Advancing ftir-based multimodal imaging with AI for risk stratification of oral precancer

The National Institute of Dental and Craniofacial Research awarded The Curators of the University of Missouri, doing business as University of Missouri at Kansas City, $527,533 on September 19, 2026, under the Oral Diseases and Disorders Research program (CFDA 93.121) to develop an AI-driven multimodal imaging system for risk stratification of oral potentially malignant disorders.

The award funds development of machine learning and deep learning models that integrate Fourier-transform infrared (FTIR) imaging and histological imaging to predict malignant transformation risk in oral precancerous lesions. The project encompasses three specific aims: development of FTIR imaging-based ML and DL models for risk stratification, development of histological imaging-based DL models for tissue segmentation and risk stratification, and development and validation of multimodal imaging systems combining both techniques. The proposed system is designed for incorporation into existing clinical workflows to enhance precancer diagnosis, facilitate early oral cancer detection, and improve outcomes for patients with oral squamous cell carcinoma precursors.

Work is performed in Kansas City, Missouri, with a period of performance extending from September 19, 2026, through June 30, 2030. The assistance type is Project Grant, a direct research award supporting basic, translational, and clinical research in dental, oral, and craniofacial health and disease.

Abstract

Successful treatment and management of oral mucosal lesions depend on definitive, accurate, and timely diagnosis. Oral squamous cell carcinoma (OSCC) is highly aggressive and often diagnosed at late stages, leading to poor prognosis. A significant portion of OSCC is preceded by oral potentially malignant disorder (OPMD), a heterogeneous group of precancerous oral lesions associated with an increased risk of malignant transformation (MT). However, gold-standard histopathological diagnosis is subjective and lacks predictive power for assessing malignant transformation risk in individual OPMD patients, creating a critical barrier in oral precancer management. To address this unmet medical need, we propose an innovative multimodal imaging system that integrates ftir imaging and histological imaging to capture both biochemical and morphological features of tissue. Leveraging advanced machine learning (ML) and deep learning (DL) methods, we will develop predictive models based on the two complementary imaging techniques for OPMD risk stratification. Our central hypothesis is that AI-driven integration of biochemical fingerprints from ftir imaging with morphological features from histological imaging will yield a more precise model for predicting MT in opmds, outperforming traditional histological assessment or either modality alone. To test this hypothesis, we propose the following specific aims: 1) develop ftir imaging-based ML and DL models for OPMD risk stratification; 2) develop histological imaging- based DL models for OPMD tissue segmentation and risk stratification; and 3) develop and validate multimodal imaging systems for OPMD risk stratification. The proposed system can be easily incorporated into existing clinical workflows to enhance OPMD diagnosis and management, facilitate early oral cancer detection, and ultimately reduce the global burden of the disease.