Dataset Information

Enhancing intraoperative tumor delineation with multispectral short-wave infrared fluorescence imaging and machine learning.

ABSTRACT:

Significance

Fluorescence-guided surgery (FGS) provides specific real-time visualization of tumors, but intensity-based measurement of fluorescence is prone to errors. Multispectral imaging (MSI) in the short-wave infrared (SWIR) has the potential to improve tumor delineation by enabling machine-learning classification of pixels based on their spectral characteristics.

Aim

Determine whether MSI can be applied to FGS and combined with machine learning to provide a robust method for tumor visualization.

Approach

A multispectral SWIR fluorescence imaging device capable of collecting data from six spectral filters was constructed and deployed on neuroblastoma (NB) subcutaneous xenografts ( n=6 ) after the injection of a NB-specific NIR-I fluorescent probe (Dinutuximab-IRDye800). We constructed image cubes representing fluorescence collected from ∼850 to 1450 nm and compared the performance of seven learning-based methods for pixel-by-pixel classification, including linear discriminant analysis, k -nearest neighbor classification, and a neural network.

Results

The spectra of tumor and non-tumor tissue were subtly different and conserved between individuals. In classification, a combine principal component analysis and k -nearest-neighbor approach with area under curve normalization performed best, achieving 97.5% per-pixel classification accuracy (97.1%, 93.5%, and 99.2% for tumor, non-tumor tissue and background, respectively).

Conclusions

The development of dozens of new imaging agents provides a timely opportunity for multispectral SWIR imaging to revolutionize next-generation FGS.

SUBMITTER: Waterhouse DJ

PROVIDER: S-EPMC10042297 | biostudies-literature | 2023 Sep

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Enhancing intraoperative tumor delineation with multispectral short-wave infrared fluorescence imaging and machine learning.

Waterhouse Dale J DJ Privitera Laura L Anderson John J Stoyanov Danail D Giuliani Stefano S

Journal of biomedical optics 20230327 9

<h4>Significance</h4>Fluorescence-guided surgery (FGS) provides specific real-time visualization of tumors, but intensity-based measurement of fluorescence is prone to errors. Multispectral imaging (MSI) in the short-wave infrared (SWIR) has the potential to improve tumor delineation by enabling machine-learning classification of pixels based on their spectral characteristics.<h4>Aim</h4>Determine whether MSI can be applied to FGS and combined with machine learning to provide a robust method for ...[more]

PMID: 36993142

Similar Datasets

Project description:BackgroundAlthough real-time localization of gliomas has improved with intraoperative image guidance systems, these tools are limited by brain shift, surgical cavity deformation, and expense.ObjectiveTo propose a novel method to perform near-infrared (NIR) imaging during glioma resections based on preclinical and clinical investigations, in order to localize tumors and to potentially identify residual disease.MethodsFifteen patients were identified and administered a Food and Drug Administration-approved, NIR contrast agent (Second Window indocyanine green [ICG], 5 mg/kg) before surgical resection. An NIR camera was utilized to localize the tumor before resection and to visualize surgical margins following resection. Neuropathology and magnetic resonance imaging data were used to assess the accuracy and precision of NIR fluorescence in identifying tumor tissue.ResultsNIR visualization of 15 gliomas (10 glioblastoma multiforme, 1 anaplastic astrocytoma, 2 low-grade astrocytoma, 1 juvenile pilocytic astrocytoma, and 1 ganglioglioma) was performed 22.7 hours (mean) after intravenous injection of ICG. During surgery, 12 of 15 tumors were visualized with the NIR camera. The mean signal-to-background ratio was 9.5 ± 0.8 and fluorescence was noted through the dura to a maximum parenchymal depth of 13 mm. The best predictor of positive fluorescence was enhancement on T1-weighted imaging; this correlated with signal-to-background ratio (P = .03). Nonenhancing tumors did not demonstrate NIR fluorescence. Using pathology as the gold standard, the technique demonstrated a sensitivity of 98% and specificity of 45% to identify tumor in gadolinium-enhancing specimens (n = 71).ConclusionWith the use of Second Window ICG, gadolinium-enhancing tumors can be localized through brain parenchyma intraoperatively. Its utility for margin detection is promising but limited by lower specificity.Abbreviations5-ALA, 5-aminolevulinic acidEPR, enhanced permeability and retentionFDA, Food and Drug AdministrationGBM, glioblastomaICG, indocyanine greenNIR, near-infraredNPV, negative predictive valuePPV, positive predictive valueROC, receiver operating characteristicROI, region of interestSBR, signal-to-background ratioWHO, World Health Organization.

Project description:While multiplexed fluorescence imaging is frequently used for in vitro microscopy, extending the technique to whole animal imaging in vivo has remained challenging due to the attenuation and scattering of visible and traditional near infrared (NIR-I) wavelengths. Fluorescence imaging using short-wave infrared (SWIR, 1000 - 1700 nm, a.k.a. NIR-II) light enables deeper tissue penetration for preclinical imaging compared to previous methods due to reduced tissue scattering and minimal background autofluorescence in this optical window. Combining NIR-I excitation wavelengths with multiple distinct SWIR emission peaks presents a tremendous opportunity to distinguish multiple fluorophores with high precision for non-invasive, multiplexed anatomical imaging in small animal models. SWIR-emitting semiconductor quantum dots (QDs) with tunable emission peaks and optical stability have emerged as powerful contrast agents, but SWIR imaging demonstrations have yet to move beyond two-color imaging schemes. In this study, we engineered a set of three high quantum yield lead sulfide/cadmium sulfide (PbS/CdS) core/shell QDs with distinct SWIR emissions ranging from 1100 - 1550 nm and utilize these for simultaneous three-color imaging in mice. We first use QDs to non-invasively track lymphatic drainage, highlighting the detailed network of lymphatic vessels with high-resolution with a widefield imaging over a 2 hr period. We then perform multiplexed imaging with all three QDs to distinctly visualize the lymphatic system and spatially overlapping vasculature network. This work establishes optimized SWIR QDs for next-generation multiplexed preclinical imaging, moving beyond the capability of previous dual-labeling techniques. The capacity to discriminate several fluorescent labels through non-invasive NIR-I excitation and SWIR detection unlocks numerous opportunities for studies of disease progression, drug biodistribution, and cell trafficking dynamics in living organisms.

Dataset Information

Enhancing intraoperative tumor delineation with multispectral short-wave infrared fluorescence imaging and machine learning.

Significance

Aim

Approach

Results

Conclusions

Publications

Enhancing intraoperative tumor delineation with multispectral short-wave infrared fluorescence imaging and machine learning.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets