Machine Learning for extracting Muscle features using Ultrasound
From iis-projects
Contents
Short Description
Ultrasound imaging is a non-invasive imaging technique that provides visible information on the structure of musculoskeletal tissues and organs. The development of wearable ultrasound probes would enable real-time non-invasive continuous monitoring of physiological parameters during the day, which is of particular interest for medical therapies and sport science. However, efficient machine learning (ML) algorithms are required in order to automatically extract the physiological parameters of interest (i.e., the length of muscle fascicles). In the context of this thesis, you will work with the raw data acquired by a fully-digital ultrasound probe, to build ML algorithm for the extraction of such physiological features.
Goal & Tasks
The goal of this project is to investigate the possibility of extracting relevant physiological information (the length of muscle fascicles) directly from raw Ultrasound data. The main tasks of this project are:
- Literature study of ultrasound imaging principles
- Neural network training and testing, based on available datasets of labeled images
- Possible optimization and deployment of the network on an embedded platform
Prerequisites
- Basics of Machine Learning (required) and Deep Learning (desirable)
- MATLAB
- Python (sklearn, tensorflow)
- C
Status: Completed
Soley Hafthorsdottir
- Supervision: Sergei Vostrikov, Andrea Cossettini, Michael Rieder
Character
- 15% Literature Study
- 10% Dataset processing
- 35% Training and testing of networks
- 40% Deployment on FPGA or microcontroller