MDAI: Burn Study Topline Data

In This Article:

By John Vandermosten, CFA

NASDAQ:MDAI

READ THE FULL MDAI RESEARCH REPORT

Spectral AI, Inc. (NASDAQ:MDAI) reported topline data from its burn validation study in a March 17th press release confirming DeepView’s superior detection sensitivity relative to burn physicians. The announcement reminded stakeholders of the device’s background and development along with a description of the burn study’s trial design. Data regarding DeepView’s sensitivity, specificity and Dice Score were provided. Submission of the De Novo application to the FDA is expected in mid-2025. Spectral communicated other important information recently including the reiteration of access to additional financing, three device installations in Australia, progress in the intended spin-off of Spectral IP and the announcement of its 2024 financial results on March 27th.

Burn Validation Study Results

On February 7th, Spectral announced that it had completed the data analysis of the burn validation study presaging the release of topline data on March 17th. Results comparing the sensitivity, Dice score and specificity were provided along with anticipated timing for clearance and commercialization.

Burn Trial Background

On January 11th, 2024, Spectral announced that it began enrollment of its pivotal study to validate DeepView for burn injuries. The study enrolled and analyzed data from 164 patients, consisting of 115 adults and 49 children in burn centers and emergency departments throughout the US. The study was filed under clinicaltrials.gov under the designator NCT06131203 and is titled Burn Pivotal Study. Its purpose is to validate the burn healing assessment algorithm for the DeepView device. 16 sites enrolled patients.

Analysis of study data generated results related to the sensitivity, specificity and Dice score of the DeepView system at both a pixel level and full image level compared with the clinical judgment of burn physicians. Findings of the various endpoints were:

  • Sensitivity - The DeepView System demonstrated a statistically significant improvement in identifying non-healing tissue compared to burn physicians, as judged on sensitivity. At the image-wise level, the DeepView System scored 86.6% while clinical judgment annotation (CJA) of burn physicians scored 40.8%. At the pixel-wise level, the DeepView System scored 81.9% and CJA of burn physicians scored 38.8%;

  • Dice Score[1] - The DeepView System achieved statistically significant higher Dice Scores when compared to those derived from burn physicians’ CJA, representing the improved pixel-wise evaluation between predicated and true segmented wound areas with the DeepView System performing at 68.5% and burn physician’s CJA at 39.2%;

  • Specificity - The DeepView System produced a specificity of 61.2% (versus an anticipated result of 36.0%) and CJA of 79.1%.