I'll be listening in on tomorrow's conference to discover that too, but as I posted on another thread it makes a big difference if you can't get the patient back for a follow-up visit. Good hospitals aren't free in the USA like they are here. If you need to repeat a test or get another X-ray to confirm the initial diagnosis but they won't come back because it costs too much, you're down to the initial results and whatever you think it was most likely to be. I suspect that in Australia we had better support for the clinical adjudicators. Tomorrow will tell."Why no one is questioning the Australian studies but point a finger to US ones"
I suspect it will come down to the fact that we trained our algorithm with Australian adjudicators and US adjudication was different. Therefore the algorithm wasn't as good at predicting US diagnoses as it would have been at predicting Australian ones.
And on the face of it, the answer isn't going to be as simple as re-training with the current data and trying another trial during the winter after this one. It's pretty clear that for some conditions clinical diagnosis just isn't as predictable in the US as it is in Australia, for reasons referred to above. Remember we centralised adjudication for this trial for exactly this reason; to eliminate variability between teams. Now it's obvious that there is a lot of variability between adjudicators on the same team.
It comes down to this:
You can only train an algorithm to make accurate classifications if the classifier is predictable.
This is about my last post here. On the face of it ResappDx can no longer be regarded as a paradigm-shifting, revolutionary development. Instead it's starting to look like a long, hard, slow grind to bring this to market and transition to revenue. I suspect we are two years or more from where we thought we would be by Christmas this year.
Hang in there all of you - optimists and opportunists alike - I believe Resapp is going to need your support through the years and capital raisings to come.
Ann: Positive Preliminary Results from SMARTCOUGH-C-2 Study, page-289
Currently unlisted. Proposed listing date: 4 SEPTEMBER 2024 #