Intraoperative Margin Assessment in Oral and Oropharyngeal Cancer Using Label-Free Fluorescence Lifetime Imaging and Machine Learning

Abstract

To demonstrate the diagnostic ability of label-free, point-scanning, fiber-based Fluorescence Lifetime Imaging (FLIm) as a means of intraoperative guidance during oral and oropharyngeal cancer removal surgery. Methods: FLIm point-measurements acquired from 53 patients (n = 67893 pre-resection in vivo, n = 89695 post-resection ex vivo) undergoing oral or oropharyngeal cancer removal surgery were used for analysis. Discrimination of healthy tissue and cancer was investigated using various FLIm-derived parameter sets and classifiers (Support Vector Machine, Random Forests, CNN). Classifier output for the acquired set of point-measurements was visualized through an interpolation-based approach to generate a probabilistic heatmap of cancer within the surgical field. Classifier output for dysplasia at the resection margins was also investigated. Results: Statistically significant change (P < 0.01) between healthy and cancer was observed in vivo for the acquired FLIm signal parameters (e.g., average lifetime) linked with metabolic activity. Superior classification was achieved at the tissue region level using the Random Forests method (ROC-AUC: 0.88). Classifier output for dysplasia (% probability of cancer) was observed to lie between that of cancer and healthy tissue, highlighting FLIm's ability to distinguish various conditions. Conclusion: The developed approach demonstrates the potential of FLIm for fast, reliable intraoperative margin assessment without the

Mesoscopic fluorescence lifetime imaging: Fundamental principles, clinical applications and future directions

“…Examples include catheterized helical scanning mechanisms for intravascular imaging [108,113] and highly flexible free‐hand scanning during intraoperative interventions. The latter include brain tumor resection surgery, where the FLIm instrument interfaces with a neurosurgical microscope (eg, OPMI Pentero 900 (Carl Zeiss Meditec, Jena, Germany) [109]; and oral and oropharyngeal surgery for tumor removal with surgical robotic platforms (eg, da Vinci Surgical System [Intuitive Surgical, Sunnyvale, California] introducer sheaths [Si model] and graspers [SP model]) [110,123,127], which display the acquired FLIm data in an image format onto the surgeon's field‐of‐view. Finally, FLIm is also compatible with multimodal imaging and it has been combined with ultrasound [108,128], OCT [84,117,129] and Raman imaging [130–133].…”

“…Data modeling and machine learning techniques have been employed to analyze and interpret FLIm parameters (including fluorescence lifetime and intensity ratios for each spectral band) beyond conventional univariate statistics. When using the Laguerre expansion based deconvolution approach [116], a set of Laguerre expansion coefficients can also be incorporated into the analysis for additional discrimination value [127,155]. Several multivariate analysis methods using these parameters have been explored, adapting to the goal of each study and the complexity of the captured data.…”

“…Nonlinear classification models such as random forests (RF) [158], k‐nearest neighbors (KNN) [159] and support vector machines (SVM) [160] have been used to recognize acute tissue conditions that vary with experimental context. SVM and RF have been investigated for interpatient discrimination of breast cancer specimens imaged ex vivo [120,155] and oral and oropharyngeal cancer [127] (both in vivo and ex vivo). SVM has also been used to identify non‐melanoma skin lesions using FLIM microscopy [161].…”

“…A KNN classifier was used to detect actinic cheilitis in the lips and distinguish between normal tissue, mild dysplasia and moderate dysplasia [162]. While not all methods have been evaluated for every application discussed, the ensemble classification approach of random forests has been shown for multiple anatomical locations [127,155] to result in improved generalization between patients when identifying cancer using fiber‐based FLIm, suggesting this method produces robust classifiers for FLIm data.…”

“…Illustration of, A, various FLIm integration schemes, B, data visualization strategies and, C, validation against histopathology evaluation for applications in oral and oropharyngeal cancer (adapted from Reference [110, 127]), brain cancer (adapted from Reference [109]) and breast cancer (adapted from Reference [155])…”