Great stuff @jcmleng and @Strawman I definitely take your point

i am travelling currently but will definitely take a better look. I apologise for being unduly negative- it is definitely a bias in this area for me!

@Chagsy no need to apologise at all! I totally appreciate the pushback/challenge as it forces me to take a closer look, which then helps me convince myself that all is still well. So please keep them coming!

I actually had a look at the AGM slides which got you going and this was interesting.

This was the AGM slide which you based your calculations on.

This was in the ASX announcement prior to the 1HFY25 results notification:

So, there was a stuff up in the EMV 1H24 slides, which you rightly picked up ...

This is a data point that no one cared to look at the detail of the AGM slides, myself included, except you! Note to myself, RTF slides!

This is well outside my wheelhouse @Chagsy -- but I pasted your post and original bear case into chatGPT, which replied with:

To clarify the confusion around the 85% ischemic stroke sensitivity quoted by EMVision: that figure was from an updated AI algorithm performance using a cleaned and re-evaluated dataset as part of their Continuous Innovation Study, not the pivotal trial. According to the May 2025 announcement, this retrained model correctly identified 19 out of 20 ischemic cases, which equates to 95% sensitivity, not 60%.

The pivotal trial underway now is specifically designed to validate these algorithms on a statistically significant sample, and EMVision is blinded to the ongoing data, so no premature claims are being made off the pivotal study results just yet.

That said, you’re right to highlight the challenge of proving clinical utility beyond just diagnostic accuracy — particularly in acute ischemic stroke, where treatment pathways are evolving (e.g., toward mechanical thrombectomy over thrombolysis). That’s a fair concern, and one EMVision seems to acknowledge, with increased emphasis on triage applications, point-of-care settings, and longer-term AI-driven diagnostics — not just as a “tPA-or-not” tool.

Re: your point on “maybe I’m missing something” — it seems the 60% figure you're quoting is from a previous dataset or a misinterpretation of the newer AI test results, rather than the latest reported performance. But it’s absolutely worth keeping a close eye on how those pivotal trial numbers stack up once released.

Not sure if that tracks with what you're seeing? The May 2025 announcement referenced is here: Enhanced-Algorithm-Performance-From-Continuous-Innovation (1).PDF

Thanks @Strawman

I was using the numbers they provided on the slide that @jcmleng put up. The new AI model certainly changes the accuracy.....if validated. As with most post hoc improvements of previously published data sets, one should remain very sceptical until proven in a) real world conditions (ie prospectively) , b) the accuracy of imaging is confirmed in other centres not associated with EMVision, c) the numbers get into statistical significance (ie are adequately powered) and most importantly d) the introduction of a program using the EMVision device actually improves patient outcomes in comparison to the current standard of care.

So far none of those have been achieved. How much does it cost to achieve the above? Certainly many multiples of what they currently have.

Not saying they won't get there, but they have a long way to go.

During the time they are executing their proposed development pathway, one can expect that there will concomitant improvements in CT perfusion scans using AI (the current Gold standard), biomarkers of stroke (a blood test pre-hospital to diagnose stroke with out imaging, EEGs pre-hospital, Vascular US pre-hospital etc etc.

Perhaps a deep-pocketed Health care giant like GE or Siemens will be convinced by the tech and take them over so they can afford to develop it properly, but expecting an unfunded minnow to up-end the standard of care of acute ischaemic stroke is not a thesis I would sink my funds into.

Sorry to sound so gloomy.

Good luck to holders.

Fair points @Chagsy and I don't doubt you're right with much of that.

My read is just that EMVision isn’t trying to replace the gold standard - it’s more about being a triage and monitoring tool for first responders and bedside use, especially where CT isn’t immediately available.

That is, they’re aiming to augment, not upend, the standard of care. If it can help prioritise care or reduce unnecessary transfers, that’s a real value add. But agreed..it’ll still need validation in real-world settings, across independent sites, and with clear outcome benefits to prove its worth.