I’ve been using Alivecor’s Kardia personal ECG to help manage my patients with atrial fibrillation for 7 years now. Early on I pointed out (see here) that Kardia was easily confused by extra beats, both premature atrial contractions (PACs) and premature ventricular contractions (PVCs.)
Sometimes such rhythms were diagnosed incorrectly as atrial fibrillation but more commonly they were designated “unclassified.” For patients, especially those not connected like mine are to a cardiologist through KardiaPro, the inconclusive readings were often a source of consternation. This problem was not solved by the 6 lead Kardia.
Lead I (top) and Lead II (bottom) from Kardia 6L showing sinus rhythm with PVCs read as Unclassified
Kardia is not alone in being confused by PVCs and PACs. Apple Watch has the same problem. Even highly sophisticated computerized interpretations of 12-lead ECGs often confuse premature beats with atrial fibrillation. Such misdiagnoses trigger cardiology consults routinely. Even board-certified cardiologists, especially if they are reading rapidly and carelessly, not infrequently misread SR with PACs as atrial fibrillation.
A New Personal ECG Day Dawns
It was exciting, therefore, to read Alivecor’s announcement that the company, long a leader in AI-based personal ECG technology has been given a new 510(K) clearance by the FDA for use of new interpretive ECG algorithms (AI V2) which are designed to recognize and diagnose these premature beats.
I had the opportunity to discuss AI V2 with Dr. David Albert (@DrDave01) the Founder and Chief Medical Officer of AliveCor earlier this week.
The focus, Albert said, of the improved algorithm, was to reduce the number of unclassified Kardia recordings and help patients deal with irregular heart rhythms or palpitations that are not atrial fibrillation.
The AI algorithm has been improved by analysis of 90 million ECGs Alivecor has access to and Albert indicates that with AI V2 Kardia’s accuracy at detecting atrial fibrillation will be substantially better.
Alivecor’s internal studies indicate that this combination of accurate identification of PVCs and PACs plus improved AF recognition will lower the number of unclassified Kardia ECGs by 80%.
Albert indicates the accuracy of AI V2 recognizing PVCs and PACs as determined by internal testing of thousands of ECGS is high with both sensitivity and specificity >90%.
Lead I versus Lead II
Both the original Kardia single-lead device and the newer 6 lead Kardia utilize Lead I (left arm-right arm electrodes) information to identify cardiac rhythm so users can anticipate a diagnostic improvement in both versions when the new software is released in early 2021. Albert indicates several external researchers are eager to test AI V2 in real-world patients at that time.
Of course, the 6L Kardia has Lead II (left arm-left leg electrodes) available for rhythm analysis and Albert indicates that “very soon” Alivecor will receive approval for utilization of Lead II analysis for rhythm determination. Lead II is often superior to Lead I for detecting p waves, the hallmark of normal sinus rhythm.
Atrial flutter, because it is often very regular is often misdiagnosed by computerized algorithms that are utilizing only Lead I whereas it is typically easy to spot in Lead II.
Albert anticipates that the Kardia 6L utilizing Lead II analysis will receive approval for diagnosing atrial flutter and will be proven even more superior for diagnosing PVCs, PACs, and AF.
I am eager to see the accuracy and utility of AI V2 in my Kardia-carrying patients and will update readers when I have my own evaluation of the new algorithms.
I’ll talk more about Alivecor’s recently announced KardiaCare program and about Lead II recordings with Apple Watch in future posts here.
AIEIOingly Yours,
-ACP
35 thoughts on “Alivecor’s AI V2: Upgraded AI-powered Algorithm Will Identify PVCs and PACs Thereby Reducing Unclassified Mobile ECGs”
Thanks, Dr. Pearson. I emailed you my comparisons and comments accordingly.
I love the 6L vs. my old Kardia single lead device. The 6L uses Bluetooth Low Energy technology and therefore does not need to access my iPhone XS’s microphone like the single lead Kardia device did. Therefore, the 6L can be used with MFI (made for iPhone) hearing aids without having to turn them off before taking a EKG reading. The 6L also has a better range capability between the device and the iPhone too.
Dr. Pearson, did the 6L ever prove itself more useful to you than the single lead Kardia device in your management of your patient’s AFIB and other arrhythmias? I find occasional PVC’s now with the updated V2 algorithm when those used to be considered “unclassified” in the original algorithm. The AFIB interpretation seems about the same to me in either device.
John,
Thanks for these observations.
I can’t say the 6L has been significantly more helpful than the single lead in my patient care.
I’m interested in more details of your comparative analysis of the two.
Perhaps you’d like to write up your experience and we could post it on the blog here.
Feel free to email me at drp@theskepticalcardiologist.com to discuss further
Dr P
Hello,
I’m interested in purchasing one of these two units. I have something going on with respect to my heart’s rhythm. Working with my Dr., we have not been able to catch or record an event for evaluation. I have worn a monitoring device for 24hrs and nothing shows up, the condition is very infrequent. I’m a bit confused about the Kardiacare subscription…is it required, will either unit still detect and record events without Kardiacare? Thank you
Hi. Loving your blogs! Been reading up about at home and wearable ECG devices since my mum has been having heart trouble.
I’m thinking of buying the KardiaCare 6L but I wanted to get your opinion on the Withings devices, the ScanWath, Move ECG and BMP Core…
I think your followers and myself would love to get a review of their technology.
Sid
Sid,
I haven’t reviewed those products you mentioned but will ask Withings if they can provide me with their devices to demo.My sense is these aren’t available here in US. Are you in UK?
Regarding Kardia be aware that I am not recommending the 6L over the original and cheaper single lead ECG device right now.
Dr P
Hi
Yes I am in the UK actually but I think they are available in the US from a very limited number of retailers.
If you could get hold of the Withings that would be amazing.
I do understand that you are not recommending the Kardia 6L over the basic model. I haven’t decided what I am going to get just yet.
I appreciate you must be very busy so thanks for taking the time to reply.
Sid
I spoke to Withings and they told me there was nothing they could send me here in the US right now.
Dr P
Oh that’s a shame. Thanks anyway!
I use the KardiaMobile 6L. I like it’s integration with the IOS Omron app so I can track BP and ECG. The improved Kardia Algorithm will be very helpful to reducing my “Unclassified” or “Possible AF” instant analysis indications.
I often send my non NSR instant analysis to Kardia for further analysis (@$25 each send). I’m pretty disappointed in the “professional cardiologist” analysis from Kardia so I’m not using that service much anymore.
Out of necessity, I’ve started looking at my ECG chart and I now appreciate the trace information better. To gain this appreciation, I looked at Paula Johnson’s YouTube videos. The one on PVC naming was helpful to me.
Main page: https://www.youtube.com/channel/UCcu-wBRz1kY6afLnZ3ZxFng
PVC Naming: https://www.youtube.com/watch?v=uiH8TEnF14g
Next I did a test and obtained a Kardia ECG using my 6L. The instant Analysis was NSR. My HR via Kardia was 56 BPM (I’m on Sotalol so my normal HR is low). Proceeding with my test I then examined the trace and saw it was uniform with 3 “hiccups” which were spread out within the 30 seconds trace. Based on the PVC naming video it looked like these 3 “hiccups” were unifocal PVC’s. I’m not an expert but remember I’m doing this out of necessity. Let’s see why:
So I sent the trace to Kardia for analysis.
Result from Kardia: Sinus Bradycardia stating: “Normal Heart Rhythm with HR less than 60 bpm.” No mention of the “hiccups” at all. I was certainly expecting some mention of these “hiccups,”, especially since there was mention of “Multiple PVC’s ” in a past trace I sent to Kardia.
Back on to the Withings; I see they have a USA store:
https://www.withings.com/us/en/
Jim Kirk
Let’s have a look at the hiccups.
Professor: FYI, the IOS OmronConnect BP/EKG app you speak of is not yet compatible with the AliveCor Kardia V2 algorithm, at least as of 2 days ago. The Omron app states “Unsupported – Update App” but there is no update yet.
Also, the updated V2 algorithm is only available with a $99 USD per year annual KardiaCare subscription. However, if you do upgrade to the V2 algorithm just use the V2 upgraded app from AliveCor Kardia as it works flawlessly for both EKG and Omron BP functions.
John,
FYI
“The new determinations are the result of an update to AliveCor’s KardiaAI software that was cleared by the FDA in November. According to the filing, the updated software is specifically compatible with the KardiaMobile System, the Triangle System and the Omron Model BP7900 Blood Pressure Monitor + EKG (for which the two heart health device companies partnered back in 2018).”
From: AliveCor updates its ECG software to determine three additional heart conditions
https://www.mobihealthnews.com/news/alivecor-updates-its-ecg-software-determine-three-additional-heart-conditions
Jim Kirk
dear TSC, how will the new algorithms yet into our old devices? If at all?
Dan
Dan,
I assumed that this would just be a free app software upgrade but that is an excellent question for which I shall seek clarification.
A while back Kardia announced a new program called KardiaCare which I am looking into and will post on as time permits
Dr P
Dear Dr. Pearson, I am a regular reader of your blog. As an atypical cardiologist, I have received an MSEE and a Ph.D. in computer science and cardiology from Stanford University. It is amazing how people are overenthysiastic about Artificial Intelligence, forgetting that a single line of computer code, one decision based on one single “if” is a pure demonstration of AI. Back in the 80´s, I had the privilege of taking classes with Prof. Feigenbaum, the “father” of AI. It is quite disappointing to see that in the last 40 years, the major advance of AI stems from the increase in computer power.
Ricardo,
It’s nice to hear from the atypical cardiologist! I’m not a fan of AI and I try to minimize use of the term. it seems like adding AI to whatever is being done is a reason for publicity, excitement and publication.
I’m aware of Harvey Feigenbaum, of Indiana U., who isoften called the father of echocardiography. Is this the same Feigenbaum?
Dr P
Dear Dr., I am pleased with your return. The computer scientist is Edward Feigenbaum, who also served as a Chief Scientist of the US Air Force. I do have a lot of projects and ideas that I would be happy to discuss with you. Your analysis is sharp and precise like a scalpel incision.
Ricardo,
I wonder if the two Feigenbaums were related?
Dr. P
I will check it out.
Very good to hear that continuous improvements are occurring in the instant analysis algorithm.
i’m a 6L user and I often send my unclassified ECG to AliveCor to ease my consternation. Be nice to not have to do this so often.
So, I’m pretty disappointed that the send-it-in for analysis service was costing $19 up till November 2020 and it has now, as of Dec 2020, jumped to $25 per analysis. If AliveCor had the instant analysis improvement in the wings they erred by raising the price of the analysis.
I wasn’t aware of the price change as I and my patients don’t utilize that service but good to know for readers.
Perhaps they upped the price to make up for the 80% drop in unclassified interp business?
I have had PVC’s for about 5 yrs now and have had every test under the sun with no good results. I have been told by my consultant to buy a ALIVCOR machine so L can self check my PVS’c BUT I spoke to Alivcor in the USA who said that this wouldn’t work for me because you have to hold the pads in your finger tips and wait till you get a palpitation. My doc knows very well that my palps are only at nightime when I am lying down and are very random. They seem to be positional (if there is such a thing)….I get nothing during the day or even when I run or go to the gym….back to square one I guess !
Michael,
If the nocturnal PVCs last long enough to grab the Karadia (presumably at bedside near your smartphone) you should be able to make the recording and capture them
Is flutter just rapid succession of p-waves and is that what Kardia’s lead 2 will be able to “see”? With lead 1, are the extra p-waves in flutter during PR interval drowned out by the QRS complex?
With atrial flutter there is a characteristic circular pathway of electrical activity in the right atrium at rates of 250-350 BPM. In typical atrial flutter this activity manifests as “sawtooth” waves in the inferior leads, one of which is lead II. The flutter waves march through the QRS complexes but should be visible also in between. Lead II sees them better due to the vector of electrical activity which is different from the normal sinus node originating p waves.
Dr P
Thank you for answering my questions. Question regarding PACs vs flutter, are PACs just AV conducted flutter or is the pathology different? It appears to me that if the AV node has not finished its refractory period the conduction stops there and would result in just another p-wave. In looking at ECG tracings that show PACs I don’t see the saw tooth pattern in lead II like flutter but see a longer PR interval? I don’t understand why there are no distinct pwaves with PACs like there are in flutter because it appears to me that the AV node is blocking the conduction circuit to the bundle of His and therefore no QRS wave? (Note: Dr. P, Please don’t post the part in parenthesis, If this is too far off topic just delete. This is very fascinating to me, however this might deviate to much from the discussion of mobile EKGs and I don’t want to clog up the blog with discussions not relevant to the topic)
Christian,
I’m leaving this discussion in the post comments because others may be interested.
Pacs are isolated atrial contraction, not a continuous recurring cyclical wave of RA electrical activity like flutter. I think the large wavelet of cyclical RA electrical activity is why there are the “sawtooth ” waves.
Sometimes PACs are blocked in the AV node when they occur very prematurely and the p wave is all you see.
The pr interval varies depending on exactly where in the atria the PAC originates, the farther from the AV node, the longer the pr interval.
I use the Kardia device to monitor my rhythm. A few years ago I had paroxysmal atrial fibrillation and I didn’t want to take blood thinners forever so I underwent a five hour ablation procedure and it worked! No more a. fib. for this cowboy!
William,
Ablation is not a cure. Recurrent atrial fibrillation post-ablation is common and can be silent so recommendations are for continued use of blood thinners post ablation.
Dr. P
Dr P
This is good to hear. As you know, I currently use the 2 lead KardiaMobile device. I have periods of PACs (either identified as Normal or Unclassified), AFIB (generally correctly identified) and Atrial flutter (either identified as AFib or Tachycardia). I look forward to the improved algorithm.
Mike,
I think you mean the single lead Kardia ECG. It’s got two electrodes on one side and typically you put left hand fingers on the left side, right hand fingers on the right side.
The 6 L has 3 electrodes, two on the top for the hands and one on the bottom traditionally placed on left leg.
Dr P
So my PVCs are very hard to capture with a normal chest band monitor and are so random to you can’t predict when Ime gonna get them. A typical event would be me being asleep with my HR in the 48-55 range(via a wristwwatch monitor) Then I get awoke by a jolt in my chest and my reading goes up to around 95.Will the new Alivecor capture this do you think ?
It will capture what is going on with the cardiac rhythm at the time you make the recording. If the episode resolves within the 30 seconds it takes to grab the device and start a recording it won’t
I need something thats going to record automacally over a 8 hr period, that why I was using a halter monitor but results were sketchy. I cannot anticipate when i am going to get them to cannot press a button to record manually…..all a bit of a nightmare.