Pattern definitions

Uterine contraction — a transient increase in uterine tone recorded on the tocograph (TOCO), rising from the baseline uterine tone and returning to it.
For annotation, only the start (the trace leaving the baseline TOCO) and the end (its return to the baseline TOCO tone) are marked — the peak is not annotated. ^[1]

1. Ayres-de-Campos D, et al. FIGO consensus guidelines on intrapartum fetal monitoring: Cardiotocography. Int J Gynecol Obstet 2015;131(1):13-24

Baseline TOCO — the stretches of the toco trace between contractions, representing the resting uterine tone. With external tocodynamometry only a relative value is obtained (it drifts with sensor and patient movement), so the absolute basal tone cannot be read reliably — internal intra-uterine pressure monitoring is required for that.
^[1]

1. Ayres-de-Campos D, et al. FIGO consensus guidelines on intrapartum fetal monitoring: Cardiotocography. Int J Gynecol Obstet 2015;131(1):13-24

Artifact (TOCO) — something appears on the toco channel that is not a uterine contraction (e.g. maternal movement, repositioning of the sensor, coughing or pushing on the transducer).

The TOCO tracing is too poor to locate the uterine contractions: it is impossible to identify individual contractions over this period (weak signal, displaced probe, excessive noise...). Different from an artifact, which is something appearing on the TOCO channel that is not a contraction.

Not analysed / Unsure (TOCO) — a segment of the toco channel left unannotated, or where the annotator is unsure whether it is a contraction, baseline tone or artefact. Such periods are excluded from the ground truth for this channel (they do not count in training/evaluating the IA)

True signal (MHR) — a portion of the maternal-rate channel that faithfully reflects approximately the true maternal heart rate. During labour it usually lies below the FHR, rises with contractions and maternal pushing, and may show accelerations coinciding with expulsive efforts.

False signal (MHR channel) — a portion of the trace annotated on the maternal-rate channel that does not reflect the true maternal heart rate. It arises from the same causes as FHR false signals: the sensor may report a harmonic (double, half, triple) of the MHR or may lose/ambiguously capture the signal.

A false continuous line interpolating true and false signals over a period of 15s is technically possible.




^[1]

1. Boudet S, et al. Use of Deep Learning to Detect the Maternal Heart Rate and False Signals on Fetal Heart Rate Recordings. Biosensors 2022;12(9):691

Not analysed / Unsure (MHR) — a segment of the maternal-rate channel left unannotated, or where it is uncertain whether the trace is a true or a false maternal signal. Excluded from the ground truth for this channel.

True signal (FHR) — a portion of the FHR trace that faithfully reflects the true fetal heart rate. Only these segments undergo morphological analysis (baseline, accelerations, decelerations) and variability assessment. On the Doppler channel a true signal must be distinguished from maternal-rate pick-up and from harmonics (double or half the true rate).
^[1]

1. Boudet S, et al. Use of Deep Learning to Detect the Maternal Heart Rate and False Signals on Fetal Heart Rate Recordings. Biosensors 2022;12(9):691

False signal (FHR) — a portion of the FHR trace that does not reflect the true fetal heart rate. On a Doppler (ultrasound) sensor this is frequent, especially during the second stage of labour, and can be dangerous when it masks fetal distress.
Main mechanisms (Boudet et al. 2022):
Maternal heart rate (MHR) recorded instead of the fetal rate — the sensor locks onto the maternal pulse. During the second stage of labor,
one should be very cautious of the “double mountain peak” sign when
large amplitude accelerations coincide with ongoing uterine contractions
(Figure below) because of erroneous monitoring of the maternal heart rate as the FHR.



Harmonics of the true rate — the sensor reports double, half or (more rarely) triple the true FHR (or a harmonic of the MHR).
Missing / ambiguous signal — dropout or an unstable, uninterpretable trace.

Generally, there are missing signal periods between false signals period and true signal periods, but sometimes it can be interpolated over a few seconds.


The internal (scalp-ECG) channel is far less prone to false signals, but it can still have aberrant signals.

^[1]
[2]

1. Boudet S, et al. Use of Deep Learning to Detect the Maternal Heart Rate and False Signals on Fetal Heart Rate Recordings. Biosensors 2022;12(9):691
2. Al Fahdi M, Chandraharan E. True versus Spurious Intrapartum FHR Accelerations on the CTG. Glob J Reprod Med 2020

Not analysed / Unsure (FHR true/false) — a segment of an FHR channel left unannotated, or where it is uncertain whether the trace is a true or a false signal. Excluded from the ground truth for this channel.

Baseline (FHR) — the mean level of the FHR when it is stable, read over segments of at least 2 minutes (excluding accelerations, decelerations and marked-variability periods).

The Baseline period annotation marks the stretches of trace that represent this stable baseline (i.e. neither acceleration nor deceleration).

Neither baseline nor A/D — a fluctuation that is neither a stable baseline nor a true acceleration/deceleration, typically a small acceleration or deceleration that does not reach the 15 bpm / 15 s criteria.

Deceleration — unsure / not analysed type — a period recognised as a deceleration but whose sub-type (early, variable, late, prolonged) has not yet been determined or is uncertain.

FIGO Definition: decelerations that are shallow, short-lasting, with normal variability within the deceleration and are coincident with contractions. They are believed to be caused by fetal head compression [25] and do not indicate fetal hypoxia/acidosis

FIGO Definition (2015): V-shaped deceleration with a rapid drop (onset to nadir in < 30 s), good variability within the deceleration, rapid recovery to the baseline, and varying size, shape and relationship to uterine contractions. Variable decelerations constitute the majority of decelerations during labour and reflect a response to increased arterial pressure, as occurs with umbilical cord compression and resultant transient hypoxaemia - not hypoxia/acidosis. The intervening baseline and variability should be assessed to judge central-organ oxygenation.
^[1]

1. FIGO 2015 - Variable decelerations (V-shaped)

FIGO Definition: (U-shaped and/or with reduced variability): decelerations with a gradual onset and/or a gradual return to the baseline. Gradual onset and return occurs when more than 30 seconds elapses between the beginning/end of a deceleration and its nadir. When contractions are adequately monitored, late decelerations start more than 20 seconds after the onset of a contraction, have a nadir after the acme, and a return to the baseline after the end of the contraction.
These decelerations are indicative of a chemoreceptor-mediated response to fetal hypoxemia. In the presence of a tracing with no accelerations and reduced variability, the definition of late decelerations also includes those with an amplitude of 10−15 bpm.

^[1]

1. Ayres-de-Campos D, et al. FIGO consensus guidelines on intrapartum fetal monitoring: Cardiotocography. Int J Gynecol Obstet 2015;131(1):13-24

FIGO Definition: (U-shaped and/or with reduced variability): decelerations with a gradual onset and/or a gradual return to the baseline and/or reduced variability within the deceleration. Gradual onset and return occurs when more than 30 seconds elapses between the beginning/end of a deceleration and its nadir. When contractions are adequately monitored, late decelerations start more than 20 seconds after the onset of a contraction, have a nadir after the acme, and a return to the baseline after the end of the contraction.
These decelerations are indicative of a chemoreceptor-mediated response to fetal hypoxemia. In the presence of a tracing with no accelerations and reduced variability, the definition of late decelerations also includes those with an amplitude of 10−15 bpm.

FIGO Definition: lasting more than 3 minutes. These are likely to include a chemoreceptor-mediated component and thus to indicate hypoxemia. Decelerations exceeding 5 minutes, with FHR maintained at less than 80 bpm and reduced variability within the deceleration, are frequently associated with acute fetal hypoxia/acidosis and require emergent intervention.

^[1]

1. Ayres-de-Campos D, et al. FIGO consensus guidelines on intrapartum fetal monitoring: Cardiotocography. Int J Gynecol Obstet 2015;131(1):13-24

FIGO Definition: lasting more than 3 minutes. These are likely to include a chemoreceptor-mediated component and thus to indicate hypoxemia. Decelerations exceeding 5 minutes, with FHR maintained at less than 80 bpm and reduced variability within the deceleration, are frequently associated with acute fetal hypoxia/acidosis and require emergent intervention.

Acceleration for which it is unsure whether it is spurious or a true acceleration, or this detail is not analysed.

True acceleration. Abrupt, transient increase of the FHR >=15 bpm lasting >=15 s (but <10 min), arising from a stable, normal baseline and returning to the same baseline, with normal baseline variability. It must not be part of a deceleration. Before 32 weeks, >=10 bpm / >=10 s may suffice.

^[1]

1. Al Fahdi & Chandraharan 2020 - definition of a true acceleration

Shoulders (grouping overshoot and sloping cliff) - spurious accelerations that are an integral part of a deceleration, not stand-alone accelerations from a stable baseline.
How to identify:
Shoulder of a variable deceleration: a rise immediately before and/or after a variable deceleration, of roughly equal amplitude on both sides.

Overshoot: an exaggerated, wider, abrupt rise on the recovery (ascending) limb of a variable deceleration, larger than a true acceleration.

Sloping cliff sign: a shallow, gradual rise immediately preceding a shallow deceleration, with reduced baseline variability (it masks the underlying shallow deceleration).
^[1]

1. Al Fahdi & Chandraharan 2020 - Shoulders & overshoots, Figures 4-5

Not analysed / Unsure (morphology) — a true-signal FHR segment whose morphology (baseline vs acceleration vs which deceleration) has not been annotated or is uncertain. Excluded from the morphological ground truth.

Normal variability — not pseudo-sinusoidal — ordinary normal baseline variability (bandwidth 5–25 bpm) that does not adopt a pseudo-sinusoidal shape. This is the default sub-type of normal variability, separated from the pseudo-sinusoidal pattern, which transiently mimics a sinusoidal wave in a benign context.

Pseudo-sinusoidal pattern — a pattern resembling the sinusoidal pattern but with a more jagged, “saw-tooth” appearance rather than a smooth sine wave. Its duration seldom exceeds 30 minutes and it is characterised by normal patterns before and after. It has been described after maternal analgesic administration and during fetal sucking / mouth movements. It can be difficult to distinguish from a true sinusoidal pattern, the short duration being the most useful discriminator.
The typical / non-pseudo distinction rests largely on the clinical association; pseudo-sinusoidal is not the same as a saltatory pattern (cf. issue #11).

^[1]

1. Ayres-de-Campos D, et al. FIGO consensus guidelines on intrapartum fetal monitoring: Cardiotocography 2015 — pseudo-sinusoidal pattern

Reduced variability — unsure type — a period of reduced baseline variability (bandwidth < 5 bpm) whose cause has not been settled: physiological deep sleep, pathological (hypoxic) reduction, or absent variability.

FIGO Definition: During deep sleep, variability is usually in the lower range of normality, but the bandwidth amplitude is seldom under 5 bpm. Deep sleep can last up to 50 minutes and is associated with a stable baseline, very rare accelerations, and borderline variability.

Note: This is also called Pattern A in the literature

In this category, we also put physiological reduced variability, for example for patients of sub-Saharan origin. If you are not sure just put "unsure type" instead

FIGO Definition: more than 50 minutes in baseline segments. Reduced variability can occur due to central nervous system hypoxia/acidosis and resulting decreased sympathetic and parasympathetic activity, but it can also be due to previous cerebral injury, infection, administration of central nervous system depressants or parasympathetic blockers. [...] Following an initially normal CTG, reduced variability due to hypoxia is very unlikely to occur during labor without preceding or concomitant decelerations and a rise in the baseline.

(silent pattern) Totally absent variability and reactivity, with or without decelerations and bradycardia. Absent variability and pre-terminal pattern not classified by FIGO, but constitute a fourth CTG class (pre-terminal CTG) in the STAN CTG classification system

Zigzag pattern.
An abrupt and erratic up-and-down fluctuation of the baseline FHR variability exceeding 25 bpm, irregular in both frequency and morphology, lasting at least 1 minute. Unlike the saltatory pattern it is transient and not uniform.

It occurs during labour when the intensity of the hypoxic stress increases with insufficient time at the baseline to ensure adequate gas exchange (rapidly evolving hypoxia). The mechanism is an autonomic instability, predominantly mediated by the parasympathetic nervous system.

Significance: a ZigZag pattern lasting > 2 min is associated with an ~11-fold increase in neonatal-unit admission, and > 1 min with a ~2-fold increase in neonatal acidosis. When it accompanies an increased baseline FHR without repetitive decelerations, it should raise the suspicion of fetal neuroinflammation (chorioamnionitis). A ZigZag pattern persisting > 1 min requires immediate action to improve fetal oxygenation (reduce/stop oxytocin, consider tocolysis; stop active pushing if seen with a subacute hypoxic pattern).

^[1]
[2]
[3]

1. Chandraharan E, et al. International expert consensus statement on physiological interpretation of CTG: first revision (2024). Eur J Obstet Gynecol Reprod Biol 2024;302:346-355 - ZigZag pattern (Fig. 3)
2. Jia YJ, Chandraharan E, et al. Pathophysiological interpretation of fetal heart rate tracings in clinical practice. Am J Obstet Gynecol 2023 - ZigZag pattern (Fig. 6)
3. Derme M, et al. Zigzag Fetal Heart Rate Pattern in an Uncomplicated Pregnancy with Dual Intrauterine Infection Detected During Labor with Intact Membranes: A Case Report. Healthcare 2025;13(14):1726

Saltatory pattern (FIGO "increased variability").
A uniform increase of the baseline bandwidth exceeding 25 bpm that lasts for more than 30 minutes.

A true saltatory pattern (> 25 bpm for > 30 min) is rare (< 5%) during labour: ongoing uterine contractions intermittently interrupt fetal oxygenation, so a uniform increase of the bandwidth usually cannot be sustained for 30 min. When present, it is mostly seen antenatally, reflecting recovery from an acute and profound (non-fatal) hypoxic-ischaemic insult, and is attributed to autonomic CNS instability.
During labour, an erratic (non-uniform) fluctuation of > 25 bpm should instead be classified as a Zigzag pattern.
^[1]
[2]
[3]

1. Ayres-de-Campos D, et al. FIGO consensus guidelines on intrapartum fetal monitoring: Cardiotocography. Int J Gynecol Obstet 2015;131(1):13-24 - Increased variability (saltatory pattern), Fig. 6
2. Chandraharan E, et al. International expert consensus statement on physiological interpretation of CTG: first revision (2024) - saltatory vs ZigZag
3. Jia YJ, Chandraharan E, et al. Pathophysiological interpretation of fetal heart rate tracings in clinical practice. Am J Obstet Gynecol 2023 - saltatory pattern

Sinusoidal pattern — a regular, smooth, sine-wave-like oscillation of the baseline FHR, amplitude about 5–15 bpm (a broad, lower-amplitude definition down to ~2–5 bpm may be used when the clinical context fits), frequency 2–5 cycles per minute (classically 3–5), lasting at least 10 minutes, with absent baseline variability and no accelerations.
Pathophysiology: classically severe fetal anaemia (anti-D alloimmunisation, feto-maternal haemorrhage, twin-to-twin transfusion, ruptured vasa praevia); also described in acute fetal hypoxia, infection, cardiac malformation, hydrocephalus and gastroschisis.

This annotation currently groups both sinusoidal sub-types — they are not separated for the challenge, for lack of data (sub-annotation remains possible when the information is available, cf. issue #11):
Typical (smooth, rounded, symmetrical sine wave; Graça type I): seen in fetal anaemia, may occur antenatally.


Atypical — jagged / saw-tooth, the “Poole Shark Teeth” pattern (Graça type II): secondary to sudden feto-maternal haemorrhage, almost always intrapartum; usually mandates immediate delivery.


Distinguish from the pseudo-sinusoidal pattern (jagged, short-lived < 30 min, normal patterns before/after, benign clinical context) and from the saltatory pattern.
^{[1]
[2]
[3]
[4]}

1. Pinas A, Chandraharan E. Continuous cardiotocography during labour: Analysis, classification and management. Best Pract Res Clin Obstet Gynaecol 2016;30:33-47 — typical vs atypical (Poole Shark Teeth) sinusoidal
2. Graça LM, et al. An approach to interpretation and classification of sinusoidal fetal heart rate patterns. Eur J Obstet Gynecol Reprod Biol 1988 — SHR type I (smooth) vs type II (sawtooth)
3. Savirón-Cornudella R, et al. Diagnosis of cardiotocographic sinusoidal patterns by spectral analyses. Eur J Obstet Gynecol Reprod Biol 2024 — typical / atypical (saw / shark teeth) sinusoidal
4. Ayres-de-Campos D, et al. FIGO consensus guidelines on intrapartum fetal monitoring: Cardiotocography 2015 — sinusoidal pattern

Not analysed / Unsure (variability) — a baseline segment whose variability has not been annotated or cannot be classified with confidence. Excluded from the variability ground truth.

Force in window of interest — mark a segment so it is forced into the window of interest: it is kept as part of the data made available to the artificial-intelligence model even if not annotated.

Not in window of interest — a portion of the recording that lies outside the window of interest and will not be used for AI training / evaluation.

Clear — an eraser: applied over a period it removes any existing annotation there, leaving the trace blank / unannotated (it does not itself create a labelled period).

Window of interest (WOI) — the portion of the recording that is used for AI training / evaluation. By default it is computed automatically around the annotated events (with a few minutes of margin); it can be overridden with Force in WOI.