Fully automated segmentation of a hip joint using the patient-specific optimal thresholding and watershed algorithm. (February 2018)

• Record Type: Journal Article
• Authors: Kim, Jung Jin; Nam, Jimin; Jang, In Gwun
• Is Part Of: Computer methods and programs in biomedicine. Volume 154(2018)
• Page Start: 161
• Page End: 171
• Publication Date: 2018-02
• Subjects: Image segmentation -- Hip joint -- Optimal thresholding -- Watershed algorithm -- Automated segmentation; Medicine -- Computer programs -- Periodicals; Biology -- Computer programs -- Periodicals; Computers -- Periodicals; Medicine -- Periodicals; Médecine -- Logiciels -- Périodiques; Biologie -- Logiciels -- Périodiques; Biology -- Computer programs; Medicine -- Computer programs; Periodicals; Electronic journals; 610.28
• DOI: 10.1016/j.cmpb.2017.11.007
• Languages: English
• ISSNs: 0169-2607

Abstract

Highlights: Fully automated segmentation is proposed using the patient-specific optimal thresholding and watershed algorithm. The patient-specific optimal thresholding offers regional information on the femur but may offer its disconnected boundary. Watershed algorithm always offers closed patches but does not offer regional information on the femur. Using a complementary property between the two methods, a 3D proximal femur can be efficiently and precisely extracted. Clinical case studies demonstrate the performances of the proposed method. Abstract: Background and Objective: Automated segmentation with high accuracy and speed is a prerequisite for FEA-based quantitative assessment with a large population. However, hip joint segmentation has remained challenging due to a narrow articular cartilage and thin cortical bone with a marked interindividual variance. To overcome this challenge, this paper proposes a fully automated segmentation method for a hip joint that uses the complementary characteristics between the thresholding technique and the watershed algorithm. Methods: Using the golden section method and load path algorithm, the proposed method first determines the patient-specific optimal threshold value that enables reliably separating a femur from a pelvis while removing cortical and trabecular bone in the femur at the minimum. This provides regional information on the femur. The watershed algorithm is then used to obtain boundary information on the femur. The proximal femur can be extracted by merging the complementary information on a target image. Results: For eight CT images, compared with the manual segmentation and other segmentation methods, the proposed method offers a high accuracy in terms of the dice overlap coefficient (97.24 ± 0.44%) and average surface distance (0.36 ± 0.07 mm) within a fast timeframe in terms of processing time per slice (1.25 ± 0.27 s). The proposed method also delivers structural behavior which is close to that of the manual segmentation with a small mean of average relative errors of the risk factor (4.99%). Conclusion: The segmentation results show that, without the aid of a prerequisite dataset and users' manual intervention, the proposed method can segment a hip joint as fast as the simplified Kang (SK)-based automated segmentation, while maintaining the segmentation accuracy at a similar level of the snake-based semi-automated segmentation.