This might be a good new tool for PCa patients and their doctors making tough treatment decisions. It may be more useful for patients at early versus advanced stages but still pretty interesting in general.
targetedonc.com/view/nccn-r...
Here's the PowerPoint presentation from the creators. Information related to this new test is later in the presentation.
Big mistake, IMO. I can't imagine a patient who wouldn't prefer to be cured.
Can you elaborate on why you consider it a big mistake? This appears to me to be taking something similar to say DECIPHER to a new level. My understanding was tests like these were more often used to actually go with a more aggressive treatment early on than a less aggressive treatment which would increase not decrease the chance of cure. But obviously it is a 2-way street so a low score could avoid possible overtreatment.
They built their AI model using met-free-survival as an endpoint. This is why it calls for PSAs to climb twice as high before using ADT compared to using bRFS as the endpoint.
I think the point was to show the ability of an AI model to predict survival. The fact that they used an endpoint that it not what you consider to be ideal is irrelevant. The study could have been done with BRFS as opposed to met free survival and would have likely produced similar results. What do you have against the use of AI? It is the future. It can provide guidance in targeting clinical trials.
It is the wrong endpoint for men who are going into salvage therapy. I have nothing against AI, but it is only as good as the parameters that specify it.
I don't disagree about the appropriate endpoint because I have not studied this. I did not read the study carefully enough to understand the details. I am just saying that AI has great potential for helping to see what to study.
Actually, the criticism usually leveled against AI is that it does not "see what to study." It arrives at somewhat mysterious conclusions because it may find small aberrations in large datasets. This can lead to a problem in statistics called "overspecification." It leads one to question whether the same AI model will be replicable on new patients.
An AI that is not able to generalize (within its scope of course, not talking about AGI) would not even be taken into consideration...at least I hope so! The first thing you do when preparing a dataset is to avoid duplicates because they could influence the generalization ability of a neural network (ok I am using an example I am very familiar with)...but I need to take a good look at those links, it feels like 1000 years ago when I was studying the subject daily 
AI is only a tool and it is as good as the person instructing it. I am now using AI for two tasks. One is converting audio piano solo music to MIDI files which can be played on my huge player piano. I did not write the model and there are serious flaws in spite of it doing a very impressive job. I took me 9 months (I could not wait for someone to produce a better model since I might not live very long) working with a friend to write a program that corrects systematic errors and makes the music from quite good to remarkable depending on the source.
My other use is to help me with programming. I have a horrible memory for syntax and am not great at the more complex nested loops and complex data structures. Chat GPT helps a lot with the grungework but what I get back is only as good as how carefully I ask the question. Some time I need to go back and forth several times and at other times I realize that it is hopeless. Chat is just never going to understand the question to the extent that I can specify it. This is no different than a collaboration with another programmer who has a different set of skills. I doubt that I could have written the "improve" program or that I would have the patience to do it at my own speed. I now use AI for all my programming but have gotten much better at guessing what it can and can not do. I would say that it has improved the efficiency of my programming by over 50%. I also think that AI will be improving tremendously as programmers provide feedback as to the weak points. I am utterly amazed at what I have been able to do with it as an end user.
Do not let AI write your legal briefs and submit them without checking 
technically, it instructs itself...but you need persons to prepare data (at least for time being)
but you need to ask the right question and that can be tricky
Different story, you are referring to something like how to use chat gpt properly, I am referring to how to train it (not precisely chat gpt as it’s based on transformers if I am not wrong while I often refer to neural networks), but yes it’s amazing and wait to see quantum computers applied to it…
Ai = is a girl's name of Japanese origin. Meaning "love" and "affection", this name works as a reminder for multiple important facets in life. Reminding baby to love herself, that she is deserving of love and affection, and reminding you to teach her these lessons each day couldn't be more important.
Good Luck, Good Health and Good Humor.
j-o-h-n
LuPSMA prognostic model
Using AI Against Research Papers.
Prognostic Nutritional Index (PNI)
