Serum Based Multiple Disease Detection Using Deep Learning: Recent Trends
R. B. Singh, Swapnil Srivastava, Tanisha Tanisha, Lovely Chauhan, Pratistha Chauhan, Harsh Tyagi
Abstract
Detection of several diseases from a single biofluid in a given time frame is essential for precision healthcare. Human serum, which is made up of proteins, peptides, lipids, and metabolites, defines one’s systemic health. Recent progress in spectroscopy and mass spectrometry with artificial intelligence has shifted serum-based diagnostics towards scalable, data-driven approaches. FTIR, MIR spectroscopy, Raman, and SERS techniques are useful in the sense that they are able to generate biochemical signatures. When integrated with deep learning (DL), it enables highly accurate disease diagnosis. DL architectures, including Convolutional Neural Networks (CNNs), Deep Neural Networks (DNNs), and hybrid ensemble models, have the ability to capture complex spectral patterns. This capacity enables the classification of cancers and disorders related to endocrine glands in some specific scenarios. Several studies demonstrate the potential of DL-driven SERS for multi-cancer screening. One such application is the integration of DL methods with mass spectrometry. It has led to the discovery of a peptide-based biomarker. The frameworks discussed in the study emphasize the adoption of robust preprocessing and the use of explainable AI to ensure clinical reliability. Advancements in the field of serum spectroscopy are required to develop tools that are non-invasive, rapid, and scalable platforms for multi-disease detection.