SciFest National Final 2023

Stand 26

Using Machine Learning to Identify Radiolucencies on Panoramic Dental Radiographs (OPGs).

Student Kamaya Gogna
School St Joseph's Secondary School, Convent Lane, Rush, Co. Dublin
Teacher Kevin Delahunty
Venue DCU

Radiolucencies are problematic for dentists to diagnose. The current method of diagnosis is a confusing and long flowchart. This flowchart alone can lead to inconclusive results. The last resort is a painful biopsy. This inefficient method of diagnosis has led to a misdiagnosis rate of 49%.

I talked to dental science students in their fourth year at TCD, and they complained about this flowchart to me, about its length and complexity. I was studying machine learning on MIT open courseware and was in awe. So, after my discussions with the students, I recognised a gap in the diagnosis procedure and I realised that ML can help mitigate this gap.

My project involves making an algorithm that identifies radiolucencies. My model has been trained on 1,000 radiographs and can diagnose 21 types of radiolucency. After training my model and testing the classification algorithm with test sets, I found that the results were accurate.

The algorithm and the associated website were developed in response to the discussions with the TCD dentistry students. The COVID-19 pandemic put massive stress on the healthcare system and there is still a backlog to this day! The algorithm and website I created will help reduce stress on the healthcare system by quickly diagnosing patients. Ireland is facing a chronic lack of dentists as there are only 44 dentists per 100,000 people. An algorithm could never replace a dentist, but it could definitely be a useful tool. I want to push the boundaries and integrate technology into dentistry.

Poster Click here
Contact Us
+353 86 379 6143

SciFest National Final 2023
© 2024 SciFest National Final 2023