haldujai

65 ms·

661.
by haldujai 4y ago
In my opinion (physician/radiologist) we're going to see higher mortality and health care costs for a while as we are just starting to feel the effects of covid/prolonged lockdowns on missed screening. As one example, at my l
662.
by haldujai 4y ago
Sorry I should have explained to the acronyms I was using, bad habit. Yeah exactly. Apologies I use them inconsistently and interchangeably, Epic’s electronic medical record system changed it to ED for some reason.
663.
by haldujai 4y ago
I got called once at 4am to perform an urgent CT scan on a patient with “borborygmus” as a junior resident. At the time I was unfamiliar with this term (which means a rumbling or gurgling sound in the belly) and did not for a moment conside
664.
by haldujai 4y ago
Point-of-care ultrasound in the Western world is a hot trend. There’s the idea that clinicians can use this tool to aid in diagnosis as an extension of the physical exam (e.g. I’m worried this patient might have a kidney stone, let me slap
665.
by haldujai 4y ago
Yeah that’s insanely upcharged. Being generous a high end ER CT scanner is in the $200,000 to $300,000 range. Contrast is a few bucks. Technologist + porter are generously paid $100 combined + whatever you want to give for overhead, noting
666.
by haldujai 4y ago
I’m aware of what SOTA means, but how does that apply to a review? How do we or the authors assess that it is in fact the state of the art (best) without an objective measure? If they mean it’s the most current, every review is state of the
667.
by haldujai 4y ago
Abstractly I wouldn’t mind that, but it would take massive system level changes to make this feasible. We used to do ICU rounds are my institution but we’re struggling to keep up with the ever increasing volume of cross sectional imaging st
668.
by haldujai 4y ago
I’m SHOCKED to read a staff physician, especially as young as you are, is performing egophany and percussion. If your comments didn’t so strongly suggest that you are a very competent and diligent physician I would honestly call BS. Good fo
669.
by haldujai 4y ago
Do you know what the facility/hospital fee component is out of curiosity?
670.
by haldujai 4y ago
Good to know, I don’t have personal experience with POCUS devices (radiologist). I think traditionally Butterfly was much cheaper than Clarius, and had better software assistance / AI for things like TGC sliders and gain. Looking at th
671.
by haldujai 4y ago
Right, people do it but it has zero clinical utility as the history is way more significant (are you passing gas/stool). If you had an ileus (non moving bowel) for 2 weeks you would know without auscultating. If you’re passing gas or h
672.
by haldujai 4y ago
I can’t speak to US billing practices on the facility fee side, in general though Canadian fee codes for the professional fee/reporting component are higher than the US. The tendency to proceed to CT early is largely because: 1. A lot
673.
by haldujai 4y ago
Haha I was being a bit conservative, but my centre has UC in the tertiary care hospitals and I don’t see many asthma indications so I assume that population is not getting imaged routinely.
674.
by haldujai 4y ago
I’m impressed by your comfort level and curious about the specific indications you’re doing, would appreciate your perspective as I’m in a tertiary care/level 1 centre so POCUS practitioners aren’t as skilled here imo. The training in
675.
by haldujai 4y ago
Butterfly iQ is another one worth checking out and popular with my ER colleagues.
676.
by haldujai 4y ago
Bowel auscultation has zero clinical utility.
677.
by haldujai 4y ago
I personally hate reporting NICU brains (not a pediatric radiologist), even for sonographers it’s usually just a subset who are qualified to do them. Slightly poor technique can make a huge difference in vascular assessment and if your sett
678.
by haldujai 4y ago
If you need imaging for pneumonia / consolidation it’s definitely better to X-ray (very accessible), you may get lucky and see a peripheral pneumonia if it’s dense enough but the negative predictive value isn’t there as you have very l
679.
by haldujai 4y ago
Thats a very generous view on the current state of POCUS. I don't believe this has been studied yet but the % of patients who undergo POCUS in the ED (which is billed for) and leave without formal imaging in the radiology department (a
680.
by haldujai 4y ago
Ultrasound can't penetrate lungs (bags of gas) so probably not, you can see things like effusions, pneumothoraces, and maybe a peripheral consolidation. Regardless, the stethoscope has (for better or worse) already been replaced for a
681.
by haldujai 4y ago
"love is all you need" is the Beatles lyric they're referencing, the title could also be read as a statement on the basic human need for attention, like love, so it's a play on words for me. Do you prefer humorous allusi
682.
by haldujai 4y ago
> The size of the training data far exceeds the size of the model's weights. I'm not expecting strict recall but I imagine there is a non-insignificant amount of text in the training data about cysts relative to other entities
683.
by haldujai 4y ago
Yeah that's correct, but sometimes it seems what we are calling hallucination is recall of bad training data, Sebastian Raschka had a short post about this[1]. I was asking it a medical question about the imaging criteria for character
684.
by haldujai 4y ago
I was hoping this would link me to a deeper discussion on hallucination. I'm intrigued that it's hallucinating sequences that appear to have never written before (at least not on Google) and not just recalling some crappy training
685.
by haldujai 4y ago
Thanks for sharing, I'll give it a read. Perhaps I wasn't clear. I fully believe in transformers and foundation models, my criticism is more on creeping model size and whether trying to use huge models is even the right approach f
686.
by haldujai 4y ago
Fair enough. I guess I'm biased by my working environment and current belief that we're scaling transformer models unnecessarily, but I guess that is also partly influenced by their cost.
687.
by haldujai 4y ago
Maybe I'm missing the point of the article then. What's the low-resource scenario where inference speed is the bottleneck for transformer adoption at scale?
688.
by haldujai 4y ago
Sure, admittedly I was being overly hyperbolic and a bit snarky. However I am genuinely curious what sort of industrial "real world task" there is that requires edge inference on GPT3.5 or PaLM-sized models where you would run int
689.
by haldujai 4y ago
Yes, aware of the reference and hence the cringe. 'Attention is all you need' is a clever play on words for a seminal paper whose title directly related to its main scientific contribution. Not sure that every other AI-related pap
690.
by haldujai 4y ago
Or you could keep it simple and just not use a 500B parameter model which is unnecessarily large for 99.9999999999999% of use cases.

More ›