Salesforce Claims Its AI Can Spot Signs of Breast Cancer With 92% Accuracy (venturebeat.com) 24
Salesforce today peeled back the curtains on ReceptorNet, a machine learning system researchers at the company developed in partnership with clinicians at the University of Southern California's Lawrence J. Ellison Institute for Transformative Medicine of USC. From a report: The system, which can determine a critical biomarker for oncologists when deciding on the appropriate treatment for breast cancer patients, achieved 92% accuracy in a study published in the journal Nature Communications. Breast cancer affects more than 2 million women each year, with around one in eight women in the U.S. developing the disease over the course of their lifetime. In 2018 in the U.S. alone, there were also 2,550 new cases of breast cancer in men. And rates of breast cancer are increasing in nearly every region around the world.
In an effort to address this, Salesforce researchers developed an algorithm -- the aforementioned ReceptorNet -- that can predict hormone-receptor status from inexpensive and ubiquitous images of tissue. Typically, breast cancer cells extracted during a biopsy or surgery are tested to see if they contain proteins that act as estrogen or progesterone receptors. (When the hormones estrogen and progesterone attach to these receptors, they fuel the cancer growth.) But these types of biopsy images are less widely available and require a pathologist to review. In contrast to the immunohistochemistry process favored by clinicians, which requires a microscope and tends to be expensive and not readily available in parts of the world, ReceptorNet determines hormone receptor status via hematoxylin and eosin (H&E) staining, which takes into account the shape, size, and structure of cells. Salesforce researchers trained the system on several thousand H&E image slides from cancer patients in "dozens" of hospitals around the world.
In an effort to address this, Salesforce researchers developed an algorithm -- the aforementioned ReceptorNet -- that can predict hormone-receptor status from inexpensive and ubiquitous images of tissue. Typically, breast cancer cells extracted during a biopsy or surgery are tested to see if they contain proteins that act as estrogen or progesterone receptors. (When the hormones estrogen and progesterone attach to these receptors, they fuel the cancer growth.) But these types of biopsy images are less widely available and require a pathologist to review. In contrast to the immunohistochemistry process favored by clinicians, which requires a microscope and tends to be expensive and not readily available in parts of the world, ReceptorNet determines hormone receptor status via hematoxylin and eosin (H&E) staining, which takes into account the shape, size, and structure of cells. Salesforce researchers trained the system on several thousand H&E image slides from cancer patients in "dozens" of hospitals around the world.
CRS (Score:2)
Oh, great... (Score:2)
Oh, great... no more paid jobs looking at pictures of tits.
Re:Oh, great... (Score:5, Funny)
Re: (Score:2)
Become an ornithologist!
I'm cryin' here!
Re: Oh, great... (Score:2)
No, just tissue samples from a biopsy, which is taken from underneath the skin. Anyways it doesn't sound like this test will detect a either HER2 or triple negative cells, so even if it was 100% accurate, you still can't rule out cancer.
All About That Sale (Score:3)
Re: (Score:2)
I didn't realize breast cancer patients were such a valuable demographic. I guess marketing to pregnant women [forbes.com] just isn't as profitable as it used to be.
https://en.wikipedia.org/wiki/... [wikipedia.org]
Prevailance:
Worldwide, breast cancer is the leading type of cancer in women, accounting for 25% of all cases. In 2018 it resulted in 2 million new cases and 627,000 deaths. It is more common in developed countries and is more than 100 times more common in women than in men.
On early detection:
Prognostic factors
The stage of the breast cancer is the most important component of traditional classification methods of breast cancer, because it has a greater effect on the prognosis than the other considerations. Staging takes into consideration size, local involvement, lymph node status and whether metastatic disease is present. The higher the stage at diagnosis, the poorer the prognosis. The stage is raised by the invasiveness of disease to lymph nodes, chest wall, skin or beyond, and the aggressiveness of the cancer cells. The stage is lowered by the presence of cancer-free zones and close-to-normal cell behaviour (grading). Size is not a factor in staging unless the cancer is invasive. For example, Ductal Carcinoma In Situ (DCIS) involving the entire breast will still be stage zero and consequently an excellent prognosis with a 10-year disease free survival of about 98%.
Stage 1 cancers (and DCIS, LCIS) have an excellent prognosis and are generally treated with lumpectomy and sometimes radiation.
Stage 2 and 3 cancers with a progressively poorer prognosis and greater risk of recurrence are generally treated with surgery (lumpectomy or mastectomy with or without lymph node removal), chemotherapy (plus trastuzumab for HER2+ cancers) and sometimes radiation (particularly following large cancers, multiple positive nodes or lumpectomy).[medical citation needed]
Stage 4, metastatic cancer, (i.e. spread to distant sites) has a poor prognosis and is managed by various combination of all treatments from surgery, radiation, chemotherapy and targeted therapies. Ten-year survival rate is 5% without treatment and 10% with optimal treatment.
So can I (Score:4, Insightful)
Here's the algorithm:
boolean hasCancer(Image img) {
return false;
}
Since less than 8% of women in the world have breast cancer, it's going to be > 92% correct. Give both a false positive and a false negative rate or stats like that are garbage.
Re: (Score:3)
So, I know it's a sin, but I looked up the original Nature article [nature.com] and it seems it's not as bad as suggested - a) it's "achieving an area under the curve (AUC) of 0.92 for sensitivity and specificity" and b) it's designed to cooperate with a person so the overall diagnosis is likely to be even better.
Re: (Score:2)
I have you beat. My AI can detect 100% of cases:
10 PRINT "POSITIVE"
Accuracy =/= false negative (Score:3)
In any case, the method they report is not about detecting breast cancer, but about detecting a particular characteristic of breast cancer that typically requires addit
Re: (Score:2)
Unless you think 1-in-8 means 8%. Otherwise you are making up percentages to fit your algorithm.
Does it detect itself? (Score:2)
The question in the subject is almost obligatory, given how many people would diagnose Salesforce as a cancer.
Re: Does it detect itself? (Score:2)
Salesforce is cancer by itself. Or a very unfortunate name.
healthcare should not be an sale / have middle man (Score:2)
healthcare should not be an sale / have middle man driving costs up.
Ahh the old joke.... (Score:2)
One day Joe complained to his friend, âMy elbow really hurts. I guess I should go to the doctor.â(TM)
His friend advised âDonâ(TM)t do that. There is a computer at the drugstore that will diagnose anything quicker and cheaper than a doctor. Just put in a sample of your urine and the computer will diagnose your problem and tell you what to do about it. It only costs $10.â(TM)
Joe figured he had nothing to lose, so he filled a jar with urine and deposited the $10.
The computer started ma
I question the timing of this announcement. (Score:2)
The last thing I'd trust Spammerforce to do (Score:2)
is make any kind of health care decision.
Jackasses at spammerforce got into the spam for hire game a few years back, and started blasting my mail server with bullshit.
They also bought the Exact Target spam group, which has been a dirty spammer for hire group for many many years.
Firewall em and forget em.
Why??? (Score:1)
Why is Salesforce looking at my breasts?
Hold on a sec (Score:1)