Direkt zum Inhalt springen
login.png Login    |
de | en
MyTUM-Portal
Technical University of Munich

Technical University of Munich

Sitemap > Bulletin Board > Diplomarbeiten, Bachelor- und Masterarbeiten > Master Thesis: Deep Learning for Wear Particle Analysis Based on Microscopy Imaging Data
up   Back to  News Board    previous   Browse in News  next    

Master thesis for Mechanical Engineering or Robotics, Cognition, Intelligence students.

Master Thesis: Deep Learning for Wear Particle Analysis Based on Microscopy Imaging Data

18.09.2022, Diplomarbeiten, Bachelor- und Masterarbeiten

Abstract: The generation of wear particles is one of the major reasons for revision of a total joint prosthesis. The reaction of the human body on wear debris highly depends on the size and shape of the particles generated. It is therefore of crucial importance to characterize those parameters in pre-clinical testing. Currently, the identification of wear particles is based on image analysis software which results in a high rate of wrong detections. As a result, the images and their resulting data need to be reinvestigated manually. The hypothesis of the study is that AI is capable to generate a lower failure rate as well as to distinguish more precisely between particles and their agglomerates. Over the last decades, Endolab GmbH has collected a large pool of image analysis data that can be used to train AI models. This dataset can be regarded as unique worldwide and shall be processed by state-of-the-art methodology to allow for better pre-clinical testing.

Tasks
Detection and classification of particles based on microscopy imaging data
Comparison of results to current tools
Adaption to task specific issues with AI
Presentation and discussion of results

Offer
Unique data with high potential for publication
Project in collaboration with academia and industry
Highly educated & interdisciplinary environment
Top level hardware for scientific computing

Prerequisites
Solid knowledge of deep learning and imaging data
Experience with detection and classification models
Beneficial but not mandatory: prior knowledge in biomechanics

References
Redmon, J., Divvala, S., Girshick, R. and Farhadi, A., 2016. You only look once: Unified, real-time object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 779-788).
Deng, Y., Yin, J., Wang, Y., Chen, J., Sun, L. and Li, Q., 2021, April. ResNet-50 based Method for Cholangiocarcinoma Identification from Microscopic Hyperspectral Pathology Images. In Journal of Physics: Conference Series (Vol. 1880, No. 1, p. 012019). IOP Publishing.
Grupp, T.M., Saleh, K.J., Mihalko, W.M., Hintner, M., Fritz, B., Schilling, C., Schwiesau, J. and Kaddick, C., 2013. Effect of anterior–posterior and internal–external motion restraint during knee wear simulation on a posterior stabilised knee design. Journal of Biomechanics, 46(3), pp.491-497.
https://www.endolab.org/simulator-knee-implants.asp


If you are interested, please send a CV and a transcript of records to: florian.hinterwimmer@tum.de & Christian.Kaddick@endolab.de

Kontakt: florian.hinterwimmer@tum.de, Christian.Kaddick@endolab.de

More Information

offer Master Thesis: Deep Learning for Wear Particle Analysis Based on Microscopy Imaging Data, (Type: application/vnd.openxmlformats-officedocument.presentationml.presentation, Size: 544.7 kB) Save attachment