SURG-95. Spectral Imaging and AI for Brain Tumor Characterization: Toward Data-Driven Surgical Guidance
Bibliographic record
Abstract
Abstract Precise delineation of brain tumor margins remains a clinical challenge, even with the aid of 5-aminolevulinic acid (5-ALA) fluorescence guidance — particularly in low-grade gliomas or infiltrative zones where visible fluorescence is weak or absent. Hyperspectral imaging (HSI), which captures detailed spectral data per pixel, offers the potential for refined tissue characterization based on spectral signatures. However, converting raw spectral data into clinically meaningful overlays involves complex processing, which can be optimized through machine- and deep learning. We developed a data-driven pipeline for ex vivo hyperspectral fluorescence imaging of brain tumor biopsies. The workflow comprises automatic biopsy segmentation, spectral feature extraction, and deep learning-based normalization to correct for optical variability across samples. Spectral unmixing is then applied to estimate the relative abundance of key fluorophores, including PpIX and various autofluorescent compounds. These abundance profiles serve as input features for machine learning classifiers trained to predict tumor type, WHO grade, margin type, and IDH mutation status. The dataset consists of 891 hyperspectral image cubes from 184 patients with diverse brain tumor pathologies. Deep neural networks enhanced the normalization process by accounting for complex, tissue-specific optical properties, leading to more robust abundance estimations. Classifiers trained on the processed spectral data achieved test accuracies of 87.3% (tumor type), 96.1% (WHO grade), 85.7% (tumor margin), and 93% (IDH mutation), surpassing the performance of previous non-fluorescence-based methods. The integration of deep learning for normalization and both classical and machine learning-based unmixing significantly enhanced data interpretability. We demonstrate that HSI, when combined with deep learning-based normalization and data-driven analysis, enables accurate classification of key brain tumor features. The fusion of spectral unmixing and machine learning facilitates the extraction of molecular and histopathological signatures, offering real-time potential to support surgical decision-making. These results underscore the clinical value of integrating AI with HSI for fluorescence-guided brain tumor resection.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.008 | 0.007 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".