Disentangled global and local features of multi-source data variational autoencoder: An interpretable model for diagnosing IgAN via multi-source Raman spectral fusion techniques

Wei Shuai, Xuecong Tian, Enguang Zuo, Xueqin Zhang, Chen Lu, Jin Gu, Chen Chen, Xiaoyi Lv, Cheng Chen

Abstract: Highlights•Deep learning combined with multi-source Raman spectroscopy for diagnosis of IgAN•Increasing sample size and improving spectral signal-to-noise ratio based on VAE•Latent vectors decouple into globally shared and locally unique representations.•Multi-source data fusion identifies common potential biomarkers.•Interpretable deep learning helps to understand the disease diagnosis process.