Behnam Javanmardi

Research 

The rapid developments in Artificial Intelligence (AI) and, at the same time, the emergence of Big Data in medical fields have provided researchers with unprecedented opportunities for advancement of Precision and Personalized Medicine. In our research group, we exploit state-of-the-art AI techniques in Computer-Vision and Automated Image Analysis to develop tools and methods for computer-assisted diagnosis of Genetic Disorders.

Projects

Although individually rare, skeletal dysplasias collectively constitute an important group of genetic disorders often resulting in short stature, altered movement biomechanics, pain, fatigue, and reduced functional performance. The clinical diagnosis of these disorders usually requires recognizing patterns on skeletal X-Ray images which is a challenging task since most of these disorders are extremely rare, and even experienced clinicians might have seen only some of these disorders. This is where AI and computer-vision can play an important role in providing assistive tools for prompt detection and identification of these genetic disorders.

In this project, we are collecting hand X-Ray images of patients diagnosed with skeletal dysplasias and we are using deep Convolutional Neural Networks (CNN) to build accurate AI models for the identification of phenotypic patterns and classification of different rare skeletal diseases.

The number of people with diabetes rose from 108 million in 1980 to 422 million in 2014 and is estimated to rise to 700 million by 2045. Up to 80% of patients with diabetes are affected by Diabetic Retinopathy (DR) also known as diabetic eye disease. While DR is one of the major causes of blindness, the risk of vision loss can be reduced with early detection and treatment. Therefore, frequent monitoring and screening of diabetic patients are crucial for preventing vision-threatening outcomes. AI-based methods have already proven to be effective in the identification of different stages of DR. 

In this project, we aim at improving the sensitivity of deep learning models in the recognition of early stages of DR using both fundus images and polygenic risk scores. 

Group Members