A computer vision system for deep learning-based detection of patient mobilization activities in the ICU

Published by Daniel at December 1, 2022

This helpful tool can achieve these goals, and more comprehensive mobility data can be discovered using this innovative CVT...

A computer vision system for deep learning-based detection of patient mobilization activities in the ICU

Nature Partner Journals – Digital Medicine

Background

Most patients who need high-intensity care suffer from post-intensive care syndrome. To explain, once the patient has not used their cognitive functions or physical functions on their own, a decline in their functional status will be present. There isn’t a wide scope of studies on this particular topic since it is difficult to record data. However, the development of computer vision technology (CVT) algorithms that detect patient mobilization activities has been important in recent research. Computer algorithms use data of patients and staff from the clinical environment. Overall, there has been a higher interest in using CVT method to deliver better patient care and to perform activity recognition.

Methods

We present the results for two models predicting the occurrence of mobility and healthcare personnel involvement. Each model represents a subject of data on its own. This study focuses on validating computer vision algorithms to discover the occurrence of patient mobility activities, including other descriptive elements of mobility. To describe the subject’s duration and the quantity of assisting personnel.

Results

The computer algorithm for the detection of mobility occurrence can continue to be used since it is one of a few methods to detect mobility. It provides great sensitivity and specificity, respectively. The overall mean percentage for sensitivity was 87.2%, and the mean percentage for specificity was 89.2%. Therefore, the computer algorithm system accurately detects patient mobility through the four activities, the number of staff assisting, and their duration. This helpful tool can achieve these goals, and more comprehensive mobility data can be discovered using this innovative CVT.

Conclusions

This computer vision algorithm is one of very few ways to collect mobility data. There is a similar tool upon which CVT is built (Ma et al.); however, CVT displays more accurate and detailed data. This efficient tool allows researchers to study specific mobility events, and their variation in duration and frequency will impact clinical outcomes. Additionally, CVT will not only aid in the refinement of protocols for mobility, but it will give a better understanding of how they can be used more effectively.

Relevance to Healthcare Field

Patients suffering from post-intensive care syndrome have suffered enough. The mobilization of these patients will shorten the time of weaning off ventilation, reduce delirium, and prevent muscle dysfunction. With CVT, these patients will not only avoid these preventable harms but also acquire improved patient care in hospitals. In addition, CVT enhances our understanding of contributing factors for the best possible experience. Overall, this method will ensure that healthcare workers deliver the most optimal care.