Application of deep-learning algorithms on automated MRI data processing

PhD study program: Measurement Technology
Academic year: 2023-2024
Advisor: RNDr. Andrej Krafčík, PhD. (andrej.krafcik@savba.sk)
External educational institution: Institute of Measurement Science SAS
Accepting university: Faculty of Electrical Engineering and Information Technology, Slovak University of Technology in Bratislava, Institute of Electrical Engineering

Annotation:

Automated identification and segmentation of clinical data, obtained primary by MRI, is very desirable. The reason is typically large size of data and therefore enormous time, which radiologist has to invest into the manual segmentation. Availability of powerful hardware open new capabilities to automate this processes and speedup via deep learning techniques using convolutional neural networks (CNN). Therefore, student will learn the fundamental functionality principles of MRI device (theoretically and practically), try manual segmentation of volumetric MRI data, and theoretically and practically learn principles of CNN. Student will design own architecture of CNN for automated segmentation of volumetric data, further train, validate and implement on testing data.
The output of this dissertation should be a CNN capable of deployment in clinical practice, in the diagnosis and quantitative analysis of selected tissues (cartilage, ligaments, tendons, menisci, subcutaneous fat, etc.). It is theoretical work, in which programming basics and knowledge of some programming language are necessary. As the programming environment, for design and implementation of CNN, will be used Python with module TensorFlow.

Literature:
1. Juras, V., Szomolanyi, P., Schreiner, M.M., Unterberger, K., Kurekova, A., Hager, B., Laurent, D., Raithel, E., Meyer, H., Trattnig, S. Reproducibility of an Automated Quantitative MRI Assessment of Low-Grade Knee Articular Cartilage Lesions. Cartilage, 13, 646S-657S, 2021.
2. Chollet, F. Deep learning with Python – 2nd edition, Manning Publications Co., Shelter Island, NY, 2021.
3. Cicek, O., Abdulkadir, A., Lienkamp, S.S., Brox, T., Ronneberger, O. 3D U-net: Learning dense volumetric segmentation from sparse annotation. Lecture Notes in Computer Science (including subseries Lecture Notes in Arti cial Intelligence and Lecture Notes in Bioinformatics), 9901 LNCS, 424-432, 2016.