A geometric alignment for human temporal bone CT images via lateral semicircular canals segmentation. Issue 10 (14th August 2022)

• Record Type: Journal Article
• Authors: Li, Xiaoguang; Fu, Peng; Yin, Hongxia; Wang, Zhenchang; Zhu, Ziyao; Qin, Yating; Zhuo, Li
• Is Part Of: Medical physics. Volume 49:Issue 10(2022)
• Page Start: 6439
• Page End: 6450
• Publication Date: 2022-08-14
• Subjects: alignment -- deep learning -- lateral semicircular canals -- segmentation -- temporal bone CT image; Medical physics -- Periodicals; Medical physics; Geneeskunde; Natuurkunde; Toepassingen; Biophysics; Periodicals; Periodicals; Electronic journals; 610.153
• DOI: 10.1002/mp.15889
• Languages: English
• ISSNs: 0094-2405

Abstract

Abstract: Purpose: Due to the different posture of the subject and settings of CT scanners, the CT images of the human temporal bone should be geometrically aligned with multiplanar reconstruction to ensure the symmetry of the bilateral anatomical structure. Manual alignment is a time‐consuming task for radiologists and an important preprocessing step for further computer‐aided CT analysis. We propose a fully automatic alignment algorithm for temporal bone CT images via lateral semicircular canals (LSCs) segmentation. Methods: The LSCs are segmented with our proposed multifeature fusion network as anchors at first. Then, we define a standard 3D coordinate system and propose an alignment procedure. Results: The experimental results show that our LSC segmentation network achieved a higher segmentation accuracy. The acceptable rate is achieved 85% over 910 raw temporal bone CT sequences. The alignment speed is reduced from 10 min by manual to 60s. Conclusions: Aiming at the problem of bilateral asymmetry in the raw temporal bone CT images, we propose an automatic geometric alignment method. Our proposed method can help to perform alignment of temporal bone CT images efficiently.