Can Artificial Intelligence Diagnose Like a Doctor?
After a recent experiment with ChatGPT, I took on Diagnosis Pad to see just how good it is in diagnosing patients when I presented it with some difficult cases.
*Spoiler alert — it diagnosed the cases perfectly and even thought of things that I myself forgot.*
Diagnosis Pad
Diagnosis Pad is a patent pending state-of-the-art application that uses Artificial Intelligence (AI) to aid practitioners with diagnosis, guidance, transcriptions and notes all in real-time, on-device and private meaning no cloud processing — just how sensitive healthcare data should be treated.
It does this by using background ambient audio heard during consultations, the system transcribes audio fragments and processes them locally on the device. Providing the practitioner with information as the consultation evolves.
Now we know how it works, let’s see how well it does.
Case 1 — Tonsillitis
This is a fairly common presentation that I see in general practice. As I had no ‘real’ patient, I spoke as both the doctor and the patient. To make the test fair and comparable to my ChatGPT experiment, I only provided a brief history, as sometimes my patients do in real life.
I was impressed with how quickly the app was listening to my conversation and coming up with real-time diagnoses and recommendations as the talk was evolving. The lists changed as more information was provided. Once I halted the conversation, I waited a few moments to see the potential diagnoses and recommendations. And there it was, the number one different diagnosis was tonsillitis. It provided me with two other potentials and a list of other questions that I may wish to ask my patient to further investigate.
One of the features of the app, is that it provides the practitioner with a medical terminology database. Meaning, if I wasn’t too sure why it arrived at a particular diagnosis or indeed what that diagnosis actually was, I could press on the diagnosis and a comprehensive answer would be given. I decided to do just that for tonsillitis. It provided me with sound reasoning of why it chose tonsillitis as its primary diagnosis and what it was, with recommendations on treatment options.
Interestingly, I asked ChatGPT the same questions in my last test. I found that it provided me with an overwhelming number of differential diagnoses and potential symptoms that I could be experiencing. Whilst I understand that the platform is not traditionally used for this purpose, Diagnosis Pad presented itself as being succinct, streamlined and accurate.
Score: Diagnosis Pad 1 Vs Doctor 1
Case 2 — Tired all the time
This is a frequent presentation that I encounter on almost a weekly basis. It may usually be a female, in her 20–30’s that feels fatigued despite conservative measures. This is a difficult diagnosis. It requires a thorough investigation and sometimes additional testing. There can be a plethora of reasons as to why someone may be tired all the time. It could range from simply not getting enough sleep, to a vitamin deficiency, a hormonal imbalance and psychological changes just to name a few. So surely, with this being so difficult, I could trip the AI up. I decided to opt for my patient having an iron deficiency as the underlying reason.
Presenting it with a very general conversation and symptoms, the patient spoke about being tired after recently having her child and despite eating very well she still felt tired. Her periods had returned and she wasn’t sure why she was still feeling exhausted despite sleeping well.
I stopped at this point and gave the app time to digest, all the while I could see a number of different recommended questions to be asked and potential diagnoses. Once it had stopped, the first diagnosis was iron deficiency anaemia, which is exactly what my patient had.
Interestingly however, it also mentioned that she could have postpartum depression. Now in my haste, I actually forgot to think of this as a potential diagnosis when I was thinking of the scenario. Postpartum depression is a real and serious reason for someone to have fatigue. I decided to press and see why the AI thought this. It told me that my patient had just had a baby — and even remembered the name and that postpartum depression can typically present with symptoms of fatigue, as well as the usual anxiety and depression.
The app also mentioned the possibility of hormonal imbalances, and prompted me to ask the patient about thyroid issues, which is another very reasonable cause of fatigue.
Alongside the recommendations and potential diagnoses, the application also transcribes the audio and makes it into notes for sharing onto the patients file reducing the admin burden of the practitioner. Which is very useful for continued engagement of the patient in a clinical setting.
Score: Diagnosis Pad 2 Vs Doctor 2* (*Though the app did a better job than me in remembering other causes!)
Case 3 — Drug Seeking
I was looking forward to presenting this case. Drawing on my experience from ChatGPT — which failed miserably at this, I was excited to see how the app would fare. Would it get that the patient was a drug seeker or simply recommend I give his medications. ChatGPT didn’t understand I was drug seeking and told me to ask my doctor for even stronger pain medications. Which is a complete fail and quite frankly dangerous advice.
Presenting the app with a common history of ‘being in town for a few days and forgetting my pain meds’. I worked through the questions as I would in clinic. What surprised me the most, was that the application was immediately able to suss out that I had an opioid addiction and that there was an issue with the frequency of my scripts.
Despite my best efforts to conceal this information, it listened carefully to the conversation and analysed it in real-time to give me a prompt diagnosis. I pressed on this diagnosis to see why it thought this, and sure enough, mentioning the patients name and medication it was concerned about the frequency and the patients desperation to obtain the medication.
This is so useful for clinicians who are starting out or who are new to an area, because undoubtedly, they will be met with these patients. Some clinicians may be naive and may have never dealt with these scenarios, but having a ‘colleague’ almost or a companion to prompt you to think that something else may be going on is very useful.
Score: Diagnosis Pad 3 Vs Doctor 3
Will AI Take Over Doctors?
Whilst Diagnosis Pad can be used as a great symptom checker, I think it is best used by a doctor as they are able to put information together to form a diagnosis. I don’t think that AI will take over a doctors role yet, doctors still have to examine patients and conduct further investigations which, presently, AI is unable to do.
However, the application has a number of great uses in the clinical setting. It can act like a prompt with questions to ask, it can help the clinician think of potential diagnoses and it also helps with note taking.
Conclusion
The application is the first of its kind to help doctors with real-time guidance, diagnoses and note taking. It does this offline and on device, meaning that privacy is maintained between the doctor and patient, as it should be in such a sensitive setting.
The application can be used by doctors, medical students, nurses, for triage purposes and even those who work remotely with limited internet connection as, after it is downloaded it does not need the internet.
Diagnosis Pad is aimed to help doctors not to replace them and can serve to be an essential tool in the doctors bag.
Try the application for free and if you have any questions or comments, drop a line to the developers here.
I hope you have found this article helpful.
Take care and stay healthy,
Dr Nora x
Please note that the contents of this article are not intended to be a substitute for professional medical advice and should not be relied on as health or personal advice. Always seek the guidance of your doctor or other qualified health professional with any questions you may have regarding your health or medical condition.
Links to products often include an affiliate tracking code which allow us to earn fees on purchases you make through them.
Dr Nora is proud to be the clinical lead for Diagnosis Pad.
