[SubiculumCode]

I do a lot of MRI analysis including segmentation of small structures in the hippocampus called the hippocampal subfields. To collect these segmentations, we collect partial-field-of-view high in-plane resolution T2-weighted images on a 3 or 7 Tesla magnet. These sequences are generally only included in research protocols if the research specifically cares about hippocampal subfields...therefore they are rarely collected. There have been attempts to enable segmentation of these small structures using lower resolution T1-weighted scans, leveraging deep-learning or other models trained on concurrent T2w high resolution scans and the lower resolution scans within the same subject, allowing the model to predict the higher resolution information from the lower resolution inputs. This produces spectacularly beautiful segmentation on shitty data. Data whose resolution is about the same as the thickness of the structures you are segmenting or less. The problem is this: 1. The lower resolution image barely has any information in it on these smaller structures 2. The accuracy of the resulting segmentation depends entirely on how much the person fits the training distribution. But much research is on specific populations: children, autism, etc. 3. Some big names in imaging analysis tools have published these tools, lending their credibility to them. 4. The beautiful segmentations and (3) tend to convince non hippocampal experts that the resulting data is trustworthy, especially to an eager beaver researcher trying to maximize the impact of their already collected datasets.

I've rejected a number of papers for this.

But my point is this. Midjourney Medical might train a model to produce pretty images with this technique, but the more they need to depend on deep-learning models to get usable data, the more that the match between the training distribution and patient will matter.

[jmhmd]

This is a critical point. I am curious what the team building this looks like? Do they have ultrasound physicists and clinical practitioners in addition to the AI researchers?

[hn_throwaway_99]

> I am curious what the team building this looks like? Do they have ultrasound physicists and clinical practitioners in addition to the AI researchers?

Related to that point, I was so turned off by Midjourney's post about this because it absolutely reeked of all the red flags I've seen over the past 2 decades when Silicon Valley-types try to "disrupt" the medical field. The post was titled "Midjourney Medical", but it approached everything from a "wellness" perspective, i.e. how can individuals "optimize" their health by numerous scans/tests/analyses, it's a spa not a clinic, etc. etc.

Where were the research papers? The peer-reviewed articles? If you're truly trying to "disrupt" some area of medicine, it doesn't make you immune to the peer-review process.

I'll fully admit I'm not qualified to assess the claims made in the article, but if Midjourney wants to be taken seriously, I would hope they would avoid an approach that looks a lot more like Theranos in their early days, vs a sober analysis of how their tech compares to existing standards of care.

[hannasanarion]

> The post was titled "Midjourney Medical", but it approached everything from a "wellness" perspective, i.e. how can individuals "optimize" their health by numerous scans/tests/analyses, it's a spa not a clinic, etc. etc.

Could this not be a regulatory thing? I am an idiot who knows nothing, but if I had a new device that I thought had medical potential, and I wanted to get some press, investment, maybe a little hype going before I was ready to seek regulatory approval, this is exactly what I would do, market it as a wellness device first, because if I say that my prototype is installed in a "clinic" i might start getting suspicious looks from the FDA that can make life harder when I do eventually apply for medical device approval.