Medical imaging and CNNs for monitoring of the Achilles tendon healing process
Medical imaging based on Magnetic Resonance Imaging (MRI) and Ultrasound (US) techniques is a baseline in the assessment of the healing progress in musculoskeletal disorders, including one of the most common – Achilles tendon rupture. At the same time, radiology is a part of healthcare where Deep Learning, in particular CNNs, is rapidly spreading and entering everyday practice. I will present a summary of our experience gained in the process of building a diagnostic support tool – an automatic assessment of the healing degree of the Achilles tendon based on image data. I will discuss solutions combining CNN techniques, classic imaging features, as well as regression models, allowing both binary classification and continuous quantitative assessment of MRI and US images. I will discuss the usage of a weakly-supervised approach, based on the reduction of significant features of a dense layer, and compare it with the supervised model, as well as late fusion of texture features, in the light of the evaluation of selected tissue characteristics. The developed solution enables a quantitative assessment with an accuracy similar to that of a radiology expert.
Bio
For over a decade, he has been involved in R&D in the area of Computer Assisted Medicine methods. He’s mainly focusing on areas related to image information and data analytics. Physicist by profession – in the field of Computer Methods and Optical Information Processing. Since 2006, he’s been working with the Interdisciplinary Centre for Mathematical and Computational Modelling at the University of Warsaw, where he’s currently leading the Artificial Intelligence and Image Analysis in Medical Diagnostics group. His main areas of interest are around topics of medical image acquisition and processing, assistance in radiology, diagnostics and surgery, using mathematical models, machine learning, artificial intelligence, augmented reality and visual methods. He also specializes in the fields of large-scale computations (High Performance Computing) and scientific visualization. More recently, focused on bringing AI solutions for radiology within orthopedics and sports medicine to the market through Smarter Diagnostics start-up.