Tuesday, May 24, 2016

ECG Blog #126 - Computerized ECG Interpretations

NOTE: This Blog post is a reproduction of Section 13.0 from my ECG-2014-ePub (www.ecg2014.com):

13.0 – Computerized ECG Interpretations:
A frequent question that arises is, “How best to use (or not use) the computerized ECG interpretation?” Opinions vary. We feel the answer depends on the goals and experience level of the interpreter.
  • Computerized ECG analysis systems are not infallible. Although they clearly have merit in certain regards — they are far from perfect at ECG interpretation. Our task is to appreciate the positives of computer systems while being aware of their drawbacks.
13.1 – Computerized Systems: Pros & Cons
At the current time — virtually all modern ECG machines automatically provide a computerized interpretation. This has benefits and drawbacks. Consider the following:
  • Computerized systems excel at computing values. This is because that’s what computers do. As a result — computerized systems are extremely accurate in calculating: i) Rate; ii) Intervals (PR/QRS/QT intervals); and iii) Axis.
  • Computerized systems are usually reliable in recognizing sinus rhythm mechanisms and normal tracings.
  • For the Expert Interpreter — the best feature of computerized systems is that they save time! There is no longer need to calculate rate, intervals or axis — since the computer instantly provides legible and accurate print-out of these values. IF the computer says, “Normal ECG” — it may literally take no more than 2-3 seconds for an experienced interpreter to peruse the tracing and sign the report (provided there is agreement with the computer interpretation).
  • For the Non-Expert Interpreter — the major benefit of computerized systems is the backup opinion the system provides. The computer may suggest findings not initially thought of by a less experienced interpreter. This encourages more careful, targeted review of the tracing. It may also be educational by the suggestions it makes. Finally — confidence is boosted when computer analysis agrees with the clinician’s interpretation.
NOTE: The computer backup opinion may also help the expert-in-a-hurry by reducing the chance that any ECG findings will be overlooked.
  • Interpretation of any one ECG by an expert provided with: i) a moment of time to sit down and give full attention to interpretation; and ii) the clinical history — will always be superior to interpretation by a machine. That said — this is not reality.
  • Reality in the “real world” — is that the clinician assigned to interpret all tracings on a given hospital or ambulatory service usually has limited time to interpret a large number of ECGs and is often asked to do so without the benefit of clinical history. As a result — it becomes easy for even an expert interpreter to overlook certain findings on occasional tracings. Knowing how to use the computerized interpretation as a “backup opinion” can be invaluable even for the most experienced of interpreters! (Grauer, Nelson, Marriott et al: J Am Bd Fam Prac 1:17-24, 1989).

CAVEATS (What the Computer May Miss): Computerized systems do not do nearly as well in evaluation of abnormal tracings as they do in assessing ECGs with minimal abnormalities. The more complex the abnormal ECG is — the more difficult it becomes for a computerized system to render an entirely accurate interpretation.
  • Computerized systems are far less accurate interpreting rhythms that do not have a sinus mechanism.
  • They may miss subtle infarctions.
  • They tend to overinterpret the J-point ST elevation that is commonly seen with early repolarization patterns. As a result — computerized systems may be prone to mislabel these normal variants as “acute MI”.
  • Computerized systems may miss pacemaker spikes/WPW/tall R in V1. They are unlikely to appreciate certain clinical entities such as Wellens’ syndrome or DeWinter T waves.
  • Many hospitals do not utilize special computer programs for interpretation of ECGs obtained on pediatric patients. Obvious problems with interpretation will arise IF a pediatric ECG is interpreted by a computer program using adult criteria.
  • Finally computerized systems by definition lack the “human Gestalt” by an expert of the overall tracing.

13.2 – Suggested Approach: How to Use the Computer
The most important point to emphasize in this Section — is that clinical use of the computerized report by non-expert interpreters should be very different than use of this same report by the expert who regularly interprets a large volume of tracings. Expertise of the interpreter therefore dictates the approach we recommend (Grauer: Practical Guide to ECG Interpretation; Mosby, St. Louis; pp 375-379, 1998).
  • For the Non-Expert Interpreter — Do not initially read the computer report. Instead — WRITE OUT (or at least think out) your interpretation first. Check findings you note with each computer statement. Then delete, modify and/or add to the computer interpretation as needed.
  • For the Expert Interpreter — Review the computer report either before or after evaluation of the ECG itself. Minimize time devoted to determination of heart rate, intervals and axis (since the computer is very accurate for these parameters). Consider more careful evaluation IF the rhythm is not sinus — or IF the ECG is interpreted by the computer as abnormal. Overread each computer statement. Place a check mark next to those that are accurate. Delete, modify or add to incorrect statements.
KEY Point: The expert interpreter is not using the computerized report to “learn”. This is because by definition — the interpretation of an expert electrocardiographer is the “gold standard”. Since computerized systems are programmed by experts — the best they can realistically hope for is to put out interpretations that equal the level of accuracy of the expert that programmed them.
  • The Expert uses the computer: i) to save time; and ii) to prevent overlooking findings when forced to read many ECGs in a limited period of time.
  • Less Experienced Interpreters do look to the computer to assist in accuracy. They are usually called on to read no more than one ECG at any one time. Therefore — the most important step for the non-expert is to first COVER UP the computerized report. It is otherwise all too easy to be biased by what the computer says. Used in this way — comparing one’s own interpretation with what the computer says optimally incorporates potential benefit from any discrepancy in interpretation that may exist.

13.3 – FIGURE 13.3-1: Do You Agree with the Computer?
Perhaps the best way to illustrate potential pros and cons of computerized interpretations — is by clinical example. Consider the ECG shown in Figure 13.3‑1 — obtained from a 78 year old woman with atypical chest pain.
  • The computerized interpretation was: Sinus rhythm; left axis (-10 degrees) — but otherwise “normal” ECG.
  • Do you agree with the computerized interpretation?
  • HINT: Be sure to interpret this ECG in its entirety by the systematic approach first — before you compare what the computer said with your interpretation.
Figure 13.3-1: ECG obtained from a 78 year old woman with atypical chest pain. The computerized report interpreted this tracing as, “left axis but otherwise normal”. Do you agree with the computerized report?

Answer to Figure 13.3-1: The rhythm is sinus. All intervals are normal. The axis is leftward (predominantly negative QRS in lead aVF) but not negative enough to qualify as LAHB (since the QRS in lead II is still upright). No chamber enlargement.
  • Regarding Q-R-S-T Changes — There are QS complexes in leads V1,V2. An r wave develops by lead V3 — and transition occurs normally between lead V3-to-V4. Although there is no more than minimal (at most) ST elevation — T waves are dramatically peaked in anterior precordial leads (especially in lead V2). There is shallow T inversion in lead III, and perhaps some nonspecific ST‑T wave flattening in lead aVF.
IMPRESSION: This example highlights the importance of overreading the computerized interpretation after you have independently arrived at your own conclusion. This is not a “normal” ECG. That statement should be crossed out on the computerized report. This is because the computerized interpretation is a medical record — and statements you disagree with should therefore be crossed out.
  • Clinical correlation is needed to determine the meaning of the abnormal findings you identified. Of Concern — is the fact that i) this woman is of a “certain age” (78 years old — so clearly old enough to have coronary disease); — and ii) she is having “chest pain” (even though it is described as “atypical” in nature).
  • While not definitive — the QS complexes in leads V1,V2 could reflect septal infarction of uncertain age. This should at least be noted in your interpretation (it was ignored by the computerized report).
  • There is marked T wave peaking — especially in leads V2,V3. This is not normal (despite also being ignored by the computerized report). Possible explanations for this abnormal T wave peaking include: i) Hyperkalemia (less likely because T wave peaking is not generalized and the base of these T waves is not narrow — but a serum K+ level should nevertheless be checked to rule this out); ii) Ischemia (which when posterior in location sometimes manifests as anterior T wave peaking); and iii) DeWinter T waves. Given the history of chest discomfort — we are most concerned with this 3rd possibility. While the J‑point ST depression that is usually seen with DeWinter T waves is missing in Figure 13.3‑1 — the ECG picture in this tracing is otherwise perfectly compatible with this harbinger sign of possible impending proximal LAD occlusion.

BOTTOM Line: It might be easy to overlook the QS complexes in leads V1,V2 of this tracing IF you allowed the computerized report to bias you prior to rendering your own independent interpretation. Hopefully — you did not overlook the obviously abnormal T wave peaking in anterior leads that somehow escaped detection by the computer. Recognition of DeWinter T waves is indication for immediate cath/acute reperfusion — so this possibility mandates immediate attention. This would have been missed had the computer report been accepted without overread. Computerized interpretations can be extremely helpful to both expert and non-expert interpreters — but knowing HOW to use the computer report always assumes first priority.
  • NOTE: For more on DeWinter T waves — Please see our ECG Blog #53
  • We refer to Computerized ECG Interpretations in our ECG Blog #12 — 


EDITORIAL COMMENT by the AUTHOR:  Prior to researching this topic (references of my work below) — I thought computerized interpretations were a waste of time. However, once I learned to appreciate their benefits and drawbacks — I learned to love computerized interpretations. They literally tripled my speed of interpretation — especially when sinus rhythm and minimal abnormalities are present. On the other hand — I ignore what the computer says when the rhythm is anything other than sinus. I’m fully aware of the need to carefully overread early repolarization patterns and tracings on patients with chest pain. Finally — I’ve observed that when less experienced interpreters truly make an honest attempt to interpret the ECG first before they look at what the computer says — that accuracy is increased.

—  —

Our Publications on Computerized ECG Interpretation include the following:
  • Grauer K: Chapter 21  Does the Computer Know Better? – from Grauer K: Practical Guide to ECG Interpretation (2nd Edition) – Mosby, St. Louis, 1998, pp 374-379.
  • Grauer K, Kravitz L, Ariet M, Curry RW, Nelson WP, Marriott HJL: Potential Benefits of a Computer ECG Interpretation System for Primary Care Physicians in a Community Hospital. J Am Bd Fam Prac 1:17-24, 1989.
  • Grauer K, Kravitz L, Curry RW, Ariet M: Computerized Electrocardiogram Interpretations: Are They Useful for the Family Physician? J Fam Prac 24:39-43, 1987.
  • Grauer K, Curry RW: Chapter 11: Use of Computerized ECG Interpretation Programs. — from Clinical Electrocardiography (Grauer & Curry Blackwell Scientific Publications, Boston, 1992, pp 418-425.


  1. I highly appreciated your post. Thank you!

    Your very instructive arguments apply also (perhaps to even an high degree) when the computer interprets ECGs of patients with pacemakers.