Rice University smartphone imaging spots oral cancer rapidly

20 Oct 2025

Autofluorescence plus white-light approach could support early detection and referral decisions.

A project at Rice University has developed a low-cost, smartphone-based imaging system intended to assist the early identification of cancer in the mouth.

Called mDOC, for mobile Detection of Oral Cancer, the device is designed in particular to assist dentists and dental hygienists, who are often the first to spot suspicious lesions inside the mouths of patients but may lack the training to easily identify the problem.

As described in Biophotonics Discovery, mDOC meets the current need for tools to support early detection and help non-expert clinicians make appropriate referral recommendations.

Advances in optical components and handheld devices have led to several recent advances in photonics-based oral health platforms, such as the project at the University of Massachusetts Boston that combined three light sources in one device, allowing autofluorescence and white-light imaging.

The new Rice University mDOC device also incorporates white light and autofluorescence imaging, along with machine learning to assess oral lesions. Blue light at 405 nanometers is used to trigger the autofluorescence and to detect the changes in tissue fluorescence which can signal abnormal growth. The device's field of view was 2.5 x 3.5 centimeters at 44 microns transverse resolution.

But an autofluorescence method alone can be misleading, commented Rice University, as benign conditions like inflammation also reduce fluorescence. So the mDOC system uses a deep learning algorithm that analyzes both image data and patient risk factors, such as age, smoking habits, and anatomic location, to make referral recommendations.

"The mDOC device collects wide-field white light and autofluorescence images of the oral cavity as well as oral cancer risk factors, and uses machine learning methods to recommend whether patients should be referred for further evaluation by an oral cancer specialist," said the project in its paper.

Fitting into clinic workflows

In trials, Rice researchers collected data from patients at two community dental clinics in Houston, Texas. Each patient underwent imaging of up to five oral sites using the mDOC device, with the images reviewed by expert clinicians and their referral decisions serving as the ground truth for training the algorithm.

The final model was tested on a holdout dataset representing a low-prevalence population. The system correctly identified 60 percent of the sites that experts recommended for referral, while avoiding unnecessary referrals in most cases. Notably, the mDOC algorithm outperformed dental providers, who missed all cases that required referral.

While the system also misclassified two of five referral sites, those lesions had resolved by the time of the specialist visit, suggesting that mDOC may have correctly predicted that no further evaluation was needed. The algorithm did also produce some false positives, indicating room for improvement.

Even so the study highlights the potential of mDOC to support early detection and referral decisions in dental clinics, said the project, especially where access to specialists is limited. The process was also rapid, with an average imaging time of 3.5 minutes.

"This is a reasonable amount of time to integrate within existing workflows," noted the team in its paper. "With the addition of a near-real-time implementation of the model on the mDOC device, mDOC has the potential to aid in the detection of oral conditions that require further evaluation by an oral specialist in a timely manner."