Student Develops AI-Powered Prototype to Screen for Early Signs of Oral Cancer

Rian Sanghi, a Grade 12 student at Scottish High International School, Gurugram, has developed OroScan, an AI-powered prototype designed to help screen images of the mouth for signs that may require further medical examination. Combining engineering, artificial intelligence and a focus on community healthcare, the project aims to address gaps in access to preliminary oral screening.

Driven by an interest in engineering and experimentation with emerging technologies, Rian conceived OroScan after recognising the challenges associated with access to routine oral examinations, particularly in underserved communities. Early detection can play an important role in the management of oral cancer, yet limited access to screening and delayed medical consultation can prevent potentially concerning signs from being identified promptly.

OroScan was developed with the aim of making preliminary screening more accessible and encouraging people to seek professional medical advice when needed. The device is designed as a screening and referral aid, not a replacement for diagnosis by a qualified medical professional.

Combining AI with a Portable Device

OroScan integrates an endoscopy camera, a custom-built printed circuit board (PCB) and a convolutional neural network (CNN)-based AI model trained to analyse images from inside the mouth. The model classifies images as suspicious or clear and generates a heatmap highlighting the regions that influenced its assessment, offering a visual explanation of the result.

The prototype also features an onboard display that allows users to view images and results directly on the device without connecting it to a laptop. Its self-contained design is intended to support ease of use in community screening settings, including outreach camps.

Developing the AI model involved multiple rounds of testing and refinement. According to the project’s reported results, OroScan achieved 96% accuracy, 98% recall and 94% precision on a held-out test dataset. Rian placed particular emphasis on recall, a metric that measures how effectively a model identifies cases belonging to a target category.

For a preliminary screening application, the objective is to minimise the likelihood of potentially suspicious cases being missed. The prototype is designed to flag images that may warrant further examination, with medical assessment remaining essential to determining the next steps.

Moving beyond software development, Rian integrated the model with the camera, custom PCB and a 3D-printed enclosure, along with an onboard screen, to create a portable prototype designed for practical use outside a laboratory.

Taking the Innovation to Communities

OroScan has received recognition across several platforms. The project was shortlisted for Samsung Solve for Tomorrow, received a Gold award at the International Greenwich Olympiad, was presented at IIT Delhi and has been acknowledged by the Ministry of Health and Family Welfare, Government of India.

An important milestone in the project’s development has been its introduction to community outreach through a screening camp conducted in collaboration with TCTF. The initiative provided an opportunity to take OroScan beyond a controlled testing environment and explore its use in a community setting.

Through its outreach initiatives, OroScan has reached more than 1,500 people, according to the project, with further activities underway. These efforts are helping inform the project’s next phase of development, including improvements in usability and engagement with healthcare professionals.

Expanding Reach and Refining the Technology

Rian is now working to expand OroScan’s reach through partnerships with additional non-governmental organisations (NGOs) and community organisations. Plans include conducting more screening camps, gathering feedback from healthcare professionals and refining the prototype to improve its usability and performance.

As the project progresses, further validation will be important to establish its effectiveness across different screening environments. The broader objective is to explore how accessible AI-powered tools can complement existing screening initiatives and encourage timely consultation with qualified medical professionals.

Through OroScan, Rian is bringing together engineering and AI to explore a practical approach to a public health challenge, with a focus on making preliminary screening more accessible to communities that may otherwise have limited opportunities for routine examination.

Ultimately, he doesn’t envision OroScan as a replacement for diagnosis in the existing system. But Rian hopes it can help bridge the gap that exists by providing an early first opinion.

Share on