A Learning-Based System for Automatic Intentional Non-Adherence Detection from Dosing Videos
Shaolei Feng, Xiaoguang Lu, Deshana Kaushal Desai, Lei Guan
Abstract
Patient adherence is pivotal in clinical trials for new pharmaceuticals. Ensuring adherence is essential for robust safety and efficacy analyses. Intentional non-adherence, marked by the patient’s deceptive actions during dosing, complicates the accuracy of measurement. This paper proposes a novel learning-based system combining vision and metadata for detecting potentially deceptive dosing videos. Exploiting neural networks’ image understanding, it integrates visual and contextual data through ensemble learning. The system, efficient and adaptable, pioneers real-world deception capture, boosting adherence precision in trials. Our experiments show its remarkable real-world performance <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup>.
BibTeX
@inproceedings{icassp2024_alearningbasedsy,
title = {A Learning-Based System for Automatic Intentional Non-Adherence Detection from Dosing Videos},
author = {Shaolei Feng and Xiaoguang Lu and Deshana Kaushal Desai and Lei Guan},
booktitle = {ICASSP 2024},
year = {2024}
}