Slashdot Mirror
Google AI Claims 99 Percent Accuracy In Metastatic Breast Cancer Detection
Researchers at the Naval Medical Center San Diego and Google AI, a division within Google dedicated to artificial intelligence research, are using cancer-detecting algorithms to detect metastatic tumors by autonomously evaluating lymph node biopsies. VentureBeat reports: Their AI system -- dubbed Lymph Node Assistant, or LYNA -- is described in a paper titled "Artificial Intelligence-Based Breast Cancer Nodal Metastasis Detection," published in The American Journal of Surgical Pathology. In tests, it achieved an area under the receiver operating characteristic (AUC) -- a measure of detection accuracy -- of 99 percent. That's superior to human pathologists, who according to one recent assessment miss small metastases on individual slides as much as 62 percent of the time when under time constraints. LYNA is based on Inception-v3, an open source image recognition deep learning model that's been shown to achieve greater than 78.1 percent accuracy on Stanford's ImageNet dataset. As the researchers explained, it takes as input a 299-pixel image (Inception-v3's default input size), outlines tumors at the pixel level, and, in the course of training, extracts labels -- i.e., predictions -- of the tissue patch ("benign" or "tumor") and adjusts the model's algorithmic weights to reduce error.
In tests, LYNA achieved 99.3 percent slide-level accuracy. When the model's sensitivity threshold was adjusted to detect all tumors on every slide, it exhibited 69 percent sensitivity, accurately identifying all 40 metastases in the evaluation dataset without any false positives. Moreover, it was unaffected by artifacts in the slides such as air bubbles, poor processing, hemorrhage, and overstaining. LYNA wasn't perfect -- it occasionally misidentified giant cells, germinal cancers, and bone marrow-derived white blood cells known as histiocytes -- but managed to perform better than a practicing pathologist tasked with evaluating the same slides. And in a second paper published by Google AI and Verily, Google parent company Alphabet's life sciences subsidiary, the model halved the amount of time it took for a six-person team of board-certified pathologists to detect metastases in lymph nodes.
In tests, LYNA achieved 99.3 percent slide-level accuracy. When the model's sensitivity threshold was adjusted to detect all tumors on every slide, it exhibited 69 percent sensitivity, accurately identifying all 40 metastases in the evaluation dataset without any false positives. Moreover, it was unaffected by artifacts in the slides such as air bubbles, poor processing, hemorrhage, and overstaining. LYNA wasn't perfect -- it occasionally misidentified giant cells, germinal cancers, and bone marrow-derived white blood cells known as histiocytes -- but managed to perform better than a practicing pathologist tasked with evaluating the same slides. And in a second paper published by Google AI and Verily, Google parent company Alphabet's life sciences subsidiary, the model halved the amount of time it took for a six-person team of board-certified pathologists to detect metastases in lymph nodes.
That's somewhere between 17x17 and 17x18, which makes it even more impressive.
Distrubingly so.
Because they are breasts? Admittedly there is a greater publicity potential with breasts than colons, and having to explain to junior what a colon is and how to test for colon cancer at the dinner table would be awkward, whereas he'd get the idea readily about breast cancer since it's like being on the playground.
This will allow Google to properly target ads to breast cancer patients.
I just tell everyone I meet they have cancer. I haven't missed someone with cancer yet.
This is no different from the Waymo claims. If it is contextually irrelevant it means precisely jack shit, and human physiology is far more individuated than roads are. They are just trying to placate investors, and it's painfully obvious. Will Google be the next Soylent or Theranos? It seems incredibly likely with each passing year.
Women are cancer so maybe thats why God puts it in their boobs
It is simply amazing. The operator hits the "yes" or "no" key and voila. Analysis complete. As artificial an intelligence as a wife-beating Sherrod Brown Democrap.
... a static I didn't like.
It little behooves the best of us to comment on the rest of us.
If Google is starting to feel everyone up "looking for cancer" I'd say it's more than time to go use DuckDuckGo!
"There is more worth loving than we have strength to love." - Brian Jay Stanley
Classifier NN are particularly very good at this task, classifying images among predetermined image set. We already know that the given images are images of cells, possibly affected by cancer. The network that was trained specifically to classify such images will do very well because it will pick up the most subtle clues that are often missed by humans. It can pick up a single digit difference in pixel brightness level, which human can't generally do.
The same network however can't tell that the given picture is something other than cancer, like an elephant or a couch. It will just say - x% probability of cancer, which would be totally wrong.
I expect that we would need so much fewer trained radiologists, pathologists and in general people who look at pictures and tell us things. We can train inception like networks to do this with much more accuracy and less mistakes. So all these radiologist jobs that were outsourced to India and other places soon will be done by AI.
1 in 99 is really bad
1000 women, about 120 will get breast cancer, if we miss-diagnose 10 cases, that could be as bad as 8% failure
fuck statistics
Go well
The end result of this line of research will clearly be really really accurate inappropriate touching by robots...
G-tards
wont stand up in the real word, google is not capable, 'hint' hint if you get MI drift a 'ton' of mistakes in their thinking.
They have corporate hand cuffs and poor collaboration skills, if you share MI point, they cant even write papers that don't loose the plot rather quickly and you can clearly see when the authors are changing thru paragraphs when reading multi author papers .
Are not 'good' enough OR 'fellows' of the correct academic branches OR collaborate correctly OR know the 'difference' and
All their so called "training" is NOT able to stand up in the real world and their crap Chinese atheist ethics don't have any intelligence OR know the 'difference' take 'knaw-pig' for example a 'temporal' turd
Only good for games and toys . they are not JEDI material
also spatial challenged and full of poop
these guys will now get a pack of suppository when they go in their smelly gents, they have little helpers to help insert then they SQu33L