Causal responsibility based explainable AI for vibrational spectroscopy applied to oral FTIR and oesophageal Raman diagnostics
4 affiliations
- 1Department of Medical Physics and Biomedical Engineering, University College London
- 2Department of Informatics, King’s College London
- 3Department of Cell and Developmental Biology, University College London
- 4School of Dentistry, University of Missouri–Kansas City
Abstract
Biomedical applications of vibrational spectroscopy increasingly use deep learning, but these models often give classifications without revealing which spectral features drive their decisions. We introduce Spec-ReX, a causal responsibility-based explainable AI method for vibrational spectroscopy, and compare it with the widely used explainability methods SHAP and Grad-CAM across synthetic spectra, biochemical mixtures, oral FTIR tissue spectra and oesophageal Raman tissue spectra. In a definitive in silico experiment with known discriminative peaks, Spec-ReX achieved the highest ground truth localisation, with an Intersection over Union of 0.10 and attribution concentration of 0.56, compared with 0.09 and 0.26 for the next best method. In an ambiguous in silico experiment, SHAP achieved the highest positive hit rate, while Spec-ReX was most robust to removal of low-importance regions, consistent with a sensitivity–specificity trade-off. In the in vitro biochemical experiment, Spec-ReX was again the most robust to removal of low-importance regions, while SHAP performed best when removing high-importance regions. In the ex vivo datasets, where no definitive attribution ground truth was available, the methods returned different and architecture-dependent attribution patterns. Overall, Spec-ReX provides sparse and specific model-causal responsibility maps for spectral classifiers. These explanations pertain to trained model behaviour rather than biological causality, and their clinical utility remains to be validated through user-centred studies.
Related
- Decoding Cell–Cell Interactions in Neurovascular Unit Co-Cultures Through Molecular Vibrational Spectroscopy and Biology
- Alcohol, tobacco and oesophageal cancer: effects of the duration of consumption, mean intake and current and former consumption
- Defining the role of HER2-targeted therapy as perioperative treatment in HER2-positive gastric and gastro-oesophageal junction cancers: protocol for a systematic review and an individual patient data meta-analysis
- Swallowed dental prosthesis causing late oesophageal obstruction and tracheo-oesophageal fistula in a young patient.
- Genomic correlates with survival after treatment of oesophageal adenocarcinoma by surgery alone
BibTeX
@article{blake2026causal,
title = {{Causal responsibility based explainable AI for vibrational spectroscopy applied to oral FTIR and oesophageal Raman diagnostics}},
author = {Nathan Blake and David A. Kelly and Sarah Kapllani-Mucaj and Rong Wang and Geraint Thomas and Hana Chockler},
journal = {Scientific Reports},
year = {2026},
month = jul,
doi = {10.1038/s41598-026-62343-3},
url = {https://doi.org/10.1038/s41598-026-62343-3},
note = {CC BY},
}
RIS
TY - JOUR
TI - Causal responsibility based explainable AI for vibrational spectroscopy applied to oral FTIR and oesophageal Raman diagnostics
AU - Nathan Blake
AU - David A. Kelly
AU - Sarah Kapllani-Mucaj
AU - Rong Wang
AU - Geraint Thomas
AU - Hana Chockler
PY - 2026
DA - 2026/07/01
JO - Scientific Reports
DO - 10.1038/s41598-026-62343-3
UR - https://doi.org/10.1038/s41598-026-62343-3
AB - Biomedical applications of vibrational spectroscopy increasingly use deep learning, but these models often give classifications without revealing which spectral features drive their decisions. We introduce Spec-ReX, a causal responsibility-based explainable AI method for vibrational spectroscopy, and compare it with the widely used explainability methods SHAP and Grad-CAM across synthetic spectra, biochemical mixtures, oral FTIR tissue spectra and oesophageal Raman tissue spectra. In a definitive in silico experiment with known discriminative peaks, Spec-ReX achieved the highest ground truth localisation, with an Intersection over Union of 0.10 and attribution concentration of 0.56, compared with 0.09 and 0.26 for the next best method. In an ambiguous in silico experiment, SHAP achieved the highest positive hit rate, while Spec-ReX was most robust to removal of low-importance regions, consistent with a sensitivity–specificity trade-off. In the in vitro biochemical experiment, Spec-ReX was again the most robust to removal of low-importance regions, while SHAP performed best when removing high-importance regions. In the ex vivo datasets, where no definitive attribution ground truth was available, the methods returned different and architecture-dependent attribution patterns. Overall, Spec-ReX provides sparse and specific model-causal responsibility maps for spectral classifiers. These explanations pertain to trained model behaviour rather than biological causality, and their clinical utility remains to be validated through user-centred studies.
SN - 2045-2322
ER -