Truace
Subscribe

ATR-FTIR spectroscopy and machine learning for colorectal cancer risk in IBD biopsies

Source article: Infrared spectroscopy with statistical analysis and machine learning for cancer risk assessment in inflammatory bowel disease patients

Patients with inflammatory bowel disease (IBD) undergo regular colonoscopic surveillance due to their elevated risk of developing colorectal cancer (CRC). However, current clinical, endoscopic and histopathological risk stratification methods can be limited in sensitivity and objectivity, highlighting the need for complementary molecular approaches. In this study, Attenuated Total Reflectance-Fourier Transform Infrared Spectroscopy (ATR-FTIR) was used to acquire mid-infrared (MIR) spectra from IBD-associated end…

By the Truvace desk2 min read1 sourcePeer-reviewedTRV-2026-2047· cite & verify
Infrared spectroscopy with statistical analysis and machine learning for cancer risk assessment in inflammatory bowel disease patients
"Endoscopy" by Yuya Tamai is licensed under CC BY 2.0. To view a copy of this license, visit https://creativecommons.org/licenses/by/2.0/.
Trace impact readingNegative state
P 74The P score combines the specificity and measured human impact of the grounded problem claim with the strength of this Trace’s cited sources.

Both sides are scored from claims and sources, not community votes.

G 67The G score combines the specificity and measured human impact of the grounded gain claim with the strength of this Trace’s cited sources.

In brief

Researchers collected mid-infrared spectra from endoscopic biopsies of 30 people with inflammatory bowel disease, 10 of whom later developed pre-cancerous dysplastic lesions. They analysed the spectra with principal component analysis, hierarchical clustering and machine learning classifiers to look for signs of future cancer risk.

The work matters because current risk stratification in IBD surveillance can lack sensitivity and objectivity, so molecular complements are sought. However, the classifiers performed no better than chance, so the approach remains exploratory and would need larger cohorts before it could inform patient care.

Main points

  1. The study analysed biopsies from 30 IBD patients, 10 of whom developed dysplastic pre-cancerous lesions during long-term follow-up.
  2. Samples were grouped by baseline clinical cancer risk and by later lesion development, targeting future risk rather than current disease status.
  3. PLS-based classifiers reached balanced accuracy of 0.61-0.70, but a permutation test gave p = 0.51.
  4. The lipid-associated 2950-3030 cm-1 spectral region was highlighted as carrying risk-related information.

The gain

Infrared spectral fingerprints of IBD biopsies, especially in a lipid-associated region, carried information linked to long-term colorectal cancer risk, suggesting a possible complement to surveillance.

The problem

Machine learning classifiers trained on 30 patients' biopsy spectra did not perform better than chance, so no definitive risk classifier could be established.

The rundown

The study used Attenuated Total Reflectance-Fourier Transform Infrared Spectroscopy to obtain mid-infrared spectra from endoscopic biopsy samples. Spectral variation was examined with chemometric methods, and the first three principal components explained 76.8%, 19.2% and 2.7% of variance.

Hierarchical clustering explored similarity patterns across defined spectral regions, and machine learning classifiers were evaluated under repeated nested cross-validation. No classifier significantly outperformed another.

The authors present the technique as a complementary, hypothesis-generating tool alongside conventional surveillance rather than a replacement for it.

What this doesn’t fix

The cohort of 30 patients is too small to establish a definitive classifier, and classifier performance was not significantly above chance.

Sources

  1. Peer-reviewedThe Analyst2026-10-13

The debate