01
The numbers are measured for you
Heart rate, PR interval, QRS, QT, QTc and heart axis are measured on all the leads together, so you start from numbers instead of from a ruler on the paper.
For hospitals
The ECG is in your hand, and no cardiologist is on site
You are the doctor on duty, and you are not a heart specialist. TraceRead gives you a structured AI first read to check the ECG against: the rebuilt tracing, the measurements and the patterns TraceRead AI sees, each with a score from 0 to 100. The decision stays with you.
- Who
- The doctor on duty who is not a heart specialist
- Their moment
- An ECG is printed on the ward and there is no cardiologist on site to ask.
- What they get
- A structured first read to check the ECG against: rate, intervals, axis and the patterns TraceRead AI sees, each with a score from 0 to 100.
TraceRead must not be relied on alone in an emergency
A patient who is unwell is treated on clinical grounds, without waiting for an AI first read. A normal AI first read does not rule out heart disease, and TraceRead AI does not reliably find ST elevation, a sign of a heart attack. The numbers are on the clinical safety page.
What changes
What changes for you
Four things change for the doctor on duty. The numbers behind them are further down this page.
02
A list of patterns to check against
TraceRead AI checks the tracing for 71 patterns of rhythm, conduction, thickened heart muscle and ST and T changes, each with a score from 0 to 100. You keep, edit or remove each one.
03
The paper stays in front of you
The rebuilt trace is drawn back over the paper as a red line, so you compare the two. A map shows which parts of each lead the paper shows. If the line and the paper disagree, trust the paper.
04
Urgent patterns are worded as possible
A pattern that may need quick attention is written as "possible ..., review now", never as a confirmed finding. No generative AI writes clinical text: every pattern name comes from a fixed list.
Words used on this page
- 12 lead ECG
- The standard ECG: 12 views of the heart's electrical activity, printed as 12 wavy lines.
- Rebuilt trace
- The digital line TraceRead makes from the photo, drawn back over the paper so a doctor can compare the two.
- Pattern
- Something TraceRead AI sees on the tracing. It is not a finding that the patient has a disease.
- Score
- How strongly TraceRead AI sees a pattern, from 0 to 100. It is not the chance that the patient has a disease. Each pattern has its own flag line, set on other ECGs so that patterns are not missed.
- PR interval
- The time from the start of the upper chambers' beat to the start of the lower chambers' beat.
- QRS
- How long the main beat of the lower chambers lasts.
- QT and QTc
- The time the lower chambers take to beat and recover. QTc is the QT corrected for heart rate.
- Heart axis
- The main direction of the heart's electrical activity, in degrees.
Your day
How it fits your day
The ECG is printed on the ward as it always is. Four steps follow.
- Step 1
The ECG is printed
TraceRead reads the printed page: a 12 lead ECG from an adult (18 and over), at 25 mm a second and 10 mm per millivolt.
- Step 2
A nurse or technician photographs it
The photo check says what to fix. They cover the header and send. If the network is down, the photo waits on the phone.
- Step 3
TraceRead AI makes the first read
In testing, 95 in 100 ECGs were read in 7.85 seconds or less. If the rebuilt trace looks unreliable, there is no first read and a request for a new photo.
- Step 4
You check it and decide
You compare the red line with the paper, keep, edit or remove each pattern and check every number. You treat the patient on clinical grounds. Nothing reaches the patient until a doctor has checked and signed it.
The numbers
What we measured
Each number says how many ECGs it was measured on and where it came from. They come from public research ECGs and from photos made by computer, not from patients at a clinic.
0.902
how well 71 patterns are told apart (AUROC, where 1.00 is perfect and 0.50 is a coin toss)
Measured on 2,198 PTB-XL test ECGs (fold 10), original recordings. Source: Spike S2.
0.931
the same score for the 5 broad groups (normal, heart muscle damage, ST and T changes, conduction problems, thickened heart muscle); the published benchmark is 0.928
Measured on 2,198 PTB-XL test ECGs (fold 10). Source: Spike S2.
0.955
the same score for 12 rhythm patterns; the published benchmark is 0.957
Measured on 2,198 PTB-XL test ECGs (fold 10). Source: Spike S2.
20 of 29
patterns scoring 0.90 or more on ECGs from China that TraceRead AI never trained on
Measured on 45,152 ECGs from ECGs from Chapman Shaoxing Ningbo (China). Source: Spike S2.
0.6 beats a minute
average heart rate difference from a reference ECG program
Measured on 2,198 PTB-XL test ECGs against PTB-XL+. Source: Spike S4.
16.8 ms
average QTc difference from a reference ECG program; two reference programs differ from each other by 10.4 ms
Measured on 2,198 PTB-XL test ECGs against PTB-XL+. Source: Spike S4.
These are engineering measurements, not clinical claims. The page on how TraceRead AI works shows every step of a real run with its numbers. See the evidence
Features
The features you use
TraceRead has ten named features. These six are the ones the doctor on duty meets.
Trace rebuild
ReadTraceRead AI follows the ink of every lead and rebuilds it as a digital signal, 500 points a second.
- What code checks
- The rebuilt line is drawn back over the paper. On 500 photos made by computer it matched the original recordings at a median of 0.908.
- What a person does
- The doctor compares the red line with the paper.
Coverage map
ReadMarks exactly which seconds of each lead the paper shows, because most printouts show each lead for only 2.5 seconds.
- What code checks
- The pattern finder reads the signal and the coverage map together, so a missing part is never treated as a flat line.
- What a person does
- The doctor sees which parts were read.
Layout reader
ReadTraceRead AI finds the calibration pulse at the start of each row and works out where each of the 12 leads sits on the page.
- What code checks
- The pulses are matched against 14 known page layouts. It read the layout right on 99.0% of 500 photos made by computer.
- What a person does
- The doctor sees which layout was read.
Pattern finder
InterpretTraceRead AI checks the tracing for 71 patterns of rhythm, conduction, thickened heart muscle and ST and T changes, each with a score from 0 to 100.
- What code checks
- Scored on 2,198 public test ECGs it never trained on (0.902 across all 71), and on 45,152 ECGs from China.
- What a person does
- The doctor keeps, edits or removes each pattern.
Measures
InterpretHeart rate, PR interval, QRS, QT, QTc and heart axis, measured on all the leads together.
- What code checks
- Compared with reference ECG programs on 2,198 test ECGs: heart rate within 0.6 beats a minute and QTc within 16.8 ms on average.
- What a person does
- The doctor checks every number before signing.
Hold for retake
ReadWhen the rebuilt trace looks unreliable, TraceRead gives no first read and asks for a new photo.
- What code checks
- A fixed rule looks at the calibration pulses, the layout match and cut leads. It held 5.9% of 500 photos made by computer.
- What a person does
- The technician retakes the photo.
Safety
A doctor checks every AI first read
TraceRead helps a doctor read an ECG. It does not stand in for one.
An AI first read is not the last word.
- A qualified doctor reviews every AI first read. Nothing reaches a patient until a doctor signs it.
- Do not rely on TraceRead alone in an emergency.
- A normal AI first read does not rule out heart disease.
- For adults, 18 and over.
Pricing
What it costs
A single hospital chooses between Pay as you go and the Centre plan. Prices are in naira, with dollars as approximate.
Pay as you go
₦1,500per AI first read
about $1.10
Centre
₦120,000a month
about $88
A hospital group with many sites can ask for one volume price. It is on the pricing page.
Dollar figures are approximate, at ₦1,370 to the dollar.
Questions
Questions people ask
Can we rely on TraceRead in an emergency?
No. TraceRead must not be relied on alone in an emergency. A patient who is unwell is treated on clinical grounds, without waiting for an AI first read. A normal AI first read does not rule out heart disease.
Does it find a heart attack?
Not reliably. ST elevation, a rise in one part of the heartbeat line that can be a sign of a heart attack, is a pattern TraceRead AI does not reliably find. The nearest pattern scored 0.61 to 0.63 on 176 ECGs from China (0.50 is a coin toss), and a rule that measures it found 61.9% of 176 on 12 lead recordings and 50.0% on a printout layout. The full numbers are on the clinical safety page.
Which patients and which ECGs is it for?
Adults, 18 and over, and printed 12 lead ECGs at the usual 25 mm a second and 10 mm per millivolt. It is not for children, and not for rhythm strips or monitor printouts that are not a 12 lead ECG.
What does the score next to each pattern mean?
It is how strongly TraceRead AI sees that pattern, from 0 to 100. It is not the chance that the patient has a disease. Each pattern has its own flag line, set on other ECGs so that patterns are not missed. When the rebuilt trace looks unreliable, TraceRead gives no AI first read and asks for a new photo.
Where is patient data kept?
Patient data stays in the AWS Cape Town region, and it is never sent to an outside AI service. TraceRead needs only the patient's age and sex, never a name. TraceRead staff cannot see patient data by default.
See it read an ECG from a photo
We show TraceRead reading public research ECGs, and talk through how the doctor on duty gets an ECG read on your wards.