Data-driven Stroke Rehabilitation

In collaboration with the Mobilis lab at the NYU School of Medicine we designed deep-learning methodology to perform automatic identification and counting of functional arm movements in stroke patients from measurements obtained with wearable sensors and videos. We have also released the first public dataset for data-driven stroke rehabilitation. This research was supported by NIH grant R01 LM013316 and is currently supported by NSF grant IIS 2404476

• Quantifying impairment and disease severity using AI models trained on healthy subjects B. Yu, A. Kaku, K. Liu, E. Fokas, A. Venkatesan, N. Pandit, R. Ranganath, H. Schambra, C. Fernandez-Granda. npj Digital Medicine 7, 180. 2024. Video. Code and data

• Data-Driven Quantitation of Movement Abnormality after Stroke A. Parnandi, A. Kaku, A. Venkatesan, N. Pandit, E. Fokas, B. Yu, D. Nilsen, C. Fernandez-Granda, H. Schambra. Bioengineering 10(6), p.648 2023

• StrokeRehab: A Benchmark Dataset for Sub-second Action Identification A. Kaku, K. Liu, A. Parnandi, H. Rajamohan, K. Venkataramanan, A. Venkatesan, A. Wirtanen, N. Pandit, H. Schambra, C. Fernandez-Granda. Proc. 35th Conference on Neural Information Processing Systems (NeurIPS) Datasets And Benchmarks Track. 2022 Data

• PrimSeq: a deep learning-based pipeline to quantitate rehabilitation training A. Parnandi, A. Kaku, A. Venkatesan, N. Pandit, A. Wirtanen, H. Rajamohan, K. Venkataramanan, D. Nilsen, C. Fernandez-Granda, H. Schambra. PLOS Digital Health 1(6), p.e0000044. 2022 Data

• Towards data-driven stroke rehabilitation via wearable sensors and deep learning A. Kaku, A. Parnandi, A. Venkatesan, N. Pandit, H. Schambra, C. Fernandez-Granda. Proceedings of Machine Learning Research 126:143-171. Machine Learning in Healthcare (MLHC) 202