Moderate evidence
More
Blood test with machine learning spots colorectal cancer in small study with 85.5% accuracy
Study title: Label-free diagnosis of colorectal cancer using a SERS platform based on a modified glass fiber filter membrane combined with machine learning
- Evidence type
- Controlled or validated comparison
- Sample
- n = 102
- Sources
- 1 sourcePeer-reviewed
- Published
- 2 min read
- Truvace record
- TRV-2026-3001cite or check it
Study title: Label-free diagnosis of colorectal cancer using a SERS platform based on a modified glass fiber filter membrane combined with machine learning
Researchers built a flexible sensing surface from a glass fiber filter membrane modified with 4-mercaptophenol and silver nanoparticles, then used it to record surface-enhanced Raman scattering spectra of blood serum without any labels. A support vector machine trained on those spectra separated colorectal cancer patients from healthy controls.
Read the study abstract
A flexible three-dimensional surface-enhanced Raman scattering (SERS) platform using a 4-mercaptophenol (4-MP)-modified glass fiber filter membrane (GFF@4MP) for label-free colorectal cancer (CRC) serum discrimination. Silver nanoparticles were uniformly immobilized onto NaCl-modified porous scaffolds. The 4-MP modifier optimized the interfacial microenvironment and provided specific binding sites, improving target adsorption within SERS "hot spots", reducing non-specific interference, and enhancing spatial unif…

The gain
In a study of serum from 30 colorectal cancer patients and 30 healthy controls, a machine-learning model reading SERS spectra reached 93.3% accuracy internally and 85.5% on an external group of 21 patients and 21 controls.
In brief
On an internal cohort of 30 cancer patients and 30 healthy controls, the model reached 93.3% accuracy (AUC 0.99). On a separate external cohort of 21 patients and 21 controls, accuracy was 85.5% (AUC 0.92), a drop that shows how performance fell on new samples.
The authors present the approach as a minimally invasive screening tool. The cohorts are small, and the work reports no testing in routine screening populations.
Main points
- The platform uses a glass fiber filter membrane modified with 4-mercaptophenol (GFF@4MP) as a flexible three-dimensional SERS substrate.
- Silver nanoparticles were immobilised on NaCl-modified porous scaffolds, and the modifier reduced non-specific interference.
- Spatial uniformity of the signal was reported as a relative standard deviation of 7.63%.
- The internal cohort had 30 colorectal cancer patients and 30 healthy controls; the external cohort had 21 and 21.
- The support vector machine reached 93.3% internal test accuracy (95% CI 89.2-97.5%, AUC 0.99).
- On the external cohort, accuracy was 85.5% (95% CI 81.7-89.0%, AUC 0.92).
The full story
The substrate is made from a porous glass fiber filter membrane. Silver nanoparticles were fixed onto NaCl-modified scaffolds, and the 4-mercaptophenol coating gave specific binding sites that improved capture of target molecules within the SERS hot spots and reduced non-specific signals.
Serum from colorectal cancer patients and healthy controls produced spectra that a support vector machine could classify. Internal test accuracy was 93.3% with an AUC of 0.99, while external validation on 42 further people gave 85.5% accuracy and an AUC of 0.92.
The authors say the interfacial chemical modification improves SERS stability in complex biological fluids and offers a reliable route to minimally invasive colorectal cancer screening.
Sources
1 on file- Microchimica Actadoi:10.1007/s00604-026-08538-x
Talk it over
Accuracy was 93.3% on the internal cohort of 30 patients and 30 controls but 85.5% on the external cohort of 21 and 21. What might explain that drop, and what does it suggest about new samples?
Blood test with machine learning spots colorectal cancer in small study with 85.5% accuracyhttps://truvace.com/item/label-free-diagnosis-of-colorectal-cancer-using-a-sers-platform-based-on-a-modif
Accuracy was 93.3% on the internal cohort of 30 patients and 30 controls but 85.5% on the external cohort of 21 and 21. What might explain that drop, and what does it suggest about new samples?
Answer in the threadThe study compared colorectal cancer patients with healthy controls. Who would be missing from that comparison if the test were used for screening, and how could that change the result?
Answer in the threadThe authors present the SERS serum test as a minimally invasive screening tool. How would you weigh its 85.5% external accuracy against what a screening programme would need before using it?
Answer in the thread
If you have more time
What would you look up to learn how the test would perform in people with other conditions, and what result from a routine screening population would settle the question?
Answer in the threadThe signal's spatial uniformity was reported as a relative standard deviation of 7.63%. How might lab-made membranes behave differently once many clinics use them?
Answer in the thread
ace
What readers say