Age Estimation from MR Images

The Ludwig Boltzmann Institute for Clinical Forensic Imaging (LBI-CFI) has an interest in estimating the age of a subject from medical imaging data. This topic is especially relevant to decide if someone, who does not know his or her real age (e.g. a fugitive seeking asylum in Austria), is below or above 18 years old.

Traditionally, such a question is investigated using x-ray radiographs of the wrist bones (looking for the epiphyseal cartilage - Wachstumsfuge) and/or looking at the third molar (Weisheitszahn). The LBI-CFI is currently gathering a large number of MR scans involving these structures (i.e. wrist, clavicle and dental MR images) together with the information about the real age of the subjects. From this data pool we intend to derive automated detection, segmentation and age estimation (either by regression or classification below/above 18) algorithms, such that we are able to come up with models that very objectively decide on the age of fugitives. Such fugitives are currently already processed with more traditional approaches in the routine work of the LBI-CFI.

In this paid master thesis (which might start as a seminar/project work) we want to come up with a proof of concept that Random Forest based detection and classification is possible for the topic of age estimation from MR images. The goal is to implement a number of established algorithms and to improve upon the state of the art if possible. In the long run this might also become a topic for a PhD project, this depends on the feasibility which has to be shown in the master thesis.

Optimally, the student has some knowledge in medical image analysis and computer vision algorithms, and still needs a seminar/project and the master thesis for his/her studies. The master thesis part will be paid by the LBI-CFI in the form of a "Forschungsbeihilfe". Knowledge in C++ is important, knowledge and interest in CUDA GPU programming is beneficial.