UtopiaCompression leverages its expertise in Image Processing and Artificial Intelligence to provide clinical diagnostic decision-support and monitoring tools. Our solutions allow medics and clinicians to screen for presence of an injury or abnormality and aid in the determination of the next course of action. Our lightweight software solutions are designed to be portable and platform agnostic. They are compatible with Android tablets/cell phones and any custom multimodal sensing system or hub. We actively collaborate with medical sensor providers to integrate our solutions.

Diagnostic-support tool for Traumatic Brain Injury (TBI)

Based on a decade of preclinical and clinical research, UC and its collaborators developed a novel screening tool for concussion detection. The screening tool is built using state-of-the-art portable ultrasound coupled with 3D modeling and custom AI analytics. It can be used to process 3D data of the optic nerve sheath and determine the likelihood of mild TBI or concussion. The tool can be used to detect, assess, and track TBI and its recovery over time.

Physiological monitoring to determine blood loss and cardiac arrhythmia

Using genetic algorithms and hierarchical machine learning algorithms, we developed software for cardiac arrhythmia detection for first responders. Depending on the number of on-body sensors, our software can perform cardiac status assessment. The results can be relayed to a central system to evacuate injured personnel who need immediate attention. We developed blood loss monitoring algorithms using data from USAISR based on inputs from an on-body pulse oximeter. Using Kalman filtering algorithms, we achieved an accuracy of 92% on USAISR data. For TATRC, a user interface was developed for the Android platform to monitor blood loss on multiple injured soldiers.

Non-invasive multimodal imaging for burn injury diagnosis

UC published work in Burns journal on fusing data from non-contact imaging systems, mainly OCT and spectroscopy, to assess extent of burn injury. Clinical evaluation of burn injuries results in less than 50% accuracy. By extracting imaging and spectral features from OCT and spectroscopy, UC has been able to develop machine learning algorithms to diagnose burn depth with 86% accuracy across all burn depths.