cs.CVOct 13, 2026

AI-Enhanced Hyperspectral Imaging for Intraoperative Detection of Low-Grade Glioma: A Clinical Study

Authors: Siri Willems, Ishika Jain, Jens De Winne, Siri Luthman, Hamed Zivari Adab, Roeland Vandebriel, Anna Alperovich, Xiaohan Zhang, +4 more

Organizations: imec, Kapeldreef 75, Leuven, 3001, Belgium · Department of Computer Science UAntwerp-imec, Sint-Pietersvliet 7, Antwerp, 2000, Belgium · Department of Telecommunications and Information Processing, IPI-imec, Sint-Pietersnieuwestraat 41, Ghent, 9000, Belgium · Carl Zeiss AG, Corporate Research & Technology, Carl-Zeiss-Str. 22, Oberkochen, 73447, Germany · Carl Zeiss Meditec AG, Rudolf-Eber-Str. 11, Oberkochen, 73447, Germany · Research Group Experimental Neurosurgery and Neuroanatomy, Dept Neurosciences, KU Leuven, Leuven Brain Institute and Dept. Neurosurgery, UZ Leuven, Herestraat 49, Leuven, 3000, Belgium

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Current image-guided neuronavigation techniques, such as intraoperative MRI and ultrasound, are limited by high costs, limited sensitivity and operational complexity. There is a clear need for more efficient systems to enhance intra-operative tumor detection and improve patient outcomes for intrinsic brain tumor resection. Whereas previous intraoperative brain HSI studies have predominantly investigated high-grade gliomas, glioblastomas, or mixed-grade cohorts, this clinical study focuses specifically on low-grade gliomas (LGGs). LGGs strongly resemble normal brain under white light conditions on the operative microscope. This study characterizes spectral differences between healthy and tumor tissue and provides a paired comparison of HSI- and RGB-based classification at matched tissue locations and under identical patient-level data splits. Tumor regions showed consistently lower reflectance, which may be associated with differences in hemoglobin concentration and oxygenation. A range of classification models was evaluated. The 3D CNN, which integrates spectral and spatial features, achieved the highest median ROC-AUC of 74.5%, compared with 67.9% for the spectral-only XGBoost model and 64.0% for the RGB-based model. These results support further investigation of HSI for low-grade glioma classification and quantify its performance relative to conventional RGB imaging in this cohort.

Figures & tables

SpecificationsInVivo-LGGHELiCoiD [ 27 ]
Tumor typesLGGMix of brain tumors
Number of patients1834
Number of hypercubes7561
Maximum spatial resolution in pixels1600×16001600\times 1600741×1004741\times 1004
Spectral range470–787 nm400–1000 nm
Spectral resolution10–15 nm2–3 nm
Table 1: Overview of dataset characteristics for both internal and public datasets used in the study.
Figure 1: (a) Clinical setup. (b) Reference target used to align the RGB and HSI cameras. (c) Image-acquisition protocol. (d) RGB–HSI calibration results.
Figure 2: RGB image, histology-confirmed ground truth, initial superpixels, metric distributions, filtered superpixels, and final filtered ground truth.
Figure 3: Visual overview for calculating the intra-patient variability (left) and inter-patient variability (right).
Figure 4: Overview of (a) individual neural-network building blocks and (b) the evaluated neural-network architectures.
Figure 5: Overview of the patient and superpixel distribution across the six outer folds of the nested patient-level cross-validation design. Each fold is used once as the independent outer test set, while the remaining five folds are used for inner cross-validation and model optimization.
Figure 6: Mean spectra of healthy (green) and tumor (blue) tissue in the internal InVivo-LGG dataset and the public HELiCoiD dataset.
Figure 7: SAM values in degrees for (a) intra-patient variability and (b) inter-patient variability. In (a), each dot represents a patient, with green dots indicating healthy tissue and blue dots indicating tumor tissue; highlighted circles represent median values. In (b), each dot represents a pair of patients.
Models#patientsF1ACCb. ACCAUC-ROCPRECRECALL
XGB160.68 ±\pm 0.090.68 ±\pm 0.060.68 ±\pm 0.060.68 ±\pm 0.060.68 ±\pm 0.130.74 ±\pm 0.08
2DCNN-All160.60 ±\pm 0.110.64 ±\pm 0.070.67 ±\pm 0.060.73 ±\pm 0.070.67 ±\pm 0.150.66 ±\pm 0.15
2DCNN-ch12160.58 ±\pm 0.130.63 ±\pm 0.090.67 ±\pm 0.080.74 ±\pm 0.090.68 ±\pm 0.150.64 ±\pm 0.16
2DCNN-ch3160.59 ±\pm 0.110.63 ±\pm 0.080.67 ±\pm 0.080.74 ±\pm 0.100.69 ±\pm 0.130.63 ±\pm 0.16
3DCNN160.64 ±\pm 0.100.66 ±\pm 0.080.69 ±\pm 0.070.75 ±\pm 0.080.69 ±\pm 0.130.70 ±\pm 0.14
2DCNN-RGB160.56 ±\pm 0.090.59 ±\pm 0.050.59 ±\pm 0.060.64 ±\pm 0.080.63 ±\pm 0.130.55 ±\pm 0.10
Table 2: Test-set classification metrics summarized per patient as the median ±\pm 95% confidence interval. Boldface indicates a statistically significant paired difference between that HSI-based model and 2DCNN-RGB for the corresponding metric. No statistically significant differences were observed among the HSI-based model families.
Figure 8: Scatterplot of patient-level F1 and AUC-ROC values obtained from the six outer test folds.
Figure 9: Visual examples of patient 13
Figure 10: Visual examples of patient 18
Figure 11: Visual examples of patient 2
Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

Figure 12: Examples of both good and bad registrations of RGB image and corresponding contours with HSI. Bad registrations were manually registered afterwards.
  Prediction
  PositiveNegative
Ground truthPositiveTrue positive (TP)False Negative (FN)
NegativeFalse Positive (FP)True Negative (TN)
Appendix
Table 4: Binary confusion matrix.
Figure 13: Hyperparameter optimization results for the cross-validations for all runs. Each sample in the boxplot represents the cross-validation score for a run given the combination of hyperparameters

Explore similar work

CardsList

  1. A Lightweight Fiducial-Based Pipeline for 3D Hyperspectral Mapping of ex-vivo Lumpectomy Specimens

    Jun 12, 2026Anna Bicchi, Alberto Rota, Leonardo Passoni +5Breast CancerMedical Image Registration

    Department of Electronics, Information and Bioengineering (DEIB), Politecnico di Milano, Milan, Italy · Department of Physics, Politecnico di Milano, Milan, Italy

  2. A Data-Centric Framework for Intraoperative Fluorescence Lifetime Imaging for Glioma Surgical Guidance

    Apr 28, 2026Silvia Noble Anbunesan, Mohamed Abul Hassan, Jinyi Qi +4Medical Image AnalysisMedical Image Classification

    University of California, Davis · Department of Biomedical Engineering · Department of Pathology and Laboratory Medicine +1

  3. Shape-Based Inductive Bias for Glioma Grading from Tumor Contours

    Jul 27, 2026Puneet Velidi, Michelle F. Miranda, Farouk Nathoo +2BraTSGeometric Representation Learning

    Department of Mathematics and Statistics University of Victoria · Department of Computer Science University of Victoria · D´epartement de math´ematiques Universit´e du Qu´ebec `a Montr´eal