application of further processing [76,77]. This is primarily because most of them work on the feature level rather than the pixel level for image segmentation.…”

“…This is primarily because most of them work on the feature level rather than the pixel level for image segmentation. In addition, the use of deep learning methods for image segmentation is currently being impaired because of factors such as the need for more datasets for continuous training, lack of memory-efficient models for both training and inference evaluation [76], limited reference information for accurate validation [78], and the possibility of over-fitted results [79]. Considering the 5400 datasets used for evaluation experimentation, we believe our method shows generalization in relation to skin lesion segmentation.…”

AAWS-Net: Anatomy-aware weakly-supervised learning network for breast mass segmentation

“…In medical image segmentation scenarios, semantic segmentation classifies each pixel of a medical image and provides both location and rich morphology features, such as shape type (round, oval, lobular and irregular) and margin type (spiculate, ill-defined, obscured and circumscribed) [ 12 ]. DL based algorithms are the data-driven machine learning paradigm, which has an intrinsic ability to learn deep and discriminative features from data without human intervention, and thence, outperform the traditional image segmentation methods [ 13 ]. Previously, DL based solutions segmented abnormities by sliding the image block with fixed-size which is intercepted by a square window centered on the target pixel.…”