Advances in Medical Image Segmentation: A Comprehensive Survey with a Focus on Lumbar Spine Applications

TL;DR

Medical image segmentation (MIS) is vital for accurate diagnostics and treatment planning across imaging modalities. This survey comprehensively covers traditional MIS methods, deep learning architectures (including CNNs, FCNs, U-Nets, and Transformer-based approaches), semi-supervised strategies, and emerging trends such as cross-modality fusion and federated learning, with a dedicated lumbar spine case study. It highlights key challenges—data scarcity, domain shift, interpretability, and clinical integration—and discusses uncertainty quantification and lightweight architectures as critical for real-world deployment. By consolidating conventional and cutting-edge techniques, the paper provides a practical, tutorial-oriented roadmap for researchers and clinicians to advance MIS across diverse modalities and anatomies.

Abstract

Medical Image Segmentation (MIS) stands as a cornerstone in medical image analysis, playing a pivotal role in precise diagnostics, treatment planning, and monitoring of various medical conditions. This paper presents a comprehensive and systematic survey of MIS methodologies, bridging the gap between traditional image processing techniques and modern deep learning approaches. The survey encompasses thresholding, edge detection, region-based segmentation, clustering algorithms, and model-based techniques while also delving into state-of-the-art deep learning architectures such as Convolutional Neural Networks (CNNs), Fully Convolutional Networks (FCNs), and the widely adopted U-Net and its variants. Moreover, integrating attention mechanisms, semi-supervised learning, generative adversarial networks (GANs), and Transformer-based models is thoroughly explored. In addition to covering established methods, this survey highlights emerging trends, including hybrid architectures, cross-modality learning, federated and distributed learning frameworks, and active learning strategies, which aim to address challenges such as limited labeled datasets, computational complexity, and model generalizability across diverse imaging modalities. Furthermore, a specialized case study on lumbar spine segmentation is presented, offering insights into the challenges and advancements in this relatively underexplored anatomical region. Despite significant progress in the field, critical challenges persist, including dataset bias, domain adaptation, interpretability of deep learning models, and integration into real-world clinical workflows.

Figures (31)

• Figure 1: Medical image segmentation methods summary. Traditional image processing methods are mainly data-driven approaches. Traditional computer vision methods are based on extracting spatial image features. Deep learning methods are commonly based on complex neural networks that can extract image features without prior hand-crafting.
• Figure 2: Traditional methods of image segmentation.
• Figure 3: Classic (a) 2D and (b) 3D CNN and the convolution operation liu2021review.
• Figure 4: Flowchart diagram of squeeze and excitation block hu2018squeeze. GAP stands for global average pooling.
• Figure 5: Structure of FCN used for brain tumor segmentation in MRI long2015fully.