Cost-effective broad learning-based ultrasound biomicroscopy with 3D reconstruction for ocular anterior segmentation

• Published in: Multimedia tools and applications 2021-11, Vol.80 (28-29), p.35105-35122
• Main Authors: Ali, Saba Ghazanfar, Chen, Yan, Sheng, Bin, Li, Huating, Wu, Qiang, Yang, Po, Muhammad, Khan, Yang, Geng
• Format: Article
• Language: English
• Subjects: Chambers; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Glaucoma; Image reconstruction; Image segmentation; Learning; Medical imaging; Multimedia Information Systems; Optical Coherence Tomography; Special Purpose and Application-Based Systems; Tomography; Ultrasonic imaging
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-020-09303-9

Anterior Chamber Angle (ACA) assessment plays an important role for the diagnosis of glaucoma. Most of the existing techniques relied on Anterior Segment Optical Coherence Tomography (AS-OCT) or Swept Source Optical Coherence Tomography (SS-OCT). We proposed a system for 360° overview of iridocorneal angle of anterior chamber (ICAAC) via Ultrasound Biomicroscopy (UBM). UBM approach acquires the visualization of anterior segment components as well as diseased structures (glaucoma). Our system consists of a new pairing scheme of feature descriptors, i.e. (FREAK, BRISK), (SURF, BRISK) and Broad Learning System (BLS) for 3D reconstruction and segmentation of ICAAC. The 360° overview of 2D ICAAC gives global conception for ACA assessment. 3D images provide a detailed assessment with the amount of opposition’s and synechiae in angle-closure suspects, angle-closure and angle-closure glaucoma in bright light conditions. Extensive evaluations are performed on dataset consists of 650 ICAAC images in five directions of 65 subjects with 10 samples per subject (5 left eye and 5 right eye) from Shanghai Sixth People’s Hospital. Experiments showed that our approach achieves an overall accuracy of 98.72% with training and testing time 29.26( s ), 1.232( s ) respectively.