Related Experiment Video
Updated: May 26, 2025
07:54
Surface-enhanced Resonance Raman Scattering Nanoprobe Ratiometry for Detecting Microscopic Ovarian Cancer via Folate Receptor Targeting
Published on: March 25, 2019
8.1K
Multi-cancer early detection based on serum surface-enhanced Raman spectroscopy with deep learning: a large-scale
Yuxiang Lin1,2,3, Qiyi Zhang4,5, Hanxi Chen1,2,3
1Department of Breast Surgery, Fujian Medical University Union Hospital, Fuzhou, Fujian Province, 350001, China.
BMC Medicine
|February 21, 2025
Summary
A novel serum-based platform using surface-enhanced Raman spectroscopy (SERS) and deep learning enables sensitive multi-cancer early detection. This approach shows high accuracy, offering a promising new tool for clinical cancer screening.
Area of Science:
- Biomedical Engineering
- Analytical Chemistry
- Oncology
Background:
- Established cancer screening methods have limitations for multi-cancer early detection.
- Early cancer detection significantly improves treatment effectiveness and patient prognosis.
- A need exists for sensitive and accurate pan-cancer screening technologies.
Purpose of the Study:
- To develop and validate a serum-based platform for sensitive and accurate multi-cancer early detection.
- To integrate surface-enhanced Raman spectroscopy (SERS) with advanced data analysis techniques.
- To evaluate the platform's performance in distinguishing various cancer types from healthy controls.
Main Methods:
- Utilized serum samples from 1655 early-stage cancer patients (breast, lung, thyroid, colorectal, gastric, esophageal) and 1896 healthy controls.
- Employed surface-enhanced Raman spectroscopy (SERS) to obtain spectral data.
- Applied data dimension enhancement (heatmap transformation, CWT) and deep learning (ResNet/CNN) with resampling (BorderlineSMOTE) for analysis.
- Used Class Activation Mapping (CAM) for interpretability.
Main Results:
- The deep neural network (DNN) model achieved high performance (e.g., 93.15% accuracy, 0.991 AUC for HC).
- The combined SERS and ResNet (heatmap) approach demonstrated excellent discrimination (e.g., 94.75% accuracy, 0.996 AUC for HC).
- CAM analysis identified key spectral regions relevant to cancer classification.
Conclusions:
- A highly effective serum SERS-based platform for multi-cancer early detection was developed.
- This approach shows significant potential to advance cancer screening in clinical practice.
- The integration of SERS, deep learning, and interpretability offers a novel strategy for early cancer diagnosis.
