[Paper Review] Towards Vision-Based Smart Hospitals: A System for Tracking and Monitoring Hand Hygiene Compliance

Albert Haque, Michelle Guo+11 more|arXiv (Cornell University)|Aug 1, 2017

IoT-based Smart Home SystemsEngineering17 references73 citations

TL;DR

The paper presents a non-intrusive depth-image–based vision system to detect, track, and classify hand hygiene events in hospitals, outperforming covert observations and RFID-like proximity baselines for compliance assessment.

ABSTRACT

One in twenty-five patients admitted to a hospital will suffer from a hospital acquired infection. If we can intelligently track healthcare staff, patients, and visitors, we can better understand the sources of such infections. We envision a smart hospital capable of increasing operational efficiency and improving patient care with less spending. In this paper, we propose a non-intrusive vision-based system for tracking people's activity in hospitals. We evaluate our method for the problem of measuring hand hygiene compliance. Empirically, our method outperforms existing solutions such as proximity-based techniques and covert in-person observational studies. We present intuitive, qualitative results that analyze human movement patterns and conduct spatial analytics which convey our method's interpretability. This work is a step towards a computer-vision based smart hospital and demonstrates promising results for reducing hospital acquired infections.

Motivation & Objective

• Motivate smart hospitals to reduce hospital-acquired infections (HAIs) through continuous monitoring of staff and patient interactions.
• Develop a non-intrusive, privacy-safe vision-based pipeline to detect, track, and assess hand hygiene compliance in hospital units.
• Provide interpretable analytics of movement patterns and space usage to inform workflow optimization and hygiene policy.
• Demonstrate feasibility and quantify performance against baselines such as covert observation and RFID-based proximity methods.

Proposed method

• Detect 3D pedestrian positions using a sparsity-driven occupancy pursuit formulation.
• Track pedestrians across cameras by solving a MAP/linear-integer flow optimization to link detections into trajectories.
• Classify hand hygiene events from depth images using a viewpoint-invariant representation with a spatio-temporal (ST) spatial transformer network integrated into a depth+pose+foreground feature extractor.
• Fuse tracking and classification by spatio-temporal matching to associate dispenser events with specific tracks and compute compliance rates.
• Evaluate classifier with ResNet-152 and depth/pose/foreground inputs, showing improvement when incorporating spatial transformer networks.

Experimental results

Research questions

• RQ1 Can a depth-image–based, non-intrusive vision system accurately detect and track hospital staff in order to assess hand hygiene events?
• RQ2 How does vision-based hand hygiene assessment compare to covert in-person observations and RFID-based proximity baselines in terms of accuracy?
• RQ3 Does incorporating viewpoint-invariant features and spatial transformer networks improve hand hygiene event classification from depth data?
• RQ4 Can tracking outputs be fused with activity classifications to yield reliable compliance metrics and insightful space analytics?

Key findings

• The proposed system achieves 75% accuracy in hand hygiene compliance assessment, outperforming both single (63%) and group (72%) covert observations and the RFID-based proximity baseline (18%).
• The depth-based approach enables non-intrusive, privacy-safe monitoring suitable under HIPAA/GDPR constraints by using de-identified depth data instead of color images.
• Hand hygiene activity classification benefits from augmenting depth maps with foreground and pose cues, and further improves with a spatial transformer network that provides a 1% accuracy boost in their experiments.
• Qualitative visualizations demonstrate interpretable transformations and localization that highlight regions of interest and support trust with clinicians.
• The paper provides analytics of hospital space usage and movement patterns, enabling potential workflow optimization and targeted infection-prevention strategies.