Integration of the Fluorescence Based Portable Device with the AI Tools for the Real-time Monitoring of Oral Mucosal Lesions - Key Findings

Sep 23, 2026

Uploaded by Riley

Non-invasive, sensitive, real-time, user-friendly optical detection for early oral mucosal lesions is addressed through an AI-integrated, steady-state fluorescence portable device. The device uses a 405 nm laser diode, UV-visible spectrometer, and optical components to excite endogenous fluorophores (FAD and porphyrin) in diseased and normal oral cavities. Enhanced porphyrin fluorescence is observed in OSCC and dysplasia. AI-based analysis (Naïve Bayes, LDA, QDA) yields best discrimination with QDA, reaching 95.34%–100% accuracies across Normal, Dysplasia, and OSCC.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages9

Reading Time23 min

AccessFree to download

Introduction

Need for early, non-invasive detection

Limitations of conventional biopsy and survivability

Methods and Device Principle

Steady-state fluorescence-based portable device

Excitation/emission via endogenous fluorophores

AI-based Data Analysis

Naïve Bayes, LDA, QDA performance

Discrimination accuracies across lesion states

Related Research Context

Prior fluorescence spectroscopy and clinical studies

[www. nature.com/scientificreports](www. nature.com/scientificreports) OPEN Integration of the fluorescence based portable device with the AI tools for the real-time monitoring of oral mucosal lesions Pavan Kumar1,2􀀍 & Shashikant Rathod3,4 There is a need for non-invasive, sensitive, real-time, and user-friendly optical devices integrated with artificial intelligence (AI) based tools for the detection of oral mucosal lesions at early stage. Research on the development of optical devices has been executed by several research groups for the cancer detection and it is still being continued. We have also contributed towards it by developing a steadystate fluorescence-based portable device. The in-house developed device is equipped with 405 nm laser diode, UV visible spectrometer, optical components, and other accessories. Laser light irradiated on the oral cavity of diseased (cancerous) and non-diseased (normal) groups, excites the two endogenous fluorophores namely FAD and porphyrin. We observed an enhancement in the porphyrin fluorescence of cancerous patients (OSCC and Dysplasia) than the normal group. Data analysis carried out by AI tools i.e., Naïve Bayes, LDA, and QDA showed slightly higher accuracy for QDA. QDA was able to discriminate among Normal toOSCC, Normal to Dysplasia, and Dysplasia toOSCC with accuracies of 95.34%, 100%, and 97.43% respectively. Keywords Oral mucosal lesions, Fluorescence spectroscopy, Porphyrin, AI tools Tumor development is a slow process and it takes years before a tumor is felt by hand or can be visualized by the naked eye. It is therefore need of sensitive and non-invasive devices to identify the disease at the very early stage. Among the conventional methods, tissue biopsy with histopathology is the only standard screening procedure for the identification of oral lesions. Though, histopathology is the most reliable (gold standard) but its invasive ethos and time-consuming procedure make it incompetent. Patients at any stage of cancer i.e., benign, precancer (mild, moderate, or severe), and oral squamous cell carcinoma (OSCC; grade I to IV) have to undergo this painful process during the treatment. It can also not be assured that the selected area for the biopsy is rightly chosen or not. The research finding shows that a five-year survival rate after the treatment is extensively less (~ 45%), which has not remedied in the last few decades. The predominant reasons for poor survivability are late diagnosis as well as the absence of early symptoms1–5. It is therefore needed for such devices which to be sensitive, fast in detection, user-friendly, low cost, and portable. Research carried out for the detection of oral lesions indicates that optical devices (spectroscopic and imaging) such as fluorescence, Raman, lifetime, diffuse reflectance, etc. could detect oral cavity abnormalities non-invasively as well as at an early stage6–15. Among the spectroscopic devices, steady-state fluorescence-based devices are adequately utilized by the researchers for in-vivo testing of oral cancer16–30. Fluorescence devices are based on light interaction with material (e.g. biomaterial, tissue, etc.) . A biological tissue consists of several endogenous fluorophores and they have a particular excitation wavelength (λexc) and emission wavelength (λemi) at which they yield maximum emission and absorption bands. Tryptophan, Collagen, Elastin, NADH, FAD, and Porphyrin are the key fluorophores and they have excitation wavelengths near 280, 325, 350, 340, 450, and 400 nm and emission bands at 340, 390, 410, 440, 530, and 634 nm respectively. It has been established that fluorophore concentration changes with the progress of disease and these changes can be noticed in the fluorescence signal. 1Faculty of Engineering and Technology (FEAT), Datta Meghe Institute of Higher Education and Research (DMIHER), Wardha 442001, India. 2Department of Physics, Indian Institute of Technology Kanpur (IITK), Kanpur 208016

FAQs

Why is early detection of oral mucosal lesions necessary?

Tumor development is slow and symptoms may be absent until late stages. Early diagnosis is needed because survivability after treatment is limited and late diagnosis is a major cause of poor outcomes.

What optical device and excitation mechanism are used in this study?

A steady-state fluorescence-based portable device irradiates the oral cavity with a 405 nm laser diode and analyzes emitted fluorescence using a UV-visible spectrometer. The approach excites endogenous fluorophores FAD and porphyrin.

Which AI method performed best for classifying lesion types?

QDA showed slightly higher accuracy than Naïve Bayes and LDA. It discriminated among Normal to OSCC, Normal to Dysplasia, and Dysplasia to OSCC with reported accuracies of 95.34%, 100%, and 97.43% respectively.

Sep 23, 2026

Uploaded by Riley