Photo credit: phys.org

Revolutionary Machine Learning Algorithm Enhances Medical Diagnostics

Researchers at Rice University have introduced a groundbreaking machine learning (ML) algorithm that significantly improves the interpretation of “light signatures”—the optical spectra emitted by molecules, materials, and disease biomarkers. This advancement could lead to quicker and more accurate medical diagnoses and analytical processes.

“Imagine detecting early indicators of diseases like Alzheimer’s or COVID-19 merely by illuminating a sample with light,” stated Ziyang Wang, a doctoral student in electrical and computer engineering and lead author of a study published in ACS Nano. “Our research is paving the way for this by training computers to effectively ‘read’ light signals scattered from microscopic molecules.”

Each material or molecule has a unique interaction with light, producing a specific pattern similar to a fingerprint. Optical spectroscopy, which involves shining a laser on materials to observe light interactions, is extensively utilized across chemistry, materials science, and medical fields. However, the complexity and time required to interpret spectral data, particularly in cases where samples exhibit subtle differences, can be a significant barrier. The newly developed algorithm is known as Peak-Sensitive Elastic-net Logistic Regression (PSE-LR) and is uniquely tailored for analyzing light-based data.

“Optical spectra from biological samples reveal insightful information regarding physiological conditions,” noted Wang. “This is crucial since rapid and precise disease detection can enhance treatment methodologies and ultimately save lives. Our approach is not limited to health care; it also aids scientists in exploring new materials, facilitating the creation of advanced sensors and compact diagnostic devices.”

One of the prominent features of PSE-LR is its ability to classify various samples accurately while maintaining clarity about its decision-making process—a challenge for many sophisticated ML models. PSE-LR produces a “feature importance map” that identifies which spectral components were pivotal in making a classification, thus allowing for more straightforward interpretation and validation of results.

“Our algorithm emphasizes the most critical aspects of the signal—those significant peaks,” Wang elaborated, likening PSE-LR to “a detective uncovering hidden clues within light signals.”

The research team rigorously tested PSE-LR against various ML models, documenting superior performance, especially in recognizing subtle or overlapping spectral characteristics.

“Many existing models overlook minute details or are overly complex,” Wang emphasized. “Our goal was to create a model that is both intelligent and understandable.”

PSE-LR demonstrated robust capabilities in a series of real-world assessments, such as identifying extremely low concentrations of the SARS-CoV-2 spike protein in fluid samples, detecting neuroprotective substances in mouse brain tissue, classifying Alzheimer’s disease samples, and differentiating between 2D semiconductors.

“Our tool excels in decoding light-based data, discerning fine signals that traditional techniques often miss,” stated Shengxi Huang, an associate professor in both electrical and computer engineering and materials science and nanoengineering, who is also a corresponding author of the study.

This novel algorithm could lead to the creation of advanced diagnostics, biosensors, or even nanodevices. “These developments have the potential to revolutionize medical diagnostics and materials science, driving us toward a future where intelligent technologies can rapidly identify and address health issues more efficiently,” Wang concluded.

More information: Ziyang Wang et al, Machine Learning Interpretation of Optical Spectroscopy Using Peak-Sensitive Logistic Regression, ACS Nano (2025). DOI: 10.1021/acsnano.4c16037

Citation: Light signature algorithm offers precise insight on viral proteins, brain disease markers and semiconductors (2025, April 28) retrieved 28 April 2025 from https://phys.org/news/2025-04-signature-algorithm-precise-insight-viral.html

Source
phys.org