Gait analysis algorithm in neurodegenerative diseases

• Main Author: Huang, Siteng
• Other Authors: Vidya Sudarshan
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science; Generative adversarial network; Computer vision; Deep learning
• Online Access: https://hdl.handle.net/10356/175763
• Institution: Nanyang Technological University

Gait Anaylsis is an established method of evaluating the motion of a person, which aids in the diagnosis of various illnesses. In the current state, gait analysis is carried out by a trained professional in a medical facility. This means that patients will face the usual problems plaguing the healthcare system, such as long appointment times, therefore resulting in delayed or missed diagnosis which adversely affects the patient. To address the aforementioned problems, in recent years research efforts has gone into automating the diagnostics pipeline, through the use of Artificial Intelligence or creating tailored algorithms using sensor data. In the specific realm of using computer vision to classify pathological gaits, data in the form of gait videos are often difficult to obtain as it often involves patient details and are highly controlled, it is therefore a demanding task to collect sufficient gait data for training unless the researcher is able to form collaborations with specific healthcare institutions. This work is thus focused on generating high quality artificial gait data through the use of Generative Adversarial Networks, which is shown to be able to generate data that has high semblance to real training data. A new metric is also proposed to measure the quality of gait data generated and various methods are evaluated to understand if it can bring about improvements in the quality of data generated.