Rapid and Noninvasive Screening of Iga Nephropathy Based on Tear Raman Spectroscopy Combined with Deep Learning Algorithms

Abstract

IgA nephropathy is a common glomerular disease caused by immune complex, and its long-term deterioration will lead to renal failure. The morbidity and mortality of IgA nephropathy are both high. However, if it is detected and treated in time in the early stage, good results will be achieved. Therefore, early detection and timely treatment are particularly important for the cure of IgA nephropathy. At present, the diagnosis of IgA nephropathy is mainly based on serum samples. However, the composition of serum is complex, so it is difficult to identify and analyze the effective components. Compared with serum, the composition of tears is simpler and the extraction operation is convenient. Therefore, a diagnosis method based on tear Raman spectroscopy combined with deep learning algorithm is proposed for the first time in this study, which is used for noninvasive, rapid and accurate screening of IgA nephropathy. In this experiment, tear samples from 17 patients with IgA nephropathy and 48 healthy people were collected, and the Raman spectra of the two types of subjects were compared and analyzed. Then, back propagation neural network (BPNN), GoogleNet and ApproxRepSet are used to establish discriminant diagnosis models. At last, the accuracy of ApproxRepSet is 100% and the auc area is 0.98. The results showed that tear Raman spectroscopy combined with deep learning classification model had a good effect on the screening of IgA nephropathy. This method is expected to realize the early diagnosis of IgAN patients.

Note:
Funding declaration: This work was supported by Tianshan Talent-Young Science and Technology Talent Project (2022TSYCCX0060), Xinjiang Uygur Autonomous Region Youth Science Foundation Project (2022D01C695).

Conflict of Interests: This research does not involve any form of conflict of interest.

Ethical Approval: All tear samples are from the Department of Ophthalmology, people's Hospital of Xinjiang Uygur Autonomous region, and the study of human tear samples has been approved by the Ethics Committee.

Keywords: IgA nephropathy, tears, Raman spectrum, Deep learning