Soon, no more blood tests or probing for prostate cancer? AI claims 99% success rate using more relaxing methods Scientists say they have devised a way to screen for prostate cancer using a drop of urine, a sensor, and AI algorithms. And the test takes just twenty minutes, and is 99 per cent accurate, according to results from a small-scale test. The risk of developing prostate cancer increases for men as they get older, and the over 50s …
Evidence please - nothing published that I can find (unless you mean preserved foods which is not based on salt content).
Prostate cancer remains the only cancer diagnosed by random biopsy
There is real concern over over-diagnosis and the associated morbidity and mortality.
It would be more helpful if there was data on grade(s) of disease detected - excluding those where the disease has no impact on life from interventions and unnecessary healthcare activity and societal impacts would be good
I note that nobody mentioned the massive improvement in accurate diagnosis pioneered by Prof Mark Emberton, Hashim Ahmed and their teams at UCLH using non-invasive MRI scans. As someone who took part in their MASTER trial, I am very pleased that their excellent work resulted in the NHS finally generally adopting this method in preference to the "random biopsy" method previously used. This also allows targeted biopsies using a less infectious-liable method to determine severity.
As the AC asking for evidence, thank you for the down vote - as probably the only person on this forum who does this as their day job I look forward to the explanation
I have performed and reported thousands of MRIs and Prostate biopsies and am fully aware of the current state of play.
MRI location of cancers helps but there are few image based fusion systems in the UK to do this "cognitive fusion" being relied on.
You *cannot* see the "target" on US - if we could we wouldn't need MR!
Of course you are all aware of Pi-RADS 2.1 and the current debate on whether it is safe to not biopsy those with a score of 2 or less - despite massive inter-observer disagreement on scores
and I can evidence my claim :https://euoncology.europeanurology.com/article/S2588-9311(19)30033-1/fulltext re random
Surprised a nomogram hasn't been developed which would be more useful than neural net/AI
You appear to be making assumptions there.
That I have a prostate.
That I do not already have prostate cancer.
(And that I am lying face-down on the floor?)
Further, in my quick look, I could see association between salted meats and prostate cancer but neither a proved causal link nor confirmation that it is sodium chloride and not other salts (or just the sodium, or the chloride) that is being pointed out.
Testing for prostate cancer has been a real problem with lots of false positives and biopsies causing more problems than they would have solved. Because little was known about PCA base levels, biopsies were often performed when levels were high for a population but not for an indiviual – rinse and repeat for other measures such as blood pressure. I'm part of a long-term EU study that will provide data on PCA base levels and development over time that will hopefully provide better data for future inference testing.
This post has been deleted by its author
Benign prostatic hypertrophy is not however cancer. Telling the difference is the problem. The prostate enlarges naturally with advancing age but not everyone gets prostate cancer.
The more linked biomarkers one tests the higher the chances of correct detection, if they are linked in the way we understand. Using AI is a good idea, provided clinicians don’t rely on it absolutely. I’m a bit dubious about the indirect measurement method. The body uses several but that means it can be fooled.
CO2 levels in the blood for eg are monitored via pH by the body, but both metabolic acidosis and alkalosis are possible. The former can make you unwontedly breathless the latter tired and prone to blue lips.
I've always said it was healthy…
I found the copy of Rustler in the park and I'm just trying to return it!*
* Actually, I do remember coming across the odd wank mag in the park to and from school. Never understood why anyone would take them to the park except perhaps to perpetuate cliches!
I know, right! Porn for my generation was the underwear section in the home shopping catalogues. Then we started to find slimy porn mags in the woods. In my era the American women were already shaven, but we assumed that was air-brushing to hide rude pubes.
At my high school only on girl shaved - how do I know that? Gossip. It's what we used to use before social media.
and, on the sidelines, there's this innovative, ground-breaking, disruptive and truly revolutionary start-up called PISSTAKUM, lead by a gender-neutral war-orphan prodigy that promises 1000% return on your socially-aware, blockchain-linked, environmentally-balanced and life-changing investment, and it's ABSOLUTELY genuine and Mr Besos and Mr Cook and Mr/Ms Google give it a big thumbs-up so get there QUICK!!!!!
There's something weird:
"Lee said that although the false positive and negative rates of the algorithms were low at 0.024 and 0.037 respectively, the team needed to verify their results with many more patients."
With 23 people tested, why wouldn't all of your error rates be multiples of .043 (1/23)?
"Perhaps each of the four biomarkers had its own false positive rate and false negative rate..."
However I look at it, I can't see how anything that has false result rates of 2.4% and 3.7% can possibly give you a better than 99% correct diagnosis rate.
There are lies, damned lies, and statistics.
The quantity of urine needed and the purported efficacy of the test with the low numbers of results should lead to maximum caution.
The tone of this article smacks of shades of Theranos.
Those that ignore history have an unfortunate trait of repeating the mistakes of said history.
This is as ever a statistical model.
p(a|b) = p(b|a)p(a)/p(b)
given one finding, what is the likelihood of the other - I hear hoofbeats; horse or zebra?
ps I live in central Africa...
It is a prediction of likely significant disease - or most importantly, identifying a high risk population that require further testing while allowing large numbers to be eliminated from inappropriate screening intervention.
Remembeer [sic] when it comes to statistics, Bayesian's know where the bar is...
And it is how all medical practice works.
(AC for evidence again)
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
