MIT Developing AI To Better Diagnose Cancer

stowie writes: Working with Massachusetts General Hospital, MIT has developed a computational model that aims to automatically suggest cancer diagnoses by learning from thousands of data points from past pathology reports. The core idea is a technique called Subgraph Augmented Non-negative Tensor Factorization (SANTF). In SANTF, data from 800-plus medical cases are organized as a 3D table where the dimensions correspond to the set of patients, the set of frequent subgraphs, and the collection of words appearing in and near each data element mentioned in the reports. This scheme clusters each of these dimensions simultaneously, using the relationships in each dimension to constrain those in the others. Researchers can then link test results to lymphoma subtypes.

A1 is good enough already

[turkeydance]

for steaks, chicken and french fries.

Re:

[Anonymous Coward]

If you put A1 on steak you need to be institutionalized in a kitchen until you can recognize decent food.

Comment removed

[account_deleted]

Comment removed based on user account deletion

FTW

[Hangtime]

I don't need an AI to tell me I have cancer that's what WebMd does already!

Such hyperbole in TFS

[fyngyrz]

MIT Developing AI To Better Diagnose Cancer

FFS, it's not AI. It's a mindless program. Unthinking software. Data analysis software. Innovative to some degree perhaps, but AI? Hardly. No better than me stumbling in here and calling some DSP code I'd written "AI." Well, except I wouldn't do that. :/

When AI gets here, we'll have to call it something else what with all this crying wolf going on.

Re:

[blue9steel]

Well, it's not strong AI that's for sure, it's most accurately categorized as an expert system.

Yes

[Njorthbiatr]

But how do we bribe -- err I mean buy lunch for -- it so it will suggest prescribing my company's drugs?

It's Not a Tumor!

[tlambert]

"... that aims to automatically suggest cancer diagnoses..." even if you don't actually have cancer.

Higher diagnoses

[oodaloop]

I think this may contribute to an already existing problem with the high cancer rate: high detection. How many people who are diagnosed with cancer would have lived just fine and ended up dying from something else? We do find cancer in autopsies where the person didn't know they had cancer. Is the current trend of high cancer rates partly due to better means of detection, and it's just that lots of people have had asymptomatic cancer all this time? Does every form of cancer require massive amounts of chemo? My wife passed away from stage 4 colon cancer last year. It had spread to her lungs, adrenal gland, and liver. She had surgery for the original tumor, and underwent 3 years of aggressive chemo to remove the very tiny filaments elsewhere in her body. I can only wonder if, without the chemo, she would have had the same fate. There are people who forego chemo and survive. And obviously chemo is necessary for many people to beat cancer. But I have to wonder if getting better at detecting cancer will bring more good than harm.

Re:Higher diagnoses

[rmdingler]

Elderly men with slow developing prostate cancer are frequently not treated because the disease is unlikely to kill them first.

Sadly, your medical care is incentivized in the same fashion as an automotive repair: the more repairs that are necessary, the greater the final invoice.

This is not to suggest there are not a great many ethical physicians, but we would be fools to overlook the likelihood that some sociopaths have slithered into the profession.

Re:Higher diagnoses

[ColdWetDog]

Probably not - at least in this case. They are looking at a specific form of cancer, lymphoma. Lymphomas do span the gamut from being indolent to extremely aggressive, hence the need for accurate diagnosis, but we have a fairly good idea of what the natural history of each subtype is. This system is not designed to mow through a bunch of clinical data and pop out a 'cancer' diagnosis.

That said, TFA is incredibly poorly written. It is anything but clear WHAT information they are using (pathology slides? DNA samples? Chart notes?) and it is most certainly not AI.

While over diagnosing pre clinical cancers is a concern, this particular methodology won't make that worse. In fact, if it actually does work, it might decrease what are essentially false positive diagnoses by linking the testing component to the natural history of the disease (eg, 'this particular cancer is mostly harmless, don't worry about it much').

Accuracy

[__aabppq7737]

by learning from thousands of data points from past pathology reports.

I'd be worried if my future surgeon had only 1000 bullet point takeaways from college, and no experience, I'd be a little worried.

What does the term 'data point' mean?

No Frenchie

[nospam007]

"...technique called Subgraph Augmented Non-negative Tensor Factorization (SANTF)"

Obviously nobody speaks french there or they would have used a last word beginning with 'e' so that it spells SANTE which means 'health' in french.

What are they trying to do again?

[chooks]

It is unclear where in the diagnostic chain this idea fits. Is it someone that already carries a diagnosis of lymphoma, but there is a question the diagnosis is wrong? Is it using lab data to make a primary diagnosis (or suggestion of diagnoses) based on a clinic visit? Are they suggesting that this data fits an ancillary role in primary diagnosis in terms of resolving subtle discrepancies between diagnoses?

Pretty much all hematopoietic malignancy diagnoses do not come from the docs you see in the clinic

Sorry Dave

[moeinvt]

"I'm sorry Dave, I'm afraid you have testicular cancer."

Better Acronym

[VorpalRodent]

The last word needs to be something like "Analysis". I'll admit, my goal here has nothing to do with helping people or better representing the technology. I just want people to be able to say "SANTA told me I have cancer."

AI?

[clenhart]

Sounds like data mining