Technology

Our company is a joint venture partner in this

https://annalise.ai/products/annalise-cxr/

I was involved in some of the development, it’s going live in a few weeks. It’s a helpful tool at the moment but eventually it will take over the reading of basic imaging and probably at some stage, the high end stuff (hopefully after I retire!)

 

Is there any AI involved in the robots that perform prostate cancer removal ?

Mate of mine going in for operation on his prostate and I told him to be careful if robots are doing it

 

\
Greg is not a robot ,

 

Short answer, no. It’s all controlled by the surgeon.

Prostate cancer treatment is quite controversial. I’d be favouring radiotherapy over surgery if it were me. I see loads of post surgical prostate patients, incontinence and impotence can happen even with robotic techniques

 

Racist

 

I'm over sex, over-rated in my book .... cut away !!!

Haha I pissed myself , must be incontinent as well

 

.... and a bully

 

Where all racists in our own way .

No such thing as being an anti-racist .......sorry lefty flogs but that's the way it is .

 

Same thing happening with pigmented skin lesion diagnosis, but it's clunky, and still operator dependent. Once the image is acquired, it can be startlingly accurate, especially for equivocal lesions. There's been at least 20 years of tuning algorithms, but the most successful one simply fed in about 8,000 images with their diagnoses to the AI. I haven't used an algorithm for decades, except for teaching - I rely on heuristics.

 

Really, you and Greg?


Silly signature here to annoy KIAI

 

WTF has it got to do with Ferrari ..i ask?

 

AI has become a cliche - everyone claims the capability, few even know what it means. 90% of AI software is just pattern matching.

The correct term is Neural Network - software that works like the synapses in the human brain, making connections between data items that are not pre-determined by coded instructions.

 

True, but try explaining that to the great unwashed

 

Are you working on SkyNet? Give us a heads up before it goes active

 

Interesting that the focus is now on imaging; years ago AI was going to be the new diagnostic tool.

 

I was wondering why he changed the name of his company to Cyberdyne Systems

 

This is purely anectdotal, but quite a few of my old swimming mates have had prostate problems - those who went the surgical route all complained of some degree of problems after (worst case was nerve damage causing constant severe pain, wrecked their marriage...); the guys who went with radiotherapy seem to be happy with the outcomes.

 

We have a huge database of images now, about 1 million cases for fed through to come up with a reliable algorithm for Chest xrays

 

Can you give me an example of a program that uses true Neural Network? What sort of connections are being drawn? I’m always interested in learning more about this stuff.
I object to the term AI because I believe intelligence is what allows people to hypothesise - think Newton and the apple.

 

It's a real issue, surgical outcomes vary widely between the various Urologists, so if you decide on a surgical path then you need to pick your Urologist carefully

 

Curious to know if they’ve loaded it with ‘obsolete’ diseases such as TB etc?

 

Absolutely, but the interesting thing is the guys who went with the commonly pronounced ‘best’ around here had the worst outcomes... but, purely anectodal.

 

Everything, but all it does is show you findings and that's where the Radiologist comes in and puts it all together, making a diagnosis such as TB. It's value at the moment is just as a bit of an adjunct, helps check that things aren't overlooked

You may be surprised to hear that TB is still around, I see quite a few cases every year

 

The "best" are usually the best self-promoters

 

The people who write this stuff have PhD's in applied mathematics. It's not easily understood or explained.

The simplistic (and not entirely accurate) description is any computed outcome that a human wouldn't have predicted from the same set of inputs. A computer neural network can hold billions of data points in its "mind" at any point in time, the smartest human perhaps a few thousand. Think A might cause B might cause C extrapolated a billion times.

So in an engineering situation, our software can predict a failure before it happens, that a human wouldn't have seen, because of the combination of factors so remote from each other that a connection couldn't otherwise be made. I suspect this is what Greg's system is doing for cancer diagnosis.