What is it about?
This research demonstrates how combining Raman spectroscopy—a tool that reveals the biochemical fingerprint of tissues—with AI can accurately distinguish between different types of soft tissue sarcomas and normal tissues. By scanning tissue samples from seven patients and collecting thousands of Raman spectra, a custom AI model (ResNet) was trained to classify eight tissue types, achieving an overall accuracy of 97.1%. This method provides an efficient, fast, and non-invasive alternative to traditional margin-assessment techniques, which are often slow and prone to errors.
Featured Image
Photo by Harpreet Singh on Unsplash
Why is it important?
Soft tissue sarcomas (STS) are rare, highly heterogeneous tumors with over 100 histological subtypes. Their diversity makes diagnosis and surgical margin assessment particularly challenging. Current methods (e.g., frozen section histologic analysis) are slow, labor-intensive, and prone to sampling errors. This approach uses Raman spectroscopy paired with AI to minimize delays and human error. It reduces the risk of leaving malignant tissue behind and could lead to better treatment outcomes, faster recovery, and lower recurrence rates.
Perspectives
This study could pave the way for using portable, handheld laser devices in operating rooms for real-time tumor analysis. The large dataset and AI model developed by our team can serve as a reference for future research and have the potential to improve diagnostic speed and accuracy across many types of tissues. Next steps involve testing handheld Raman device during surgery, validating it with larger patient groups, and expanding it to more types of sarcomas and cancers.
Maede Boroji
Read the Original
This page is a summary of: Ex-vivo Raman spectroscopy and AI-based classification of soft tissue sarcomas, PLOS One, September 2025, PLOS,
DOI: 10.1371/journal.pone.0330618.
You can read the full text:
Resources
Contributors
The following have contributed to this page
