AI combines ECG and X-ray to diagnose arrhythmic disorders


Kobe University Hospital’s Dr. Makoto Nishimori and Project Assistant Professor Kunihiko Kiuchi et al. (of the Division of Cardiovascular Medicine, Department of Internal Medicine) have developed an AI that uses multiple kinds of test data to predict the location of surplus pathways in the heart called ‘accessory pathways’, which cause the heart to beat irregularly.

In this study, the researchers were able to improve diagnosis accuracy by having the AI learn from two completely different types of test results – electrocardiography (ECG) data and X-ray images. They hope that this methodology can be applied to other disorders based upon the successful results of this research. The research results were published online in ‘Scientific Reports’.

human heart red neon sign

Wolff-Parkinson-White (WPW) is an arrhythmic disorder. Patients with WPW syndrome are born with surplus pathways inside their hearts called ‘accessory pathways’, which can cause tachycardia episodes where the pulse speeds up. Catheter ablation involves using a catheter to selectively cauterize accessory pathways and can completely cure this disorder.  However, the success rate of catheter ablation varies depending on the location of the accessory pathways. Conventionally, a 12-lead ECG (i.e. a regular electrocardiography) has been used to predict accessory pathway location prior to treatment. However, this current method that relies solely on ECG is insufficiently accurate, which makes it difficult to give patients a full explanation that includes the success rate of treatment. This research study tried using AI to solve this problem.

The researchers used a methodology for teaching AI called deep learning. Deep learning involves entering the data for each patient and the corresponding answers into a program. By repeating this learning process, the program automatically becomes smarter. Using this methodology, the research group was able to present a solution to a previously unresolved problem, thus further promoting the application of AI to modern medicine.

Firstly, Dr. Nishimori’s team developed AI using only ECG data and compared its performance to previous methods. They conducted repeated learning where they gave the AI each patients’ ECG data and the accessory pathway location (i.e. the answer) in each case at the same time, successfully creating an AI with a higher accuracy rate than previous methods. However, the AI was unable to perform correct predictions every time from ECG data alone. The cause of this issue was thought to be that the ECG data is affected by the differences in size and position of each heart, therefore the ECG data did not match even when the location of the accessory pathway was the same. This problem was resolved by having the AI learn data, such as information on each heart’s size, from chest x-ray images at the same time (Figure below). By simultaneously learning both the pre-treatment ECG and X-ray image data, the AI was able to obtain the missing information and its diagnostic accuracy was significantly improved compared to when only ECG data was used.

Diagram of the AI’s architecture

Diagram of the AI’s architecture

Image source: Kobe University (adapted from: Nishimori et al., Scientific Reports 2021 (CC BY 4.0))

The advancement of AI technology in recent years has made it possible for AI to make highly accurate diagnoses based on various kinds of test data in the field of medicine. However, there are cases where data from a single test is insufficient for AI to perform an accurate diagnosis. This research study successfully increased the accuracy by having the AI learn not only from ECG results but also from chest X-ray images, which are a completely different type of data. AI-mediated accurate diagnoses will enable doctors to give pre-treatment patients a more accurate explanation of their condition, which will hopefully put patients at ease. In addition, this research could be applied to various other disorders and will hopefully lead to the implementation of AI diagnosis software.

FDA clears credit card-sized personal ECG device


AliveCor announced the FDA has cleared its credit card-sized personal ECG device capable of delivering medical-grade, single-lead ECGs in 30 seconds.

According to a company press release, the new device (KardiaMobile Card) pairs with a smartphone using Bluetooth to detect six of the most common arrhythmias and provides users with access to cardiologist analyses of ECGs, monthly heart health reports and automatic sharing of ECG recordings.

FDA approval
Source: Adobe Stock

The device’s algorithm is based on the company’s prior AI-enabled technology (Kardia).

According to the release, the device comes with a 1-year access to the company’s heart health service (KardiaCare) that offers features to help users manage and understand their heart health.

The new personal ECG device is priced at $149.

How to read an ECG?


This guide demonstrates how to read an ECG in a systematic & effective manner.

Always start by confirming the name and date of birth of the patient to confirm the ECG belongs to the right person. Also, confirm the date and time the ECG was performed.

Step 1 – Heart rate

Heart rate can be calculated simply with the following method:

  • Work out the number of small squares in one R-R interval
  • Then divide 300 by this number and you have your answer

e.g. If there are 4 squares in an R-R interval 300/4 = 75 beats per minute

.

If the rhythm is irregular:

  • Count the number of complexes on the rhythm strip (each rhythm strip is 10 seconds long)
  • Multiply the number of complexes by 6 (giving you the average number of complexes in 1 minute)
What’s a normal heart rate?
  • Normal = 60 – 100 bpm
  • Tachycardia > 100 bpm
  • Bradycardia < 60 bpm

Hint: If there are obviously P waves present, check the ventricular rate and the atrial rate. The rates will be the same if there is 1:1 AV conduction.

Step 2 – Heart rhythm

The heart rhythm can be regular or irregular.
Irregular rhythms are regularly irregular (i.e. a recurrent pattern of irregularity) or irregularly irregular (i.e. completely disorganised)

Mark out several consecutive R-R intervals on a piece of paper, then move them along the rhythm strip to check if the subsequent intervals are the same.

Hint – if you are suspicious that there is some atrioventricular block, map out the atrial rate and the ventricular rhythm separately (i.e. mark the P waves and R waves). As you move along the rhythm strip, you can then see if the PR interval changes, if QRS complexes are missing or if there is complete dissociation between the two.

Step 3 – Cardiac axis

Cardiac axis describes the overall direction of electrical spread within the heart

In a healthy individual the axis should spread from 11 o clock to 5 o clock

To figure out the cardiac axis you need to look at leads I,II & III

To get a better understanding of Cardiac Axis read this article

Normal cardiac axis

In normal cardiac axis Lead II has the most positive deflection compared to Leads I & III

NORMAL AXIS

Right axis deviation

In right axis deviation Lead III has the most positive deflection & Lead I should be negative

This is commonly seen in individuals with Right Ventricular Hypertrophy

Left axis deviation

In left axis deviation Lead I has the most positive deflection & Leads II & III are negative

Left axis deviation is seen in individuals with heart conduction defects

Step 4 – P waves

Next we look at the p waves & answer the following questions:

  • Are P waves present?
  • If so, is each P wave followed by a QRS complex?
  • Do the P waves look normal? (check duration, direction and shape)
  • If not present, is there any atrial activity e.g. sawtooth baseline → flutter waves / chaotic baseline → fibrillation waves / flat line → no atrial activity at all?

Hint – If P-waves are absent & there is an irregular rhythm it may suggest atrial fibrillation.

 

Step 5 – P-R interval 

The P-R interval should be between 0.12-0.2 seconds (3-5 small squares)

Prolonged PR interval (>0.2 seconds)

A prolonged PR interval suggests there is atrioventricular delay.

Is the prolonged PR interval fixed or does it vary across the ECG?

  • A fixed prolonged PR interval is a FIRST DEGREE AV BLOCK

 

  • If the PR interval slowly increases then there is a dropped beat, this is MOBITZ TYPE I SECOND DEGREE AV BLOCK(Wenckebach)

2nd degree AV block (Mobitz Type 1 - Wenckebach)1

 

  • If the PR interval is fixed but there are dropped beats, this is MOBITZ TYPE 2 SECOND DEGREE HEART BLOCK (clarify that by the frequency of dropped beats e.g 2:1, 3:1, 4:1)

Mobitz type 2

 

  • If the P waves and QRS complexes are completely unrelated, this is THIRD DEGREE AV BLOCK (complete heart block)

Complete heart block (3rd degree)

 

To help remember these degrees of AV block, it is useful to remember the anatomical location of the block in the conducting system:

  • First degree AV block:
    • Occurs between the SA node and the AV node (i.e. within the atrium)
  • Second degree AV block:
    • Mobitz I (Wenckebach) – occurs IN the AV node. This is the only piece of conductive tissue in the heart which exhibits the ability to conduct at different speeds.
    • Mobitz II – occurs AFTER the AV node in the bundle of His or Purkinje fibres.
  • Third degree AV block: 
    • Occurs anywhere from the AV node down causing complete blockage
Shortened PR interval

If the PR interval is short, this means one of two things:

  • Simply, the P wave is originating from somewhere closer to the AV node so the conduction takes less time (the SA node is not in a fixed place and some people’s atria are smaller than others!)
  • The atrial impulse is getting to the ventricle by a faster shortcut instead of conducting slowly across the atrial wall. This is an accessory pathway and can be associated with a delta wave (see below which demonstrates an ECG of a patient with Wolff Parkinson White syndrome)

Step 6 – QRS complex

There are several aspects of the QRS complex to assess.

Width

Width can be described as NARROW (< 0.12ms) or BROAD (> 0.12ms)

  • A narrow QRS complex occurs when the impulse is conducted down the bundle of His and the Purkinje fibre to the ventricles. This results in well organised synchronised ventricular depolarisation.
  • A broad QRS complex occurs if there is an abnormal depolarisation sequence – for example, a ventricular ectopic where the impulse spreads slowly across the myocardium from the focus in the ventricle. In contrast, an atrial ectopic would result in a narrow QRS complex because it would conduct down the normal conduction system of the heart. Similarly, a bundle branch block results in a broad QRS because the impulse gets to one ventricle rapidly down the intrinsic conduction system then has to spread slowly across the myocardium to the other ventricle.
Height

Describe this as SMALL or TALL:

  • Small complexes are defined as < 5mm in the limb leads or < 10 mm in the chest leads.
  • Tall complexes imply ventricular hypertrophy (although can be due to body habitus e.g. tall slim people). There are numerous algorithms for measuring LVH, such as the Sokolow-Lyon index or the Cornell index.
Morphology

This is where you assess the individual waves of the QRS complex.waves

Delta wave

The mythical ‘delta wave’ is a sign that the ventricles are being activated earlier than normal from a point distant to the AV node. The early activation then spreads slowly across the myocardium causing the slurred upstroke of the QRS complex.  Note – the presence of a delta wave does NOT diagnose Wolff-Parkinson-White syndrome. This requires evidence of tachyarrhythmias AND a delta wave.

Q  waves

Isolated Q waves can be normal. A pathological Q wave is > 25% the size of the R wave that follows it or > 2mm in height and > 40ms in width. A single Q waves is not a cause for concern – look for Q waves in an entire territory (anterior / inferior) for evidence of previous MI.

Inferior Q waves (II, III, aVF) with T-wave inversion due to previous MI - http://lifeinthefastlane.com/ecg-library/basics/q-wave/

 

R and S waves

Look for R wave progression across the chest leads (from small in V1 to large in V6) with the transition from S > R wave to R > S wave should occur in V3 or V4. Poor progression (i.e. S > R through to leads V5 and V6) can be a sign of previous MI but can also occur in very large people due to lead position.

 

Poor R wave progression (previous anterior MI) - Image sourced from http://lifeinthefastlane.com/ecg-library/poor-r-wave-progression/

 

 

J point  segment

The J point is where the S wave joins the ST segment. This point can be elevated or The ST segment that follows is then raised…High take off (or benign early repolarisation to give its full title) is a normal variant that causes a lot of angst and confusion as itLOOKS like ST elevation.

  • It occurs mostly under the age of 50 (over age of 50, ischaemia is more common and should be suspected first) and rarely over 70.
  • Typically, the J point is raised with widespread ST elevation in multiple territories making ischaemia less likely
  • The T waves are also raised (in contrast to a STEMI where the T wave remains the same size and the ST segment is raised)
  • The changes do not change! During a STEMI, the changes will evolve – in BER, they will remain the same.

Benign Early Repolarisation (High take off) -

Step 7 – ST segment

The ST segment is the part of the ECG between the end of the S wave & start of the T wave.

In a healthy individual it should be an isoelectric line (neither elevated or depressed).

Abnormalities of the ST segment should be investigated to rule out pathology.

 

ST elevation

ST elevation is significant when it is > 1mm (1 small square) in relation to the baseline.

It is most commonly caused by acute myocardial infarction.

The morphology of the ST elevation differs depending on how long ago the MI occurred.

stemi

 

ST depression

ST depression is significant when it is >1mm (1 small square) in relation to the baseline.

ST-depression lacks specificity, therefore you shouldn’t jump to any diagnostic conclusions.

It can be caused by many different things including:

  • Anxiety
  • Tachycardia
  • Digoxin toxicity
  • Haemorrhage, Hypokalaemia, Myocarditis
  • Coronary artery insufficiency
  • MI

As a result you must take this ECG finding & apply it in the context of your patient.

STD

Step 8 – T waves

The T waves represent repolarisation of the ventricles.

Tall T waves

T waves are tall if they are:

  • > 5mm in the limb leads and
  • > 10mm in the chest leads (the same criteria as ‘small’ QRS complexes).

 

Tall T waves can be associated with:

  • Hyperkalaemia (“Tall tented T waves”)
  • Hyper-acute STEMI
Peaked T
Inverted T waves

T waves are normally inverted in V1 and inversion in lead III is a normal variant.

Inverted T waves in other leads are a nonspecific sign of a wide variety of conditions:

  • Ischaemia
  • Bundle branch blocks (V4 – 6 in LBBB and V1 – V3 in RBBB)
  • PE
  • LVH (in the lateral leads)
  • HCM (widespread)
  • General illness

Around 50% of ITU admissions have some evidence of T wave inversion during their stay.

Comment on the distribution of the T wave inversion e.g. anterior / lateral / posterior leads.

You must take this ECG finding & apply it in the context of your patient.

TWI

Biphasic T waves

Biphasic T waves have two peaks and can be indicative of ischaemia and hypokalaemia.

 

Flattened T waves

Another non-specific sign, this may represent ischaemia or electrolyte imbalance.

 

U waves

Not a common finding.

The U wave is a > 0.5mm deflection after the T wave best seen in V2 or V3.

These become larger the slower the bradycardia – classically U waves are seen in various electrolyte imbalances or hypothermia, or antiarrhythmic therapy (such as digoxin, procainamide or amiodarone).

Prominent U waves in a patient with Hypokalaemia

Summary

Having a system whilst working through ECGs is essential until you gain the experience required to start using pattern recognition to speed up the process.

Novel troponin test may rule out heart attack risk.


A single measurement of the new high-sensitivity cardiac troponin T (hs-cTnT) and a normal electrocardiogram (ECG) might help doctors predict which patients with chest pain are at low risk of heart attack and can be sent home from the emergency department (ED), a study has shown.

Researchers in Sweden sought to determine the negative predictive value of undetectable hs-cTnT (<5 ng/L) and an ECG without significant ST elevation or depression for the primary endpoint of myocardial infarction (MI) within 30 days among 14,636 patients (age >25 years) with chest pain who presented at the ED of a hospital in Sweden over a 2-year period. [Abstract 403-14-LB-13200; J Am Coll Cardiol 2014; doi:10.1016/j.jacc.2014.03.017]

Nearly 9,000 patients (61 percent) who had undetectable hs-cTnT (<5 ng/L) on initial testing (as measured with Elecsys® 2010, Roche Diagnostics) were included in the study. They were younger and less likely to have diabetes, chronic kidney disease, prior MI or stroke compared with patients who had increased hs-cTnT levels (5-14 and >14 ng/L).

Hospitalization rates were lower among patients with undetectable hs-cTnT (21 percent vs 44 and 82 percent for 5-14 and >14 ng/L, respectively). Diagnosis of MI increased with increasing levels of hs-cTnT (2 percent for <5 ng/L, 3.1 percent for 5-14 ng/L, 4.1 percent for >14 ng/L).

Within 30 days, 39 patients with undetectable hs-cTnT had MI (24 had significant ECG changes, 15 had normal ECG). The negative predictive value of the tests for MI was 99.8 percent (95% CI, 99.7-99.9) and 100 percent for death (95% CI, 99.9-100). After adjusting for age, sex, diabetes, prior MI and eGFR, there was no significant difference in the risk of death between patients discharged from the ED and those hospitalized within 365 days (hazard ratio [HR], 0.73; 955 CI, 0.48-1.12).

“Patients with chest pain who have an initial hs-cTnT of <5 ng/L and no signs of ischemia on ECG, independent of duration of chest pain and other risk factors, have a minimal risk of MI within 30 days, and no risk of death,” said lead investigator Dr. Nadia Bandstein from the Karolinska Institute in Stockholm, Sweden. “These patients can be safely discharged from the ED.”

The study has some clinical implications – it can reduce overcrowding of the ED, prevent unnecessary admissions, and save doctor-patient time. If MI can be ruled out more quickly, 20 to 25 percent of all hospital admissions for chest pain can be prevented, Bandstein said.

Panelist Dr. Allan Jaffe from Mayo Clinic, Rochester, Minnesotta, US, however, cautioned that the assay is likely to be effective only in low-risk groups. “We have to be careful in defining how we rule people in and out. In the long run, we will be able to validate the strategy… we need to do it a little bit more rigorously.”

Remote, Minimally Invasive Atrial Fibrillation Detection by Medtronic.


Medtronic has just released the smallest implantable cardiac monitoring device currently available on the market. The compact Reveal LINQ monitor is also the only implantable cardiac monitor that collects data continuously and wirelessly for up to three years; not only can the device store data, but it communicates remotely with a MyCareLink patient monitor, which taps Medtronic’s Carelink® Network. This powerful tool provides doctors with patient data and notifications, empowering physicians to make earlier clinical decisions based on CareLink’s customizable, actionable and comprehensive reports. The Reveal LINQ implantable ECG monitor is designed for patients at risk for cardiac arrhythmia or patients that show symptoms that may suggest cardiac arrhythmia. It provides unparalleled preventative care by notifying cardiologists of detected atrial fibrillation events, classed by rhythm using an exclusive detection algorithm.

Remote, Minimally Invasive Atrial Fibrillation Detection by MedtronicLINQ     Reveal Linq     Reveal-XT-and-Linq