TL;DR
Researchers have developed a novel hyperspectral imaging system, HP1.1, that can rapidly and non-invasively map molecular and metabolic information within brain tissue. This "label-free" approach avoids the need for contrast agents and shows promise for improving intraoperative guidance and early tumor assessment in neurosurgery.
What Problem Does This Paper Address?
Neurosurgery for brain tumors, particularly gliomas, is challenging due to the infiltrative nature of these tumors and the limitations of current imaging techniques. While preoperative imaging (MRI, CT, PET) is crucial for planning, it becomes less reliable during surgery due to brain shift. Intraoperative guidance often relies on fluorescence-guided surgery, which uses exogenous dyes that have limitations in specificity and suitability for all patients. Postoperative histopathology, the gold standard for grading, is time-consuming and can delay critical treatment decisions. There's a clear need for a real-time, label-free tool that can provide detailed molecular and metabolic information during surgery and for early tissue assessment.
Methods
The core of this work is the HyperProbe1.1 (HP1.1) system, a hyperspectral imaging (HSI) platform designed for ex vivo analysis of brain tissue. HP1.1 employs two main illumination strategies:
- Hyperspectral Reflectance Imaging: This uses a broadband plasma light source coupled with a tunable wavelength selector to acquire reflectance images across a spectral range of 385 to 1015 nm. This allows for mapping of chromophores like hemoglobin and cytochrome c oxidase, which are related to vascularization and metabolic activity.
- Autofluorescence Imaging: This utilizes a 370-nm LED to excite endogenous fluorophores NADH and FAD, which are key metabolic cofactors. Specific filters are then used to capture their emission spectra, providing insights into cellular metabolism.
The system is designed for rapid acquisition, with a complete hypercube (a 3D dataset of spatial and spectral information) generated in approximately 30 seconds. Data processing involves calibration to correct for system variations, followed by spectral unmixing algorithms (based on the Modified Beer-Lambert Law) to quantify the relative contributions of different molecules. For autofluorescence, an optical redox ratio (ORR) is calculated to assess the balance between NADH and FAD.
Datasets
The HP1.1 system was validated using:
- Plastic Optical Phantoms (MEDPHOT): These phantoms, with varying absorption and scattering coefficients, served as reliable standards for characterizing spectral response and system performance. Disk B3 specifically mimicked brain tissue optical properties.
- Liquid Phantoms: These included fluorescein (SF) and protoporphyrin IX (PpIX) solutions at various dilutions, used to test the system's ability to distinguish between exogenous fluorescent dyes.
- Freshly Excised Human Glioma Biopsies: A cohort of 28 glioma samples from 31 patients was used for ex vivo analysis. These samples were histopathologically graded according to the 2021 WHO classification, serving as the ground truth for hyperspectral analysis.
Key Insights and Ideas
- Label-Free Molecular Mapping: The ability to generate detailed maps of hemoglobin oxygenation, total hemoglobin, and cytochrome c oxidase activity without exogenous agents is a significant step towards real-time, non-invasive tissue characterization.
- Metabolic Fingerprinting: Integrating autofluorescence imaging of NADH and FAD provides a complementary metabolic contrast, revealing insights into cellular respiration and potentially distinguishing tumor grades.
- Bridging the Gap: HP1.1 aims to bridge the gap between preoperative planning and postoperative histopathology by offering intraoperative insights into tissue characteristics that are not visually apparent.
- Spatial Heterogeneity: The system effectively reveals significant spatial heterogeneity in oxygenation and metabolic activity within tumors, which is crucial for understanding tumor aggressiveness and guiding resection.
- Reproducibility: The reproducibility of spectral patterns within individual specimens highlights the system's robustness and its ability to resolve local metabolic heterogeneity.
Key Results
- Spatial Resolution: HP1.1 demonstrated the ability to resolve features at the cellular scale over a field of view of 6.5 mm, supporting both high-magnification and wide-field imaging.
- Signal-to-Noise Ratio (SBR): A linear relationship between exposure time and SBR was observed, indicating predictable system behavior under typical glioma acquisition conditions.
- Spectral Stability: Imaging of tissue-mimicking phantoms over days and weeks showed stable spectral performance, indicating excellent temporal consistency.
- Sensitivity: The system showed sensitivity in detecting low concentrations of fluorescein (down to 1 µM), relevant for clinical scenarios.
- Tumor Grading Correlation: Preliminary analysis of glioma biopsies showed a statistically significant upward trend in mean Optical Redox Ratio (ORR) with increasing tumor grade, suggesting a metabolic shift associated with glioma progression.
- Metabolic Signatures: High-grade gliomas (HGG) exhibited metabolic signatures consistent with increased stress and hypoxia, while low-grade gliomas (LGG) showed signs of more active oxidative metabolism.
Limitations and Future Work
- Limited Dataset Size and Balance: The study acknowledges a limited and slightly unbalanced dataset of glioma samples (6 low-grade vs. 22 high-grade). The authors deliberately refrained from inferential statistics or classifier development due to this.
- Ex Vivo Analysis: The current study focuses on ex vivo analysis. While promising, further investigations are needed to adapt and validate HP1.1 for in vivo intraoperative use.
- Interpretation of ORR: Interpreting the ORR trend in gliomas can be challenging due to the complex metabolic pathways tumor cells can exploit.
- Future Work: The authors suggest exploring HP1.1's application to other tumor types and for dynamic functional monitoring in preclinical in vivo models. They envision HP1.1 integrating into neurosurgical workflows as a fast, label-free tool for metabolic tissue characterization and intraoperative guidance.
Summary updated