What I Learned Using AI When My Brother Got Sick

In 2025 my brother had a seizure.

Everything after that happened fast. The doctors ordered an MRI and a PET scan. The scans found nodules in his lungs and two masses in his brain, and the word everyone used for the spots in his brain was “tumors.” Soon after, the doctors told us what they believed it was: stage 4 lung cancer that had spread to the brain.

I don’t have a good way to describe hearing that. You go numb for a while, and then your mind starts racing.

What I did with the paperwork

I make software for a living. When I’m scared, I look for something I can work on. So I gathered everything I could get: his medical records, his blood work, the reports from his scans. Then I started going through all of it with AI.

I wasn’t asking AI to diagnose him. I wanted to understand what I was reading. Radiology reports are written for other doctors, not for worried brothers. A blood panel is a page of numbers with little H’s and L’s next to some of them. I wanted to know what each finding meant, what it pointed toward and what it pointed away from.

I also wanted to understand the odds. Given these results, how likely is cancer? What else can cause nodules in the lungs and lesions in the brain at the same time? What would make one explanation more likely than another?

At the same time I started talking to family. I asked who in our family had had cancer, and what else ran in the family that I might not know about.

A word I’d never heard

As I went through the results with AI, one possibility other than cancer kept coming up: sarcoidosis. More specifically, a form of it called neurosarcoidosis, where the disease shows up in the brain and nervous system.

I had never heard of it. Then, talking with family, I found out that sarcoidosis runs in ours.

That stopped me cold.

So what is sarcoidosis?

Here’s the very simple version, as I understand it. I’m not a doctor.

Your immune system is supposed to attack things that don’t belong in your body. In sarcoidosis, it overreacts. Immune cells bunch together into tiny clumps called granulomas. Those clumps can form in almost any organ. The lungs and the lymph nodes in the chest are the most common places, but they can also show up in the skin, the eyes, the heart and sometimes the brain and nerves. Nobody knows exactly what sets it off, but it does seem to run in families.

Here’s the part that matters for my story. On a scan, those clumps can look a lot like cancer. A PET scan lights up places where cells are very active, and both tumors and inflamed granulomas are very active. So spots in the lungs plus spots in the brain can look exactly like cancer that has spread, even when it’s something else.

Why it’s so hard to catch

There’s no single test that says “this is sarcoidosis.”

It’s diagnosed mostly by process of elimination. Doctors rule out cancer. They rule out infections like tuberculosis and fungal infections. They rule out other conditions that cause the same kind of inflammation. Usually they take a biopsy, a small piece of tissue, and look at it under a microscope for granulomas. Even that isn’t proof on its own, because other diseases make granulomas too. Sarcoidosis is what’s left when everything else has been ruled out and the whole picture fits.

The symptoms don’t help either. Some people feel tired and short of breath. Some get a rash or eye problems. Some have no symptoms at all, and it turns up by accident on a chest X-ray. When it affects the brain, it can look like a stroke, multiple sclerosis, an infection or a tumor. And it’s rare enough that a lot of doctors don’t see many cases.

Learning, not diagnosing

AI can be wrong. It can sound very sure of itself when it shouldn’t be. I knew that going in, so I was careful about how I asked. Instead of “what does he have,” I asked things like:

How likely is each possibility, and why?
What in these results supports that, and what goes against it?
What should I read to understand this myself?
What questions should I bring to his doctors?

The reasoning ended up mattering more than the numbers. When AI showed me why it thought something, I could go check that reasoning against real medical research. I did a lot of reading. I learned more than I ever expected to about lung cancer, about how cancer spreads to the brain, about how sarcoidosis works and how doctors tell the two apart.

“Impossible”

The moment that has stuck with me most was a conversation with his neurosurgeon.

I asked, as carefully as I could, whether what my brother had could be neurosarcoidosis. The answer was one word: “Impossible.” The reason given was where the granulomas were in his brain. As the surgeon saw it, sarcoidosis just doesn’t show up there.

I didn’t argue. This is someone who operates on brains for a living. But it bothered me, because what I had been reading said otherwise. Neurosarcoidosis most often affects the lining around the brain, the nerves of the face and eyes, and the area around the pituitary gland. Masses inside the brain itself are less common, but they’re well documented in the medical literature. Less common is not impossible. Depending on the study, somewhere around 5 to 15 percent of people with sarcoidosis have it reach their nervous system, and autopsy studies find it even more often than that, which suggests it gets missed.

It took multiple biopsies to get a final answer. It was neurosarcoidosis.

My brother is doing well. After he started treatment, the spots in his brain that were first called tumors shrank by about 75 percent. A follow-up MRI will show whether they’re still there.

So “impossible” turned out to be wrong. I don’t say that to score a point, and I don’t fault the surgeon for it. Which brings me to the real point of this post.

Doctors have an impossible job

The amount of medical research published every year is staggering. New studies, new case reports, updated guidelines, in every specialty, all the time. Nobody can read all of it. Nobody can remember all of it. A neurosurgeon spends years becoming extraordinary at one very hard, very specific thing, and is then also expected to keep up with every rare disease that might turn up on an MRI. Add a full schedule of patients and surgeries, and a life outside the hospital. That isn’t a fair ask of anyone.

I don’t think the answer is for patients to swap their doctors for a chatbot. Please don’t do that. Everything I learned, I took back to the people treating my brother, as questions.

What I do see is a future where AI does the grunt work. It reads every record, every lab result and every scan report. It checks all of it against the latest research and the family history, and lays out the possibilities with the evidence for and against each one. Then a doctor reviews it, brings their experience and judgment, and makes the call. The doctor stays in charge. They just don’t have to remember everything anymore.

I did a clumsy version of that by hand, as a scared brother with a laptop. If one family member with no medical training could find something worth asking about, imagine what it would mean for every doctor to have that kind of help on every case.

If you’re going through something similar

A few things I’d pass along:

Ask for copies of everything: reports, labs, imaging results. You’re allowed to have them.
Talk to your family about their health history, including the odd things nobody mentions. Write it down.
If it’s someone else’s medical information, like my brother’s was, always ask their permission before you share it with AI. It’s their information, not yours.
Never freely hand sensitive information to an AI system. Before I uploaded anything, I removed my brother’s name, his date of birth and anything else that could identify him. AI doesn’t need any of that to explain a blood test.
If you use AI, ask for its reasoning and its sources, not just an answer. Then check them.
Take what you learn to your doctors as questions, not conclusions.
Be kind to the doctors. They’re carrying more than you can see.

Thanks for reading.