Automatic detection of the onset of nursing activities using accelerometers and adaptive segmentation.

We used the Recursive Least Squares algorithm and a predictor filter to automatically identify the start and stop times of 6 simple nursing activities. The dataset included continuous acceleration recordings obtained with a single accelerometer sensor attached to the backs of 8 nurses. The algorithm requires no training. It identifies the start and stop time of each activity when at least 2 of 3 axes show significant acceleration changes not more than a second apart. The overall accuracy of the algorithm for a total of 96 start and stop events was 86.46% ± 12.55%. The accuracy was higher than 91% for 5 out of 8 subjects. The algorithm also indicated the onset of subcomponents of nursing activities for the majority of the subjects. The results of this study suggest that the presented algorithm may be useful in identifying transition points of human activities recorded with accelerometers.

Nursing activity recognition using an inexpensive game controller: An application to infection control.

It is estimated that 10% of the patients admitted to North American hospitals die of hospital acquired infections. Approximately half of these are thought to be a consequence of poor hand hygiene practices by the hospital staff. Electronic hand washing reminders that prompt caregivers to wash their hands before and after the patient/patient's environment contact may help to increase the hand hygiene compliance rate. However, the current systems fail to identify the nursing procedures happening around the patient to issue proper hand hygiene prompt. In this research we used the hardware of a low-cost wireless Sony game controller, which included a 3-axis accelerometer, to identify six nursing activities happening around a patient. We attached five sensors to eight nurses' left and right wrists, left and right upper arms, and the backs. Each nurse performed 10 trials of each nursing activity in sequence, followed by a combined nursing activities trial. We extracted mean, standard deviation, energy, and correlation among axes per sensor and compared the results of 1-Nearest Neighbour (1-NN), Decision Tree (J48), and Naïve Bayes classifiers. 1-NN classifier had the best performance and on average regardless of the sensor locations, we achieved 84% ± 2% accuracy.

Advanced technologies to curb healthcare-associated infections.

The commentary was prepared in response to the manuscript "Healthcare-Associated Infections as Patient Safety Indicators," by Gardam, Lemieux, Reason, van Dijk and Goel. Healthcare-associated infections are a severe patient safety hazard. Current patient safety initiatives targeting increased healthcare worker hand hygiene to prevent some of these infections have had limited effect. This commentary describes recent advances in electronic sensing and computational power that have provided new options to increase hand hygiene compliance as a step toward reducing healthcare-associated infections. Smart electronics can provide reasoning about a healthcare worker's circumstance and prompt the worker to perform hand hygiene when necessary. These novel approaches in technology development have tremendous potential to enhance the hand hygiene of healthcare workers and can support the prevention of this significant problem for patients in our hospitals.

Real-time classification of forearm electromyographic signals corresponding to user-selected intentional movements for multifunction prosthesis control.

Pattern recognition-based multifunction prosthesis control strategies have largely been demonstrated with subsets of typical able-bodied hand movements. These movements are often unnatural to the amputee, necessitating significant user training and do not maximally exploit the potential of residual muscle activity. This paper presents a real-time electromyography (EMG) classifier of user-selected intentional movements rather than an imposed subset of standard movements. EMG signals were recorded from the forearm extensor and flexor muscles of seven able-bodied participants and one congenital amputee. Participants freely selected and labeled their own muscle contractions through a unique training protocol. Signals were parameterized by the natural logarithm of root mean square values, calculated within 0.2 s sliding and non overlapping windows. The feature space was segmented using fuzzy C-means clustering. With only 2 min of training data from each user, the classifier discriminated four different movements with an average accuracy of 92.7% +/- 3.2%. This accuracy could be further increased with additional training data and improved user proficiency that comes with practice. The proposed method may facilitate the development of dynamic upper extremity prosthesis control strategies using arbitrary, user-preferred muscle contractions.