This repo holds the algorithm development for modeling TMS data with Gaussian processes and use active learning to select optimum locations for the next stimulation.
- Implementation of papers:
- python 3.9.6
- numpy 1.20.3
- pandas 1.3.3
- scikit-learn 1.0
- data:
-
5 Subjects each with 3 maps (GRID,RAND,USER)
-
GRID MAP: 49 locations with 6 stimulations at each location (294 total stimulation) USER MAP: center + 4 corners defined remaining 289 stimulations chosen with user expertise (294 total stimulations) RANDOM MAP: center + 4 corners defined remaining 289 stimulations chosen from a uniform random distirubution (294 stimulations total)
-
Each data file (ie AS_3MAPS.mat) contains: NavDataGRID: xyz data of grid locations (cm) MEPampGRID: amplitude at each grid location (uV) NavDataRAND: xyz data of rand locations (cm) MEPampRAND: amplitude at each rand location (uV) NavDataUSER: xyz data of user locations (cm) MEPampUSER: amplitude at each user location (uV)
-
Note: No thresholding has been done on data.
-
Note: Some data sets will contain less than 294 stimulations. Stimulations were removed for poor coil location, subject head movement, etc
-
There is a figure provided showing maps for each subject
-
Code: "MEPmapInterpWithOutcomes.m"
Inputs: NavData, MEPamp, AmpThresh (usually set to 50) Outputs: Interpolated maps, map volume, map area, mean amplitude, center of gravity in x coordinate (COGx), center of gravity in y coordinate (COGy)
-
-
Faghihpirayesh, R., Imbiriba, T., Yarossi, M., Tunik, E., Brooks, D. and Erdoğmuş, D., 2020, June. Motor cortex mapping using active gaussian processes. In Proceedings of the 13th ACM International Conference on PErvasive Technologies Related to Assistive Environments (pp. 1-7).
-
Faghihpirayesh, R., Yarossi, M., Imbiriba, T., Brooks, D.H., Tunik, E. and Erdoğmuş, D., 2021. Efficient TMS-Based Motor Cortex Mapping Using Gaussian Process Active Learning. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 29, pp.1679-1689.