Medi-Cars: One Company’s Vision to Improve Crash Victim Care – Technologue

An Israeli startup called MDGo could put a real dent in the ambulance-chasing business if its bid to leverage deep learning and biomechanical modeling succeeds in turning cars into triage doctors.

The idea is elegantly simple: Use information gathered by a vehicle’s numerous multiaxis accelerometers and other sensors during an accident in order to help predict the injuries inflicted on the occupants, and share this data with first responders so they’re prepared to provide precisely the care needed. For example, the data could be used to immediately dispatch a medevac unit instead of waiting for an ambulance to arrive, assess, and then summon medevac. MDGo claims that—for survivable accidents—such tailored care and time savings could improve victims’ long-term outlook for mortality and morbidity (chronic pain/side-effects) by as much as 44 percent.

MDGo co-founder and CEO Itay Bengad’s trauma ward experience in medical school exposed him to many crash victims who received insufficient or incorrect care at the scene and/or en route to the hospital. This got him wondering whether technology aboard vehicles, coupled with crash-structure information and biomechanical modeling of the human body, could accurately and instantly predict injuries suffered by accident victims.

Since its January 2017 founding, the company has been obtaining and analyzing crash-test information from the open databases of crash-test authorities like NHTSA, IIHS, and Europe’s NCAP—covering five government-mandated crash test scenarios. To this, MDGo has added finite-element simulations of vehicles involved in many more scenarios, some of which the company verified by conducting its own crash tests.

Each prediction model involves some 22 million elements. The company has obtained finite-element models of 11 actual vehicles; the rest have been derived in-house, using open-source structure data to which MDGo adds its own safety equipment modeling. The company also investigated 10,000 crash tests and simulations in development of its own biomechanical occupant dummy model. The predictions also appear to be valid for male or female occupants (the weight of which modern cars sense and record for the model to use in the prediction).

This research has confirmed that vehicles of similar size, type, weight, and configuration can share a simulation model and yield reasonable injury prediction results. As such, it will never be necessary to obtain actual finite-element models of every vehicle. A handful of comprehensive models for each vehicle category is sufficient. And all such models will be forever evolving. Each of the tens of thousands of crashes that happen each day can inform a massive database, noting slightly new or different crash scenarios, against which actual reports of occupant injuries can potentially tweak the prediction model.

MDGo is currently assessing real-world crash data from a pilot program running in Israel since mid-2018. In at least 120 cases, telematics data was received from vehicles for which an appropriate model was available. In these cases, the data was received, analyzed within 8 seconds, and transmitted to the nearest pertinent trauma experts and medical facilities via a smartphone app. Comparing predictions against actual trauma reported by the nation’s sole ambulance/EMS service, Magen David Adom, reveals an impressive 92 percent accuracy rate.

All work conducted so far simply involves today’s suite of vehicle sensors. But over the next few years, MDGo looks to further enrich its data—and perhaps provide drivers and passengers with instant health checkups as they drive. That’s when the next generation of sensors begin rolling out, including facial recognition cameras and biometric devices that detect responses like heart rate, respiration, and pupil dilation.

The public benefits of MDGo’s technology are obvious. But the company’s business case is based on the savings and services it can provide to insurance companies—so they are the target customers. Turns out those crash-prediction models are also great at predicting vehicle damage, providing an instant repair estimate. Knowing exactly what happened in any accident—or saving the expense of determining (and perhaps litigating) what happened—promises to return about 25 percent to an insurance company’s bottom line, perhaps at the expense of those ambulance-chasers.

The post Medi-Cars: One Company’s Vision to Improve Crash Victim Care – Technologue appeared first on Motortrend.

from Motortrend https://www.motortrend.com/news/medi-cars-companys-vision-to-improve-crash-victim-care-technologue/