The simplified image-based ML solution can assist in preliminary screening via early detection of abnormal lesions that could potentially lead to oral cancer.
In healthcare, everything essentially boils down to timely detection of diseases. It forms the basis of the future course of treatment – whether it’s drug-based, surgical or a combination of the two. Attempting to do just that, IIITH’s IHub-Data and INAI – an Applied AI Research Centre which is a collaborative effort championed by Intel and catalysed by the Government of Telangana – have used artificial intelligence in the screening of abnormal lesions of the oral cavity, medically known as oral potentially malignant disorders (OPMDs). In rural areas of the country where tobacco and areca nut consumption are high, there’s a high prevalence of oral cancer and OPMDs. Early detection, and treatment of OPMDs is needed to prevent the natural progression into cancer. While the oral cavity can be easily visualised in the absence of specialised instruments unlike that for internal organs, such a clinical assessment is unfortunately subjective. While one could fall back upon biopsies which remain the gold standard for a definitive diagnosis, unfortunately they are not ideal especially in remote locations due to non-availability of experts and limited means at the disposal of community health centres. Hence a solution that serves as screening tool to triage the population at the community level for further investigations was mooted. Such an approach of risk stratification helps to filter the patients needing immediate attention.
AI Assist
A study titled ‘AI-Assisted Screening of Potentially Malignant Disorders’ commissioned by INAI and IHub Data set out to explore the potential of AI in detecting OPMDs from smartphone-based photographic images of oral cavities. The intent was to compare the performance of deep learning models and see if they could accurately flag suspicious lesions comprising of OPMDs. For the study, smartphone-based images were obtained from a curated database of community outreach programs for early cancer detection, the Biocon Foundation, and the Grace Cancer Foundation. The research[1] led by Vivek Talwar and Dr. Pragya Singh found that the simplified AI solution could indeed label images as suspicious and non-suspicious with an F1-score – an evaluation metric that measures the model’s accuracy – of over 70%. The researchers also discovered that clicking pictures on smartphones with a white light ensures consistency in imaging and eliminates the need for methods that use other light sources. The solution by design is aimed at being efficient and easy to use by healthcare workers at community level.
Conclusion
Today, the major challenge faced by the healthcare system is coverage of screening for easily detectable diseases. The treatment cycle reduction and improved prognosis can be achieved by early detection and awareness creation at the same time. Capturing diseases, such as the likes of oral cancer which is preventable and curable, early in the disease lifecycle reduces the financial burden and improves the quality of life for rural and middle-class India. According to the research team, further innovations in the quality of screening and migration to phone app from the current web app, will accelerate solution deployment. Mr. Konala Varma, CEO, INAI, remarks, “Innovation in the quality of screening and future deployment through mobile app will help improve the capturing of Oral Cancer incidence rate in community setups.”
More details at https://www.mdpi.com/2072-6694/15/16/4120
Next post