Identifying and training deep learning neural networks on biomedical-related datasets.

This manuscript describes the development of a resources module that is part of a learning platform named 'NIGMS Sandbox for Cloud-based Learning' https://github.com/NIGMS/NIGMS-Sandbox. The overall genesis of the Sandbox is described in the editorial NIGMS Sandbox at the beginning of this Supplement. This module delivers learning materials on implementing deep learning algorithms for biomedical image data in an interactive format that uses appropriate cloud resources for data access and analyses. Biomedical-related datasets are widely used in both research and clinical settings, but the ability for professionally trained clinicians and researchers to interpret datasets becomes difficult as the size and breadth of these datasets increases. Artificial intelligence, and specifically deep learning neural networks, have recently become an important tool in novel biomedical research. However, use is limited due to their computational requirements and confusion regarding different neural network architectures. The goal of this learning module is to introduce types of deep learning neural networks and cover practices that are commonly used in biomedical research. This module is subdivided into four submodules that cover classification, augmentation, segmentation and regression. Each complementary submodule was written on the Google Cloud Platform and contains detailed code and explanations, as well as quizzes and challenges to facilitate user training. Overall, the goal of this learning module is to enable users to identify and integrate the correct type of neural network with their data while highlighting the ease-of-use of cloud computing for implementing neural networks. This manuscript describes the development of a resource module that is part of a learning platform named ``NIGMS Sandbox for Cloud-based Learning'' https://github.com/NIGMS/NIGMS-Sandbox. The overall genesis of the Sandbox is described in the editorial NIGMS Sandbox [1] at the beginning of this Supplement. This module delivers learning materials on the analysis of bulk and single-cell ATAC-seq data in an interactive format that uses appropriate cloud resources for data access and analyses.

Effects of single-variable biofeedback on wheelchair handrim biomechanics.

OBJECTIVE: To determine the effects of single-variable biofeedback on select wheelchair propulsion variables. DESIGN: Within-subject comparisons. SETTING: Biomechanics laboratory. PARTICIPANTS: Manual wheelchair users (N=31). INTERVENTIONS: Biofeedback on braking moment, cadence, contact angle, peak force, push distance, and smoothness were presented on a large monitor during propulsion on a motor-driven treadmill. For each variable, subjects were asked to make a maximum improvement, as well as a targeted 10% improvement for cadence, contact angle, peak force, and push distance. MAIN OUTCOME MEASURES: Relative differences (%) in each variable between the normal propulsion trial and the biofeedback trials. RESULTS: Subjects were able to interpret and respond to the biofeedback successfully. For the maximum change conditions, significant improvements were made to all variables except smoothness, with individual improvements of 11% in peak force, 31% in contact angle, 44% in braking moment, 64% in cadence, and 255% in push distance. For the 10% target conditions, improvements were achieved to within 1% for all variables except peak force, which was a difficult variable for most subjects to control. Cross-variable interactions were found for most variables, particularly during the maximum change conditions. Minimizing cadence led to a 154% increase in peak force, suggesting the need for multi-variable feedback if multiple training objectives, such as reducing cadence and peak force simultaneously, are desired. While subjects were unable to significantly change smoothness, efforts to push more smoothly led to improvements across most outcome variables. CONCLUSIONS: Biofeedback can be used to improve specific aspects of wheelchair propulsion. Cadence, contact angle, and push distance are well controlled by wheelchair users, and may be useful for clinical propulsion training. Clinicians should be aware of and comfortable with any cross-variable effects resulting from single-variable biofeedback training.

The effects of four different stroke patterns on manual wheelchair propulsion and upper limb muscle strain.

PURPOSE: To evaluate the effects of stroke pattern on handrim biomechanics and upper limb electromyography (EMG) in experienced wheelchair users. METHOD: Subjects propelled their own wheelchair on a level, motor-driven treadmill using each of four identified stroke patterns: arcing, double loop (DL), semi-circular (SC) and single loop (SL). Upper limb EMG and measurements taken from an instrumented wheelchair wheel were compared for each pattern. A one-way ANOVA with Bonferroni correction (p < 0.05) was used to check for significant differences. RESULTS: The DL and SC patterns produced the best overall results. The DL pattern led to a significantly longer contact angle and significantly less braking moment than the SL and arcing patterns, and a significantly lower cadence than the SL pattern. The SC pattern led to a significantly longer contact angle than the SL pattern and the lowest peak force and impact of any pattern. There were no significant differences in integrated EMG (IEMG); however, the DL and arcing patterns produced lower combined IEMG values. CONCLUSIONS: When traversing level terrain, wheelchair users should push with either the DL or SC patterns. Between the two, the DL pattern required less muscle activity and may be a better choice for experienced wheelchair users.