Home | Help Center

Endless possibilities in academia

A survey on the application of deep learning in knee joint cartilage ultrasound image segmentation

Jintao Duan1, Miao Zhou2, Yuxiang Wang1, Fangfang Chen1, Liangqing Lin3, Qinghua Wu3, Haipo Cui1


1School of Health Science and Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China. 2Jiangsu Cancer Hospital, Nanjing 213164, Jiangsu Province, China. 3The First Hospital of Putian, Putian 351100, Fujian Province, China.


Address correspondence to: Haipo Cui, School of Health Science and Engineering, University of Shanghai for Science and Technology, No. 516 Jungong Road, Yangpu District, Shanghai 200093, China. E-mail: h_b_cui@163.com.


Received September 27, 2024; Accepted January 22, 2025; Published April 1, 2025


DOI: https://doi.org/10.61189/242224ubhmhy


Highlights

● A systematic review of deep learning (DL) approaches for femoral cartilage segmentation in knee joint ultrasound images.

● Evaluation of popular DL models (U-Net, U-Net++, Siam U-Net, Mask R-CNNs) using various metrics.

● Discussion on dataset scarcity, data preprocessing methods, and future directions for DL-based ultrasound segmentation.

Abstract

Objectives: The femoral cartilage in the knee joint is prone to degenerative changes and injuries, often requir-ing Magnetic Resonance Imaging as the diagnostic gold standard. However, due to the high cost and limited availability of Magnetic Resonance Imaging, ultrasound is explored as a viable alternative. This paper presents a comprehensive review of current deep learning (DL) strategies for knee femoral cartilage segmentation in ultrasound images, focusing on commonly used datasets, data preprocessing techniques, and state-of-the-art DL models. Methods: We systematically reviewed the literature from major medical and engineering databases, sum-marizing key contributions to knee femoral cartilage segmentation. We focused on (1) the scarcity of large-scale public ultrasound datasets and its impact on model training, (2) popular DL architectures (e.g., U-Net variants, Mask Region-based Convolutional Neural Network), and (3) evaluation techniques, particularly the Dice Similarity Coefficient. We also examined image preprocessing and data augmentation strategies aimed at mitigating data insufficiency. Results: Our review shows that U-Net and its variants (e.g., Siam U-Net, U-Net++) commonly achieve competitive Dice Similarity Coefficient values (around 0.70–0.80) for knee cartilage segmentation, despite the limitations in training data. Advanced networks like Mask Region-based Convolutional Neural Network, when com-bined with robust image preprocessing and transfer learning (e.g., using COCO/ImageNet pretrained weights), can improve Dice Similarity Coefficient by over 20%. However, the absence of standardized public datasets limits direct comparisons between studies and affects reproducibility. Conclusion: DL holds significant potential for accurate and cost-effective femoral cartilage segmentation in knee joint ultrasound images, offering a feasible alternative or complement to Magnetic Resonance Imaging-based assessments. However, challenges remain due to the lack of large-scale, open-access ultrasound datasets and inconsistent evaluation protocols. Future work should focus on establishing public benchmarks, refining novel network architectures, and enhancing real-time clinical deploy-ment to foster wider adoption and greater clinical impact.

Keywords: Deep learning, knee joint cartilage, ultrasound image segmentation

Duan JT, Zhou M, Wang YX, Chen FF, Lin LQ, Wu QH, Cui HP. A survey on the application of deep learning in knee joint cartilage ultrasound image segmentation. Med Artif Intell 2025 Apr; 1(1): 1-13.

Copy