#AI reads Urine# Non-Invasive Precise Classification of Glomerular Diseases in Urine Based on Hyperspectral Technology
Published 26 September, 2025
This study, conducted by a team from Shandong University and other institutions, is the first to combine hyperspectral imaging (HSI) technology with the ResNet-50 model to carry out research on non-invasive and accurate classification of glomerular diseases using urine samples. The study collected 120 urine samples (30 samples for each of the 4 subtypes) from patients with glomerular diseases. After processing steps including centrifugation, hyperspectral image acquisition, PCA dimensionality reduction, and data augmentation, the ResNet-50 model was used for classification. The model achieved an average five-fold cross-validation accuracy of 96.8% and an AUC value of 0.982, outperforming models such as VGG-16 and EfficientNet-B3. Although there are limitations including a small sample size, high cost of HSI equipment, and relatively weak recognition performance for IgA nephropathy, the study provides a new method for the non-invasive diagnosis of glomerular diseases. In the future, optimizations can be made by expanding the dataset, exploring lightweight networks, and other means.
J Biophotonics. 2025 Sep 15:e202500208. doi: 10.1002/jbio.202500208.
Youhe Gao
Statement: During the preparation of this work the author(s) used Doubao / AI reading for summarizing the content. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the published article.
For earlier AI Reads Urine articles: