Automated segmentation of osteoblastic vertebral metastasis: a radiomics approach. Issue 6 (2nd November 2021)

• Record Type: Journal Article
• Authors: Clement, Allison; Whyne, Cari; Hardisty, Michael
• Is Part Of: Computer methods in biomechanics and biomedical engineering. Volume 9:Issue 6(2021)
• Page Start: 635
• Page End: 642
• Publication Date: 2021-11-02
• Subjects: Osteoblastic metastasis -- micro-computed tomography -- automated segmentation -- pyradiomics -- machine learning -- random forests classifier; Imaging systems in biology -- Periodicals; Imaging systems in medicine -- Periodicals; Biomechanics -- Data processing -- Periodicals; Biomedical engineering -- Periodicals; 616.0757
• DOI: 10.1080/21681163.2021.1910076
• Languages: English
• ISSNs: 2168-1163

Abstract

ABSTRACT: Skeletal metastasis can be visualised as changes in bone texture. This investigation aimed to develop an automated method to segment spinal osteoblastic metastases in μCT images using radiomic-based feature extraction. Ex vivo μCT images were acquired of rat lumbar vertebrae with osteoblastic metastases. Pyradiomics was used to calculate 3D features for voxels of trabecular bone isolated using an atlas and level-set based approach. A Random Forest Classifier was used to combine features to develop automated predictions for segmenting the osteoblastic lesions. The automated Pyradiomics approach (77 ± 10%) yielded a dice similarity coefficients comparable to the manual intra and inter-user variability (82 ± 15% and 76 ± 15%, respectively).