I Encourage My Patients to Use AI to Understand Their MRI Report
Dr. Merritt encourages patients to use AI to understand their MRI report. Here is what the research shows and what AI and the report cannot tell you.
Frequently a patient tells me they put their MRI report into ChatGPT or Claude before seeing me in clinic. I am always happy to hear that. The reason I have a website and all of these patient support resources is to encourage patients to own their healthcare journey, and now AI is one more tool. I fully support patients using AI to understand their health. A report full of words like "intrasubstance signal" and "chondral fissuring" is hard for anyone to read, and a patient who understands it walks into the room with better questions.
It is interesting to see that there has actually been some research into this topic. A randomized study in Arthroscopy found that AI-translated MRI reports significantly improved patient comprehension1. They understood the report much better! I like that patients are getting involved in their care. The one thing I want to add is what an AI tool, and the report itself, cannot tell you, so you know how to get the best of both worlds.
What the research shows
The Arthroscopy study was a randomized trial. Sixty-four patients in an orthopedic clinic (average age about 52) read either a standard radiologist MRI report on a multiligament knee injury or a ChatGPT-4 plain-language translation of the same report. Then everyone took the same comprehension quiz1.
Patients who read the AI version scored 74.7% on the quiz, compared with 54.7% for the standard report. Almost 9 in 10 preferred the AI version, and they rated it clearer. But also, they trusted it less than the radiologist's report1.
The mix of better understanding but lower trust is the right way to think about AI. It is a good translator, not a doctor. In a separate study of 300 knee reports, AI rewrites cut the reading level from about 13th grade to 7th grade, but 2% to 5% of the explanations contained inaccurate or made-up content2.
The problem: a report is not a diagnosis
An MRI report is a radiologist's visual description of the images. The radiologist does not examine you and often does not know where it hurts. Their job is to describe every finding they see, so the report lists everything, whether or not it explains your pain.
A common example: the MRI shows a meniscus tear, but the symptoms are coming from the kneecap joint (the patellofemoral compartment) or something else in the knee. The meniscus tear is real, but it is not relevant to what is bothering you. If you only read the report, you could easily worry about the wrong problem.
| Radiologist's report | What I add in clinic |
|---|---|
| Describes every finding on the images | Matches your symptoms to the findings that matter |
| Based on the images alone | Based on your exam, history, and the images |
| Often does not have your X-rays | Reviews weight-bearing X-rays, which often show arthritis better than MRI |
| May miss or underrepresent a finding | I look at the images myself, not just the report |
Why I look at the images myself
I always read the report before I see a patient, and I always look at the MRI images myself. Radiologists are very good, but sometimes a finding is missed or understated, and sometimes I find a problem that does not show up in the report at all.
Your X-rays matter too. The radiologist reading your MRI often does not have them. A standing X-ray shows how the knee looks under your body weight, and it often shows arthritis more clearly than an MRI does.
Why I want you to use AI
I want you to use it. Use AI to understand the words, to make a list of questions, and to learn the basics about your findings and your options. A patient who has done that homework asks better questions, and our visit is more fruitful. Taking a proactive step to learn about your problem is exactly what I hope to see.
Then bring the report and your questions to a visit where I can examine your knee and look at the pictures with you. The AI helps you understand the report, and the visit is where we figure out what it means for you. Just do not use an AI summary on its own to decide that you do or do not need surgery.
Frequently asked questions
Is it safe to use ChatGPT to read my MRI report?
It is a reasonable way to understand the terms and prepare questions. Do not rely on it for diagnosis or treatment decisions, and do not paste identifying details like your name or date of birth if you want to protect your privacy.
My report says meniscus tear. Do I need surgery?
Not necessarily. Many meniscus tears are not the source of pain, especially with arthritis or kneecap problems. Your symptoms and exam decide this, not the report alone.
Why do you want my X-rays if I already had an MRI?
Standing X-rays show how the joint carries your weight and often show arthritis better than MRI. The two tests answer different questions.
Can the report miss something?
Yes. Radiologists are good, but a finding can be missed or described lightly. That is why I look at the images myself.
If you have an MRI report you are trying to make sense of, go ahead and run it through AI, then bring it and your questions to clinic. We will go through it together and look at the images side by side.
References
- Berzolla E, Gosnell GG, Chen L, et al. Artificial intelligence large language models improve patient comprehension of radiologist magnetic resonance imaging reports. Arthroscopy. 2025. doi:10.1016/j.arthro.2025.04.033. PMID 40288466
- Butler JJ, Puleo J, Harrington MC, et al. From technical to understandable: artificial intelligence large language models improve the readability of knee radiology reports. Knee Surg Sports Traumatol Arthrosc. 2024;32(5):1077-1086. doi:10.1002/ksa.12133. PMID 38488217
This article is general education, not medical advice for your situation. Every knee and hip is different — talk with your own doctor about your care.