Automated analysis of multimodal fluorescence lifetime imaging and optical coherence tomography data for the diagnosis of oral cancer in the hamster cheek pouch model.

Jimi Brandon,Sebina Shrestha,Irma Gimenez-Conti,Javier A Jo,Paritosh Pande,Jesung Park,Brian E Applegate

doi:10.1364/boe.7.002000

Abstract

It is known that the progression of oral cancer is accompanied by changes in both tissue biochemistry and morphology. A multimodal imaging approach combining functional and structural imaging modalities could therefore provide a more comprehensive prognosis of oral cancer. This idea forms the central theme of the current study, wherein this premise is examined in the context of a multimodal imaging system that combines fluorescence lifetime imaging (FLIM) and optical coherence tomography (OCT). Towards this end, in the first part of the present study, the diagnostic advantage obtained by using both fluorescence intensity and lifetime information is assessed. In the second part of the study, the diagnostic potential of FLIM-derived biochemical features is compared with that of OCT-derived morphological features. For an objective assessment, several quantitative biochemical and morphological features from FLIM and OCT data, respectively, were obtained using signal and image processing techniques. These features were subsequently used in a statistical classification framework to quantify the diagnostic potential of different features. The classification accuracy for combined FLIM and OCT features was estimated to be 87.4%, which was statistically higher than accuracy based on only FLIM (83.2%) or OCT (81.0%) features. Moreover, the complimentary information provided by FLIM and OCT features, resulted in highest sensitivity and specificity for the combined FLIM and OCT features for discriminating benign (88.2% sens., 92.0% spec.), pre-cancerous (81.5% sens., 96.0% spec.), and cancerous (90.1% sens., 92.0% spec.) classes.

Highlights

Oral cancers have an overall five-year survival rate of approximately 60%
The mean accuracy for the exact fluorescence lifetime imaging (FLIM) features was higher than the approximate FLIM features, the p-value for the Wilcoxon test indicates that the difference between the overall accuracies for the two sets of FLIM features was not significantly different at α = 0.05 level of significance
This suggests that an accuracy as high as that obtained by using the exact FLIM features can, be obtained by using the approximate FLIM features

Summary

Introduction

Oral cancers have an overall five-year survival rate of approximately 60%. when detected in early stages, oral cancers can have a five-year survival rate as high as 80-90% [1]. Capable of providing information about both tissue biochemistry and morphology, could serve as an effective screening tool for possible malignancy of large number of oral lesions, which are routinely encountered by dentists and other physicians during visual examination of the oral cavity. Motivated by this idea, in this study, we primarily seek to examine the premise that using both biochemical and morphological information, obtained from fluorescence lifetime imaging (FLIM) and optical coherence tomography (OCT), respectively, increases the diagnostic accuracy for oral cancer, compared to using only one type of information

Methods

Results

Conclusion