Nuclear cataract classification in anterior segment OCT based on clinical global-local features

Zhang, Xiaoqing1,2; Xiao, Zunjie1,2; Wu, Xiao1,2; Chen, Yu1,2; Higashita, Risa1,2,3; Chen, Wan4; Yuan, Jin4; Liu, Jiang1,2,5,6

2022-09-01

10.1007/s40747-022-00869-5

Complex & Intelligent Systems   Impact Factor & Quartile

2199-4536

2198-6053

Nuclear cataract (NC) is a priority ocular disease of blindness and vision impairment globally. Early intervention and cataract surgery can improve the vision and life quality of NC patients. Anterior segment coherence tomography (AS-OCT) imaging is a non-invasive way to capture the NC opacity objectively and quantitatively. Recent clinical research has shown that there exists a strong opacity correlation relationship between NC severity levels and the mean density on AS-OCT images. In this paper, we present an effective NC classification framework on AS-OCT images, based on feature extraction and feature importance analysis. Motivated by previous clinical knowledge, our method extracts the clinical global-local features, and then applies Pearson's correlation coefficient and recursive feature elimination methods to analyze the feature importance. Finally, an ensemble logistic regression is employed to distinguish NC, which considers different optimization methods' characteristics. A dataset with 11,442 AS-OCT images is collected to evaluate the method. The results show that the proposed method achieves 86.96% accuracy and 88.70% macro-sensitivity, respectively. The performance comparison analysis also demonstrates that the global-local feature extraction method improves about 2% accuracy than the single region-based feature extraction method.

Nuclear cataract Classification Machine learning AS-OCT image Global-local features

SCI

English

First ; Corresponding

Science and Technology Innovation Committee of Shenzhen City["JCYJ20200109140820699","20200925174052004"] ; Guangdong Provincial Department of Education[2020ZDZX3043] ; Guangdong Provincial Key Laboratory[2020B121201001]

Computer Science

Computer Science, Artificial Intelligence

WOS:000854401000001

SPRINGER HEIDELBERG

Web of Science

1.Southern Univ Sci & Technol, Res Inst Trustworthy Autonomous Syst, Shenzhen, Peoples R China
2.Southern Univ Sci & Technol, Dept Comp Sci & Engn, Shenzhen, Peoples R China
3.Tomey Corp, Nagoya, Aichi, Japan
4.Sun Yat Sen Univ, State Key Lab Ophthalmol, Guangzhou, Peoples R China
5.Chinese Acad Sci, Cixi Inst Biomed Engn, Ningbo Inst Mat Technol & Engn, Ningbo, Peoples R China
6.Southern Univ Sci & Technol, Dept Comp Sci & Engn, Guangdong Prov Key Lab Brain Inspired Intelligent, Shenzhen, Peoples R China

Zhang, Xiaoqing,Xiao, Zunjie,Wu, Xiao,et al. Nuclear cataract classification in anterior segment OCT based on clinical global-local features[J]. Complex & Intelligent Systems,2022.

Zhang, Xiaoqing.,Xiao, Zunjie.,Wu, Xiao.,Chen, Yu.,Higashita, Risa.,...&Liu, Jiang.(2022).Nuclear cataract classification in anterior segment OCT based on clinical global-local features.Complex & Intelligent Systems.

Zhang, Xiaoqing,et al."Nuclear cataract classification in anterior segment OCT based on clinical global-local features".Complex & Intelligent Systems (2022).