Further investigation is required to confirm that our game design promotes physical activity.

Physical inactivity is increasing among children globally and has been directly linked to the growing problems of overweight and obesity. We aim to assess the impact of a new mobile exergame, MobileKids Monster Manor (MKMM), in a school-based setting. MKMM, developed with input from youth to enhance physical activity, is wirelessly connected to an accelerometer-based activity monitor. Forty-two healthy students (11.3 ± 1.2 years old and 0.28 ± 1.29 body-mass index [BMI] z-score) participated in a randomized 4-week crossover study to evaluate the game intervention. The two study arms consisted of week-long baseline, game intervention/control, washout, and control/game intervention phases. All participants were required to wear an activity monitor at all times to record steps and active minutes for the study duration. MKMM was used during each arm's respective intervention week, during which children were asked to play the game at their convenience. When children were exposed to the game, an increase compared with the control phase of 2,934 steps per day (p = 0.0004, 95% CI 1,434-4,434) and 46 active minutes per day (p = 0.001, 95% CI 20-72) from baseline (12,299 steps/day and 190 active minutes/day) was observed. A linear regression model showed that MKMM yielded a greater increase in steps and active minutes per day among children with a higher BMI z-score, showing 10 percent more steps per day and 14 percent more active minutes per day relative to baseline, per unit increase in BMI z-score. In conclusion, MKMM increased steps and active minutes in a school-based environment. This suggests that mobile exergames could be useful tools for schools to promote physical activity and combat obesity in adolescents.

We present a fully convolutional neural network for segmenting ischemic stroke lesions in CT perfusion images for the ISLES 2018 challenge. Treatment of stroke is time sensitive and current standards for lesion identification require manual segmentation, a time consuming and challenging process. Automatic segmentation methods present the possibility of accurately identifying lesions and improving treatment planning. Our model is based on the PSPNet, a network architecture that makes use of pyramid pooling to provide global and local contextual information. To learn the varying shapes of the lesions, we train our network using focal loss, a loss function designed for the network to focus on learning the more difficult samples. We compare our model to networks trained using the U-Net and V-Net architectures. Our approach demonstrates effective performance in lesion segmentation and ranked among the top performers at the challenge conclusion.