TB-SERS analyzer: Analysis tool for tuberculosis prediction based on Raman spectroscopy with machine learning and convolutional neural network

Jukgarin Eisiri, Chadatan Juntagran, Kanwara Trisakul, Benjawan Kaewseekhao, Noppadon Nuntawong, Chakchai So-In and Kiatichai Faksri

PLOS Computational Biology, 2026, vol. 22, issue 7, 1-16

Abstract: Raman spectroscopy (RS) and surface-enhanced Raman spectroscopy (SERS) are promising technologies that have been applied across various fields, including clinical diagnostics. In the context of tuberculosis (TB) diagnosis, RS/SERS offers significant potential for rapid, non-invasive, and sensitive biomolecular detection. However, no software currently exists that is specifically designed to analyze RS/SERS data for TB diagnosis. Our goal is to develop such a tool by integrating machine learning (ML) and a one-dimensional convolutional neural network (1D-CNN) into a user-friendly graphical user interface (GUI). We introduce TB-SERS Analyzer, a Python-based tool with a GUI for tuberculosis prediction using SERS data. A reference database of 1,000 plasma samples (500 IGRA-positive, 500 IGRA-negative) was established using the interferon-gamma release assay (IGRA). TB-SERS Analyzer allows users to input spectral data and automatically generate TB diagnostic reports. ML and 1D-CNN models were trained and optimized via five-fold stratified cross-validation. We evaluated seven algorithms to identify the most effective method for TB classification. The 1D-CNN model achieved 82.00% sensitivity and 76.00% specificity in the validation set (n = 200). In a blinded external test (n = 20), the model maintained 80.00% sensitivity with 100% specificity. The software comprises four integrated modules: (1) patient data extraction, (2) data preparation, (3) ML and 1D-CNN analysis, and (4) diagnostic report generation. TB-SERS Analyzer demonstrated high efficiency in TB screening, delivering results in under 10 seconds per sample. TB-SERS Analyzer is an effective and accessible tool for TB screening, combining RS/SERS technologies with ML and 1D-CNN models. The software is freely available on GitHub at: https://github.com/jkeisiri/TB-SERS-Analyzer.Author summary: Tuberculosis (TB) remains one of the world’s leading infectious diseases, affecting millions of people annually and posing a major global public health challenge. Early screening and diagnosis are critical for controlling disease transmission and improving patient outcomes; however, many conventional diagnostic methods require specialized equipment, trained personnel, and time-consuming laboratory procedures. Raman spectroscopy has emerged as a promising analytical technology capable of rapidly detecting disease-associated biochemical changes in biological samples without extensive sample preparation. Nevertheless, translating complex spectral data into clinically meaningful diagnostic information remains challenging and often requires advanced computational approaches. To address this need, we developed TB-SERS Analyzer, a user-friendly software platform for tuberculosis prediction based on Raman spectroscopy data. The software integrates spectral preprocessing, data visualization, artificial intelligence-based classification, and automated report generation into a single workflow, enabling users to obtain diagnostic predictions with minimal technical expertise. Using a large collection of blood plasma samples, we evaluated multiple machine-learning and deep-learning approaches and identified a deep learning model that demonstrated promising classification performance for TB detection. By simplifying Raman spectral analysis, TB-SERS Analyzer enhances the accessibility of Raman-based diagnostics and may facilitate future research, clinical translation, and the development of Raman spectroscopy-based approaches for tuberculosis screening.

Date: 2026
References: Add references at CitEc
Citations:

Downloads: (external link)
https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1014397 (text/html)
https://journals.plos.org/ploscompbiol/article/fil ... 14397&type=printable (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:plo:pcbi00:1014397

DOI: 10.1371/journal.pcbi.1014397

Access Statistics for this article

More articles in PLOS Computational Biology from Public Library of Science
Bibliographic data for series maintained by ploscompbiol ().