Home // eTELEMED 2024, The Sixteenth International Conference on eHealth, Telemedicine, and Social Medicine // View article

A Case Study for Scoliosis: How MLOps Can Help Reduce AI Challenges in Health Care?

Authors:
Gábor György Gulyás
Janis Lapins
Attila Csaba Kiss

Keywords: machine learning, edgeML, health care, MLOps, pose estimation.

Abstract:
The integration of Artificial Intelligence (AI) into healthcare diagnostics represents a significant advancement, particularly in the screening for conditions, such as scoliosis. This paper discusses the development, implementation, and evaluation of the Posture Buddy (PB) device, a machine vision driven tool designed to enhance the efficiency of scoliosis screening among school-aged children within the Hungarian health visitor system. Through the lens of Machine Learning Operations (MLOps) practices, our case study demonstrates the pivotal role of MLOps in overcoming operational hurdles at the intersection of eHealth and AI. The field-testing of PB revealed that within the context of low light conditions and slight side viewing angles the device performance decreases. In a later phase of the project, the pose estimation model of the device was put through model validation, observing the same flaw. Through these findings, the importance of proactive validation of AI models in healthcare is highlighted, whereas it also underscores the need to use MLOps to enable continuous deployment through the lifecycle of ML-based medical tools.

Pages: 29 to 36

Copyright: Copyright (c) IARIA, 2024

Publication date: May 26, 2024

Published in: conference

ISSN: 2308-4359

ISBN: 978-1-68558-167-1

Location: Barcelona, Spain

Dates: from May 26, 2024 to May 30, 2024