The popularity of wearable sensors being used to determine head impacts has been on the rise. As various research groups and organizations have been working on developing mouthguard sensors, we have been building a platform that can integrate with these sensors and provide a quantitative description of brain tissue strain and other brain injury metrics. We use custom-built cloud-based finite element modeling to compute the physics-based nonlinear response of the intracranial tissue and provide both a frontend web application as well as an Application Programming Interface (API) that will enable groups working on sensor technology to include simulation injury predictions directly into their existing research pipeline. In order to achieve scalability, multiple serverless microservice functions were developed. These serverless functions can run scripts/code in the backend when triggered by a specific event hence allowing us to decouple the frontend from the backend services.In order to test the scalability of the platform, our objective was to demonstrate that running multiple simulations would not take any longer than running a single simulation. To run these simulations, impact data from Prevent Biometrics was obtained from the Federal Interagency Traumatic Brain Injury Research (FITBIR) website for 1016 impacts. The FITBIR platform contains extensive data on TBI research that has been shared by various research groups and can be accessed by someone who has an account with the platform. Using this data, we were able to incrementally increase the number of simulations from 1 to 1016. We tested “on-demand” and “spot” instances on Amazon Web Services in the US-EAST region. The time taken for any individual job to complete is the sum of the queue and simulation times. A single simulation takes approximately 29 minutes to complete for 50 ms of impact data.For on-demand instances, all simulations were completed within 95 minutes and cost $336.85 USD. However, not all jobs started at once; 44% of the simulations started in 1 minute, but the average time a simulation was waiting was 34 minutes. For the spot instances, the quickest simulation finished within 31 minutes, however, all simulations finished running after 12 hours and 2 minutes. The maximum wait time was 11 hours and 58 minutes while the average wait time was 6 hours and 1 minute. In comparison to the on-demand instances, the spot instances have significantly larger wait times, however, the cost for running all simulations on spot instances was $154.62 USD which is less than 50% of the cost of the on-demand instances.In conclusion, for a small number of simulations (&lt; 500) all simulations can be completed within the time it takes to complete a single simulation. However, as the number of simulations increases above 500, the wait time to start a simulation also increases. Once a simulation starts it consistently takes approximately 29 minutes. &nbsp;In the future, we plan to expand to more AWS computing regions which could possibly reduce wait times.

Cost and Scalability Analysis of a Cloud-Based Brain Computing Service

Paper Proceedings: https://asmedigitalcollection.asme.org/IMECE/IMECE2021/volume/85543

ASME 2021 International Mechanical Engineering Congress and Exposition (IMECE2021)

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.