A Vision-based System for Hand Washing Quality Assessment with Real-Time Feedback

Abstract

Hand washing is a critical activity in preventing the spread of infection in healthcare environments and contamination in food processing industries. The World Health Organization (WHO) guidelines recommend a hand washing protocol consisting of different hand washing poses to ensure that all parts of the hands are thoroughly cleaned. The assessment and training of workers in proper hand hygiene technique is performed by human trainers and can be very tedious and time consuming. In this paper we present an automated computer vision system which measures the user's hand washing technique to ensure that the WHO guidelines are correctly followed. The main contribution of this work is a system which performs robust hand segmentation and hand washing pose classification. The performance of the system is analysed based on data collected in a clinical trial performed in a hospital ward. The performance of the system was compared with that of human auditors. The agreement between two human auditors and between the system and the human auditors were found to be of the same order, which emphasizes the accuracy and validity of the proposed automated system.

Designing a Computer-Vision Application: A Case Study for Hand-Hygiene Assessment in an Open-Room Environment

“…For example, we chose to passively observe people washing their hands which means they may not execute all steps and some actions may be out of camera view. In contrast, other handwashing systems have been designed to be interactive or even educational about what the next action to execute should be [ 4 , 5 ]. The interactive approach simplifies the image processing but constrains people to keep their hands over the sink in camera view during the washing activities.…”

“…Hand hygiene plays a critical role in reducing contamination in our food supply and in reducing disease transmission. Several image- and video-based systems have been developed to assess or assist hand hygiene for healthcare workers [ 5 , 6 , 8 , 9 , 10 , 11 , 12 ] and people with dementia [ 4 , 13 , 14 , 15 ]. In [ 8 , 9 ], image processing reports when hospital staff are present at a hand-washing station.…”

“…Robustness to lighting conditions is obtained by combining skin tone and motion during hand segmentation. A real-time and improved version of this system using linear discriminant analysis is presented in [ 5 ]. The COACH system (Cognitive Orthosis for Assisting aCtivities in the Home) is an interactive system that prompts patients with dementia to perform correct hand-hygiene when they are unable to perform it on their own [ 4 ].…”

“…One key difference between the system we develop here and the methods in [ 5 , 6 , 13 ] is that our cameras are not overhead, and the sink area where handwashing occurs may have extraneous objects. This increases the image and video processing challenges, but also opens up opportunities for the system to observe other food handling activities.…”

“…In this paper, we extend this work to consider an open-room environment with a single side-view stationary camera. We use the term “open-room environment” to contrast these environments with both the much smaller restrooms of our previous work and the top-down sink-only views of [ 4 , 5 , 6 ]. In the current open-room environment, the distances between the camera and the sink are much longer, and there are potentially many more distracting objects.…”