Reflections and Observations
Thanks to everyone for their extensive comments!
It was great to see everyone from HC. I believe I met three esteemed members, though unfortunately, I didn’t realize that for the afternoon AGM, the meet-and-greet happenedbeforethe formal proceedings. As a result, I arrived with only minutes to spare.
Below are my thoughts, adding my perspective where suitable (these are my opinions and not the company’s).
Gen1/Emu1 vs. Gen2/Ambulance Model
•TheGen1/Emu1device (with 16 antennas) requires scanning in two planes (scan, move, scan), whereas the more advancedGen2(ambulance model) uses 28 antennas (2 rings plus 2 down near the neck).
•Based on past conversations with Scott and others in the startup space, it’s critical to release aminimum viable product (MVP)to start generating revenue. Delaying this stage could lead to years of development and excessive cash burn.
•Emu1might be frozen at a usable stage for deployment, withEmu2later marketed as an upgrade (e.g., after three years) featuring more sensitive or additional antennas for the next capital cycle.
•When I compared theGen1 vs. Gen2 antennasa few years ago, the difference was striking:
•Gen1:~200g each, about the size of a soap bar, and all metal.
•Gen2:~10g each, resembling a wafer-thin CPU board.
•This likely explains whyGen1relies on an articulated arm due to its weight.
It was great to haveStuart Crozieron hand to address technical questions. Scott also demonstrated an excellent understanding of the technology. Stuart mentioned he works three days a week at EMV.
USA Operations and Data Validation
•Regarding Austin (USA), I understand thatChristianandForoughare currently there.
•I inquired about the one hemorrhage case missed out of 13. Stuart explained the context, but Scott clarified that it wasn’t mistaken for ischemia, which would have been more concerning.
The analysis of thethree ischemia casesmissed appears ongoing. Stuart suggested possibilities such as scanning or boundary errors, and the results are expected to be published in a high-impact journal soon.
Later, informal conversations indicated that, in addition to the two Australian sites involved in the FDA validation trial,two additional Australian siteswill collect data. This will enhance sensitivity and specificity metrics.
Challenges with AI and Generalization
AI generalizability remains a significant challenge. Many studies have shown a disparity between sensitivity and specificity during internal testing versus real-world application:
•For example, the study“Diagnostic Accuracy and Failure Mode Analysis of a Deep Learning Algorithm for the Detection of Cervical Spine Fractures”highlights these issues with results significant below expected.
•Similarly, the systematic review“External Validation of Deep Learning Algorithms for Radiologic Diagnosis”found that over 75% of algorithms showed modest or significant underperformance in external settings.
That said, EMV has an advantage: data is acquired using thesame type of machineand processed consistently. This minimizes variability compared to, say, CT scanners, where different vendors and post-processing software can significantly affect results.
Mimics and Anatomical Imaging
•Stuart noted that normal brains are clearly identifiable as such, which is promising. However, further focus will be needed on brains with anomalies (e.g., clips, coils, shunts, calcium deposits, or trauma-related scarring).
•The potential foranatomical imagingis exciting, though Scott emphasized it’s not currently intended for diagnostic purposes (likely due to a lack of validation). With proper studies, this capability could align imaging with clinical needs.
Cost and Market Opportunities
•Amobile CT scannercosts ~$1.6 million and requires a radiographer, making it resource-intensive. In contrast,Gen2could be deployed by a paramedic, making it more accessible and cost-effective.
•Theaddressable marketfor the technology is estimated at$15 billion, with a goal of capturing 20%.
Traumatic Brain Injury (TBI)
•PD provided an excellent explanation of the rotational shear forces causing gray-white matter axonal disruption in TBI.
•Thisdiagramhelps illustrate the issue, showing microhemorrhages (red spots):
Key Events to Watch for in 2025
1.Trial activation.
2.First responder activation trial.
3.Regulatory submissions toFDA, TGA, and CE.
4.Establishment of go-to-market distributors (USA).
Emerging Use Cases
Recent studies suggesttight blood pressure controlin hemorrhagic stroke can improve outcomes, presenting another potential use case forbedside monitoring of blood pressure and hemorrhage. A paper is in progress to quantify the sensitivity and specificity needed to demonstrate cost-benefit advantages.
(20min delay)