AUTHOR=Chen Yao , Streeter Samuel S. , Hunt Brady , Sardar Hira S. , Gunn Jason R. , Tafe Laura J. , Paydarfar Joseph A. , Pogue Brian W. , Paulsen Keith D. , Samkoe Kimberley S. 

TITLE=Fluorescence molecular optomic signatures improve identification of tumors in head and neck specimens

JOURNAL=Frontiers in Medical Technology

VOLUME=Volume 5 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/medical-technology/articles/10.3389/fmedt.2023.1009638

DOI=10.3389/fmedt.2023.1009638

ISSN=2673-3129

ABSTRACT=Background: Fluorescence molecular imaging is emerging as a method for precise surgical guidance during head and neck squamous cell carcinoma (HNSCC) resection. However, the tumor-to-normal tissue contrast is confounded by intrinsic physiological limitations of heterogeneous expression of the target molecule, epidermal growth factor receptor (EGFR), and the non-specific uptake of the fluorescent imaging agent. 
Objective: In this study, we extended a radiomics approach for tissue classification to optical fluorescence molecular imaging data—termed “optomics”. Optomics seeks to improve tumor identification by leveraging textural pattern differences in EGFR expression conveyed by fluorescence. We determined whether these optomic signatures improve binary classification of malignant versus non-malignant tissues relative to more conventional fluorescence intensity thresholding.
Materials and Methods: Fluorescence image data collected through a Phase 0 clinical trial (NCT03282461) for testing ABY-029, a novel fluorescent EGFR-targeted imaging agent, involved 24 bread-loafed slices from surgical HNSCC resections originating from 12 patients who were stratified into three dose groups (30, 90, and 171 nanomoles). For analysis, each dose group was partitioned randomly 75%/25% into training/testing sets, then all training and testing sets were aggregated. A total of 1,472 standardized radiomic features were extracted from fluorescence image samples. A supervised machine learning pipeline involving a support vector machine classifier was trained with 25 top-ranked features selected by minimum redundancy maximum relevance criterion. Predictive performance of model was compared to fluorescence intensity thresholding by classifying testing set image patches of resected tissues with histologically confirmed malignancy status.
Results: The optomics approach provided consistent improvement in prediction accuracy on all test set samples, irrespective of dose, compared to fluorescence intensity thresholding (mean accuracies of 89% vs. 81%; P = 0.0072).
Conclusions: The optomics approach outperformed conventional fluorescence intensity thresholding for tumor identification. Optomics mitigates diagnostic uncertainties introduced through physiological variability, imaging agent dose, and inter-patient biases of fluorescence molecular imaging by probing textural image information. The improved performance demonstrates that extending the radiomics approach to fluorescence molecular imaging data offers a promising image analysis technique for cancer detection in fluorescence-guided surgery.