Tuesday, October 17, 2023

Repost from Saturday Oct 14: The Queen gets it right with a better image

This is a case I posted Saturday, October 14.

This ECG was texted to me after the fact with no information. What do you think?

This is the ECG that I used for the post.  It is high quality screenshot of a pdf:

I had said that the Queen said this is Not OMI with low confidence


However, the image that had been texted to me and submitted to the Queen was not this one above.  It was this one below: (photo of computer screen, lower quality):

This is the image for which the Queen interpreted "Not OMI with Low Confidence"




After posting, I wanted to get explainability for the ECG, and I was unable to get the explainability from the submission made by the person who sent me the ECG.

I needed to submit it myself.

So I submitted the high quality ECG again (the one at the top of this post), and this is what I got:

Notice (upper left) now she says "OMI - High Confidence"!!


Thus, although you can take a photo of any ECG and upload to the Queen, the quality of your photo and the exact screenshot you take of it can make a difference.


PM Cardio is working on improving this.


But for the time being, the better the image, the more reliable the diagnosis.


The good news is this:

With the bad image, the Queen did not detect the acute LAD OMI.

With the good image, she got it with HIGH confidence.




===================================

MY Comment, by KEN GRAUER, MD (10/17/2023):

===================================
QOH (Queen OHearts) — is an amazing AI application. Her ability to instantly and accurately identify OMI greatly expands our potential for expediting clinician recognition of those patients in need of immediate cardiac cath with PCI.
  • I found it interesting (albeit not surprising) to learn — that QOH changed her interpretation of Dr. Smith’s October 13, 2023 post, depending on whether QOH was given a high quality resolution pdf screenshot vs the original smart-phone version that was texted to Dr. Smith.


I offer the following reflections on this change of interpretation by QOH:
  • Reflection #1: As good as QOH is for identifying acute OMI — she isn’t perfect. And while I suspect that QOH will never achieve “perfection” — she will certainly continue to get better-and-better, especially once she becomes able to: i) Integrate the clinical history into her decision-making algorithms; ii) Compare current tracing with previous ECGs on the patient; iii) Better account for suboptimal images inherent with taking no-time-to-lose bedside smart phone photos; and, iv) Add an additional umpteen thousand tracings of cases with cath results to her already incredibly extensive ECG data bank.
  • Reflection #2: The reason (in my opinion) that QOH will never achieve “perfection” — is that she is not the “Gold Standard”, and by definition, she never will become the Gold Standard. My version of the “Gold Standard” for recognition of acute OMI on ECG is the interpretation of any given tracing by Dr. Smith or Meyers — as I believe THEY have the most experience of anyone, in not only assessing emergency ECGs from patients who present with new chest pain — but more importantly, the benefit from years of detailed follow-up, including serial tracings, serial troponin assays, bedside Echos and cath results (witness Dr. Smith’s ECG Blog with probably more cases detailing the clinical course of such patients than any other resource). Since Drs. Smith and Meyers are the “teachers” of QOH (overseeing her ongoing development) — they will always (by definition) be “better” at OMI diagnosis than QOH.
  • Reflection #3: The above is not to say that Drs. Smith and Meyers are 100% accurate in predicting the culprit artery for all cases of acute OMI — since no one can perfectly predict those cases in which despite a “correct” ECG interpretation — unusual forms of multivessel disease, prior events, anatomic variants, surprising collateralization patterns and/or confounding conditions such as Takotubo cardiomyopathy or myocarditis — may occasionally derail even the best of predictions.
  • Reflection #4: Attaining 100% accuracy for when a given patient with new CP has an ongoing OMI is not the most important goal of emergency ECG interpretation — simply because 100% accuracy is not attainable. more important goal that is attainable — is knowing at the earliest possible moment when prompt cath is needed (ie,  because of the history, available lab data [ie, troponins, bedside echo, etc.] and the available ECG(s). This determination of when prompt cath is urgently needed is a teachable skill. The "good news" — is that this skill can instantly be improved when clinicians are receptive to input from QOH.

  • Reflection #5: It is because QOH is not infallible that clinician input will always be needed — just as human oversight is always needed to ensure that any computer operation achieves what it is meant to achieve. This last reflection was borne out in the October 13, 2023 post — in which Dr. Smith disagreed with the QOH prediction — and Dr. Smith's interpretation was correct.

  • Reflection #6: On those rare occasions when I find myself disagreeing with the QOH interpretation — I stop and ask myself WHY might it be that QOH is "off"? We learned from Dr. Smith's October 13 post that one of the reasons why QOH may miss a diagnosis — is that the smart phone, taken-in-haste photo of the ECG was clearly suboptimal. This is where clinician input comes in (ie, recognizing the artifact, angulation or improper lighting that might alter or distort the ECG image that QOH is given to analyze). Clinician appreciation that there is a valid reason why QOH may miss a diagnosis (ie, bad image) — reinforces the importance of clinician input in such cases.
  • Alternatively — when Drs. Smith and Meyers encounter a tracing in which despite a good quality ECG, QOH missed the diagnosis — this is an opportunity for them to program the corrected information such that QOH continues "to learn", until she achieves the "near perfection" accuracy level being aimed for.

  • Reflection #7: Regardless of whether the clinician does or does not agree with the QOH interpretation — the benefit of QOH availability is that it provides an "experienced" extra opinion. I know myself that when my interpretation matches QOH that I feel reassured — and when my interpretation differs from QOH, that this causes me to take one more look at the ECG to make sure I did not miss anything.

  • Final Reflection: QOH has already become so good at OMI recognition — that unless a clinician has extensive experience in recognizing non-ST-elevation OMIsclinicians will do well to respect the opinion rendered by QOH





No comments:

Post a Comment

DEAR READER: I have loved receiving your comments, but I am no longer able to moderate them. Since the vast majority are SPAM, I need to moderate them all. Therefore, comments will rarely be published any more. So Sorry.