Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

I used Claude Code to get a second opinion on my MRI

antoine.fi

Post reply on HN
Print view View original

691–700 of 748 posts

Re: I used Claude Code to get a second opinion on my MRI

#691

Post by kennywinker »

shiandow wrote: post #542
azan_ wrote: post #354

Earlier quoted context omitted.

How is it possible? You can't diagnose tuberculosis just based on imaging and tuberculosis hospital has to know that!

Not only that, what is the point confining someone to prevent the spread of a disease about a quarter of the world is already infected with? I suppose there could be reasons, but I don't know them.

Because it’s a nasty disease, and they’d like to prevent its spread. A quarter of the world may have TB, but there are only like 10,000 cases in the US every year

Re: I used Claude Code to get a second opinion on my MRI

#692

Post by agos »

lostlogin wrote: post #157
darepublic wrote: post #92

I would like if we could have a site where you submit your MRI then doctor commenters anonymously post their opinion. In general I want a forum where.. when people come with questions for which there are varying opinions we don't just have people leave their 2c and then jet. The thread persists, duplicated ideas get merged, erroneous statements get purged and gradually we refine shining truth

I’m wondering how many radiologist want to work all day, then come home and work. Many can get paid fee-for-service for after hours work, so would probably prefer that.

there's plenty of doctors doing volunteer work, this would be just another medium to do it

Re: I used Claude Code to get a second opinion on my MRI

#693

Post by greenail »

marcus_holmes wrote: post #481
Aurornis wrote: post #167

Earlier quoted context omitted.

I have multiple LLM subscriptions at any given time, plus an array of local models. When I ask a question outside of my domain of expertise I like to ask all of the LLMs I have access to. I also create separate sessions and ask the same question multiple ways. It’s revealing to see how many different and contradictory answers I get, most of which are presented confidently. The last time I ran a medical question throu…

In my day job we tried creating a credit assessor tool using LLM as the credit assessor. It did great, generated a report on the assessed business that was incredibly detailed and plausible. Then I started running tests and getting into the details, and found that if you ran the same report on the same data, it generated completely different, still very plausible, results. I could run the same source data through the…

you can set the "temperature" which is a lever on how stochastic the prediction is. If you are doing your own inference this is clear and easy. If you are consuming tokens this is outsourced.

Re: I used Claude Code to get a second opinion on my MRI

#694

Post by nums »

rockostrich wrote: post #672
AceJohnny2 wrote: post #117

> There's something incredibly peaceful about being in the hands of an expert you trust. [...] AI can absolutely shatter that feeling in an uncomfortable way [...] but I don't know if I can fully trust AI either. This really is key. We know we can't trust the AI, but at the same time we're also more comfortable asking the AI for clarifications or confronting it. Not having a time-bound appointment or paying by the ho…

There are 3 kinds of mechanics: Scammers who do the lowest effort diagnostic and "fix" to get you to pay a smaller amount of money to fix the problem in the short term even though it'll re-present itself a week/month/year later. Upsellers who will find other things "wrong" with your car and pressure you into paying to fix them because they sound a lot worse than they are. Good mechanics that will explain what they di…

Maybe a difference here is asking AI for conclusions. When I have it do a buyer's report for me, I ask it for "what questions should I be asking? What are typical things that go wrong with this type of vehicle?" I don't delegate conclusions to the AI but use it to educate myself. Then, I can gather further information to make MY decision .. to buy it or not.

Re: I used Claude Code to get a second opinion on my MRI

#695

Post by nums »

base698 wrote: post #604
Aurornis wrote: post #167

Earlier quoted context omitted.

I have multiple LLM subscriptions at any given time, plus an array of local models. When I ask a question outside of my domain of expertise I like to ask all of the LLMs I have access to. I also create separate sessions and ask the same question multiple ways. It’s revealing to see how many different and contradictory answers I get, most of which are presented confidently. The last time I ran a medical question throu…

My step mom was having debilitating pain. A year of going to doctors and no one was able to find a cause. I scanned her discharge paper work which had her prescriptions on it and gave it to Claude. It identified a prescription that had that exact side effect. They later confronted her primary care that concurred and took her off it. A friend of mine's wife recently passed. They were chasing a suspected heart defect f…

My mom had cancer and she was on regular, suppressive chemotherapy. I put her info into an AI and it correctly noted that her chemotherapy had stopped being effective 2 months prior based on factual lab reports. She was unaware of this. I was able to be her health advocate much more effectively by respectfully asking her oncologist targeted questions. He was already on top of it and was addressing the issue. Our conversation was respectful and, due to my educating myself, went up another level. Ultimately, it was a positive interaction. I was satisfied that he was indeed expert at his craft, and he was satisfied that we were aware of the uncertainty of the new treatment with a risk-based understanding of the viability of success. This was a positive engagement with an expert. In parallel situations around non-health issues, I've found the ego of the expert seems to be the determinative factor in whether or not the interaction goes well.

Re: I used Claude Code to get a second opinion on my MRI

#696

Post by nums »

Here's an example that illustrates a shortcoming about how medicine has historically been practiced and how it might change with the advent of AI.

A person very familiar with me was having an interaction at a review board level at Stanford. They had this rare illness that they were treating and someone had the "bright idea" of suggesting of saying ... "hey, why don't we look at the ten other times we treated this exact illness and see what worked!" Everyone was delighted with this novel idea (discussion circa 2021). My person was a bit disgusted as this is simple feedback loop style improvement and WTF! they should be doing this all the time to get probabilistic style suggestions for many treatments. I know it happens within certain healthcare systems (e.g., Kaiser had full EMR back in early 2000s and saw right away that VIOXX was killing people. So, they stop prescribing it. citation: Kaiser panel member paraphrase at a healthcare conference in 2010). If you just observe the healthcare system, you can see that the healthcare systems and most EMRs don't typically capture the feedback loop (i.e, when's the last time a doctor followed up and said "did you feel better after the last treatment?" or measures the result.) AI itself can't solve this as it doesn't have access to the data feedback loop. However, maybe AI's within the EMR will help "suggest" evidence based treatments. I could go on and on, but as a math guy, I've often been shocked at the non-evidence based assertions some doctors make. My conclusion is that if you're not "in the fairway", they are typically just guessing.

Re: I used Claude Code to get a second opinion on my MRI

#698

Post by rockostrich »

palata wrote: post #689
rockostrich wrote: post #672

Earlier quoted context omitted.

There are 3 kinds of mechanics: Scammers who do the lowest effort diagnostic and "fix" to get you to pay a smaller amount of money to fix the problem in the short term even though it'll re-present itself a week/month/year later. Upsellers who will find other things "wrong" with your car and pressure you into paying to fix them because they sound a lot worse than they are. Good mechanics that will explain what they di…

Yes, and that's a problem. Doctors (or experts in general) hate it when people don't trust them, but the thing with experts is that people have to trust them. And in my life (and a few times just in the last few years), enough doctors have been wrong enough that I cannot just trust them anymore [1]. If it is important, I will ask them to explain to me, and sometimes I will just ask for a second opinion. I have read a…

Heh, I hear stories like that everyday from my partner who is an ICU nurse. Not as dire, but there are constant inter-department arguments about moving patients because of resource constraints and the ICU could end up completely understaffed/resource constrained if the wrong NP or charge nurse is working. I'm amazed our healthcare system works at all to be honest.

Re: I used Claude Code to get a second opinion on my MRI

#699

Post by ShinyLeftPad »

coderatlarge wrote: post #673
bondarchuk wrote: post #641

Earlier quoted context omitted.

It's funny that if the LLMs had all given the same result each time (it sounds like) you would have considered it more valid, even though it might just be giving a single wrong answer more consistently.

surely getting the same answer multiple ways suggests it is nore likely in the probability space?

Slavery used to be super likely in the probability space, doesn't make it right...

Re: I used Claude Code to get a second opinion on my MRI

#700

Post by rockostrich »

nums wrote: post #694
rockostrich wrote: post #672

Earlier quoted context omitted.

There are 3 kinds of mechanics: Scammers who do the lowest effort diagnostic and "fix" to get you to pay a smaller amount of money to fix the problem in the short term even though it'll re-present itself a week/month/year later. Upsellers who will find other things "wrong" with your car and pressure you into paying to fix them because they sound a lot worse than they are. Good mechanics that will explain what they di…

Maybe a difference here is asking AI for conclusions. When I have it do a buyer's report for me, I ask it for "what questions should I be asking? What are typical things that go wrong with this type of vehicle?" I don't delegate conclusions to the AI but use it to educate myself. Then, I can gather further information to make MY decision .. to buy it or not.

I don't think so. LLMs tend to over-index on providing results in general whether it's a conclusion or not. When you ask it "What are typical things that go wrong with this type of vehicle?" you're forcing it to make a conclusion about which results to include and it will almost certainly provide results even if those issues aren't as much of a concern compared to typical issues with other vehicles.

For example, I just prompted Kimi-K2.6 with:

> I'm considering buying a used base model 2010 Honda Civic with 80k miles that's been garage kept. What are typical things that go wrong with this type of vehicle?

It listed 10 issues including the engine block cracking (which wasn't even an issue with 2010 Civics). Started a new chat and asked about a 2010 Toyota Camry, another unbelievably reliable car, and it listed 9 similar issues. Started a new chat and asked about a 2011 Jeep Grand Cherokee, a notoriously unreliable vehicle, and it listed the same number of issues.

Sure it's data to make decisions on either way, but it really all comes down to how good your prompts are and whether or not you can think critically about the output, whether or not that output is a conclusion or just data collection.

Post reply on HN