Machine Learning for Postoperative Continuous Recovery Scores of Oncology Patients in Perioperative Care with Data from Wearables.

Assessing post-operative recovery is a significant component of perioperative care, since this assessment might facilitate detecting complications and determining an appropriate discharge date. However, recovery is difficult to assess and challenging to predict, as no universally accepted definition exists. Current solutions often contain a high level of subjectivity, measure recovery only at one moment in time, and only investigate recovery until the discharge moment. For these reasons, this research aims to create a model that predicts continuous recovery scores in perioperative care in the hospital and at home for objective decision making. This regression model utilized vital signs and activity metrics measured using wearable sensors and the XGBoost algorithm for training. The proposed model described continuous recovery profiles, obtained a high predictive performance, and provided outcomes that are interpretable due to the low number of features in the final model. Moreover, activity features, the circadian rhythm of the heart, and heart rate recovery showed the highest feature importance in the recovery model. Patients could be identified with fast and slow recovery trajectories by comparing patient-specific predicted profiles to the average fast- and slow-recovering populations. This identification may facilitate determining appropriate discharge dates, detecting complications, preventing readmission, and planning physical therapy. Hence, the model can provide an automatic and objective decision support tool.

Validation of Walking Speed Estimation from Trunk Mounted Accelerometers for a Range of Walking Speeds.

Walking speed is a strong indicator of the health status of older people and patients. Using algorithms, the walking speed can be estimated from wearable accelerometers, which enables minimally obtrusive (longitudinal) monitoring. We evaluated the performance of two algorithms, the inverted pendulum (IP) algorithm, and a novel adaptation correcting for lateral step movement, which aimed to improve accuracy during slow walking. To evaluate robustness, we gathered data from different groups (healthy adults, elderly, and elderly patients) of volunteers (n = 159) walking under various conditions (over ground, treadmill, using walking aids) at a broad range of speeds (0.11-1.93 m/s). Both of the algorithms showed good agreement with the reference values and similar root-mean-square errors (RMSEs) for walking speeds ≥0.5 m/s, which ranged from 0.09-0.16 m/s for the different positions, in line with the results from others. However, for slower walking, RMSEs were significantly better for the new method (0.06-0.09 m/s versus 0.15-0.19 m/s). Pearson correlation improved for speeds <0.5 m/s (from 0.67-0.72 to 0.73-0.82) as well as higher speeds (0.87-0.97 to 0.90-0.98) with the new method. Overall, we found that IP(-based) walking speed estimation proved to be applicable for a variety of wearing positions, conditions and speeds, indicating its potential value for health assessment applications.

Evaluation of an activity monitor for use in pregnancy to help reduce excessive gestational weight gain.

BACKGROUND: Excessive weight gain during pregnancy increases the risk for negative effects on mother and child during pregnancy, delivery, and also postnatally. Excessive weight gain can be partially compensated by being sufficiently physically active, which can be measured using activity trackers. Modern activity trackers often use accelerometer data as well as heart rate data to estimate energy expenditure. Because pregnancy affects the metabolism and cardiac output, it is not evident that activity trackers that are calibrated to the general population can be reliably used during pregnancy. We evaluated whether an activity monitor designed for the general population is sufficiently accurate for estimating energy expenditure in pregnant women. METHODS: Forty pregnant women (age: 30.8 ± 4.7 years, BMI: 25.0 ± 4.0) from all three trimesters performed a 1-h protocol including paced and self-paced exercise activities as well as household activities. We tracked reference energy expenditure using indirect calorimetry and used equivalence testing to determine whether the estimated energy expenditure from the activity monitor was within the limits of equivalence. RESULTS: Overall we found an averaged underestimation of 10 kcal (estimated energy expenditure was 97% of the reference measurement). The 90% CI for the cumulative total energy expenditure was 94-100%. The activities of self-paced cycling, household activities, stair-walking, and yoga had one of their equivalence boundaries outside a 80-125% range of equivalence; for exercise on a cross-trainer, for self-paced and fixed-pace walking, fixed-paced cycling, and resting, the estimations were within the limits of equivalence. CONCLUSIONS: We conclude that the activity monitor is sufficiently accurate for every-day use during pregnancy. The observed deviations can be accounted for and are acceptable from a statistical and an applied perspective because the positive and negative deviations that we observed cancel out to an accurate average energy expenditure over a day, and estimations during exercise are sufficiently accurate to enable coaching on physical activity. The positive and negative deviations themselves were relatively small. Therefore, the activity monitor can be used to help in preventing excessive weight gain during pregnancy by accurately tracking physical activity.

The turnover of mineralized growth plate cartilage into bone may be regulated by osteocytes.

During endochondral ossification, growth plate cartilage is replaced with bone. Mineralized cartilage matrix is resorbed by osteoclasts, and new bone tissue is formed by osteoblasts. As mineralized cartilage does not contain any cells, it is unclear how this process is regulated. We hypothesize that, in analogy with bone remodeling, osteoclast and osteoblast activity are regulated by osteocytes, in response to mechanical loading. Since the cartilage does not contain osteocytes, this means that cartilage turnover during endochondral ossification would be regulated by the adjacent bone tissue. We investigated this hypothesis with an established computational bone adaptation model. In this model, osteocytes stimulate osteoblastic bone formation in response to the mechanical bone tissue loading. Osteoclasts resorb bone near randomly occurring microcracks that are assumed to block osteocyte signals. We used finite element modeling to evaluate our hypothesis in a 2D-domain representing part of the growth plate and adjacent bone. Cartilage was added at a constant physiological rate to simulate growth. Simulations showed that osteocyte signals from neighboring bone were sufficient for successful cartilage turnover, since equilibrium between cartilage remodeling and growth was obtained. Furthermore, there was good agreement between simulated bone structures and rat tibia histology, and the development of the trabecular architecture resembled that of infant long bones. Additionally, prohibiting osteoclast invasion resulted in thickened mineralized cartilage, similar to observations in a knock-out mouse model. We therefore conclude that it is well possible that osteocytes regulate the turnover of mineralized growth plate cartilage.

A mathematical model to evaluate control strategies for mechanical circulatory support.

Continuous flow ventricular assist devices (VADs) for mechanical circulatory support (MCS) are generally smaller and believed to be more reliable than pulsatile VADs. However, regarding continuous flow, there are concerns about the decreased pulsatility and ventricular unloading. Moreover, pulsatile VADs offer a wider range in control strategies. For this reason, we used a computer model to evaluate whether pulsatile operation of a continuous flow VAD would be more beneficial than the standard constant pump speed. The computer model describes the left and right ventricle with one-fiber heart contraction models, and the systemic, pulmonary, and coronary circulation with lumped parameter hemodynamical models, while the heart rate is regulated with a baroreflex model. With this computer model, both normal and heart failure hemodynamics were simulated. A HeartMate II left ventricular assist device model was connected to this model, and both constant speed and pulsatile support were simulated. Pulsatile support did not solve the decreased pulsatility issue, but it did improve perfusion (cardiac index and coronary flow) and unloading (stroke work and heart rate) compared with constant speed. Also, pulsatile support would be beneficial for developing control strategies, as it offers more options to adjust assist device settings to the patient's needs. Because the mathematical model used in this study can simulate different assist device settings, it can play a valuable role in developing mechanical circulatory support control strategies.