Off-the-shelf Algorithms for Classification and Extrinsic Regression

• Grant number: 24/07047-7
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: June 01, 2024
• End date: May 31, 2025
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Diego Furtado Silva
• Grantee: Gabriel da Costa Merlin
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil
• Associated research grant: 22/03176-1 - Machine Learning for Time Series Obtained in mHealth Applications, AP.PNGP.PI

Abstract

Monitoring physiological signals, vital signs, and other parameters that can be collected over time is essential in various healthcare tasks, such as estimating heart rate and identifying abnormal heartbeats. However, costly and usually non-portable equipment is the usual means to obtain these time series. On the other hand, with the improvement and miniaturization of sensors capable of transmitting various types of data, mobile and wearable devices have increasingly shown themselves as options to support medical decisions. Smartphones and smartwatches have increasingly precise and diverse sensors, leading the World Health Organization to consider that so-called mobile health (mHealth) could represent a revolution in how populations interact with public health systems. However, making mHealth applications viable in a practical scenario demands overcoming several scientific and technological challenges. Among these challenges are the need for low-cost methods, the heterogeneity and multimodality of data, and the difficulty in obtaining annotated data. In this scenario, this project proposes to investigate the performance of Machine Learning algorithms for time series applied to health data with different levels of resource availability. By the end of this research, we aim to have a broad and deep understanding of the quality of algorithms in physiological signals such as ECG and PPG with few and many annotated examples, low and high sampling rates, and various numbers of patients in the dataset.