Abdominal PPG for Fetal Heart Rate Monitoring: Non-Invasive Prenatal Surveillance
**Fetal heart rate (FHR) monitoring is a cornerstone of prenatal care, yet current clinical methods remain cumbersome, intermittent, or invasive.** Ph...
- Heart Rate
Fetal heart rate (FHR) monitoring is a cornerstone of prenatal care, yet current clinical methods remain cumbersome, intermittent, or invasive. Photoplethysmography (PPG) applied to the maternal abdomen offers a compelling alternative: continuous, non-invasive, and capable of extracting not only heart rate but also fetal oxygen saturation. The technical challenge is formidable -- the fetal PPG signal is buried beneath maternal cardiac pulsations, respiratory motion, and tissue absorption -- but advances in near-infrared optics, adaptive signal processing, and machine learning are bringing this approach closer to clinical viability.
This article reviews the physics, signal extraction algorithms, clinical validation studies, and remaining challenges for abdominal PPG-based fetal heart rate monitoring. For foundational context on how PPG signals are generated and processed, see our introduction to PPG technology.
Clinical Context and Motivation
Current fetal heart rate monitoring relies primarily on Doppler ultrasound cardiotocography (CTG), which has been the clinical standard since the 1960s. While CTG is effective, it has well-documented limitations. External CTG requires trained personnel to position the transducer, frequent repositioning as the fetus moves, and conductive gel that causes skin irritation during prolonged monitoring. Internal fetal scalp electrodes provide higher-quality signals but require ruptured membranes and carry infection risk. Neither approach is practical for continuous ambulatory monitoring during the antepartum period.
These limitations create a clinical gap. Approximately 3-5% of pregnancies develop fetal growth restriction, and early detection of fetal distress through continuous heart rate monitoring could improve outcomes (Alfirevic et al., 2017; doi:10.1002/14651858.CD006066.pub3). Current intermittent monitoring schedules -- typically 20-40 minutes per clinic visit -- may miss transient decelerations and abnormal variability patterns that signal developing hypoxia. A wearable, continuous fetal monitoring device based on optical sensing could address this gap.
Physics of Transabdominal Fetal PPG
The fundamental challenge of abdominal fetal PPG is photon transport through multiple tissue layers. Light emitted at the maternal abdominal surface must traverse skin, subcutaneous fat, abdominal muscle, peritoneum, the uterine wall, amniotic fluid, and fetal tissue before interacting with fetal blood. The return path through these same layers further attenuates the signal.
Wavelength Selection and Tissue Penetration
Tissue penetration depth is strongly wavelength-dependent. In the near-infrared (NIR) window between 700 nm and 1000 nm, absorption by hemoglobin, water, and lipids reaches a relative minimum, allowing light to travel deeper into tissue. Below 700 nm, hemoglobin absorption dominates. Above 1000 nm, water absorption increases sharply. For transabdominal fetal pulse oximetry, Zourabian et al. (2000; doi:10.1117/1.1289359) identified a first wavelength in the 670-720 nm range and a second in the 825-925 nm range as optimal, and showed in feasibility experiments that the fetal pulse could be discriminated from the maternal pulse.
At 890 nm, the effective attenuation coefficient in abdominal tissue is approximately 0.8-1.2 cm^-1, meaning signal intensity drops by roughly an order of magnitude per centimeter of tissue traversal. For a typical source-detector separation of 4-6 cm on the maternal abdomen, the detected fetal pulsatile signal is 10^-4 to 10^-6 of the emitted intensity, compared to 10^-2 for the maternal pulse component. This 100- to 10,000-fold signal disparity is the central technical challenge.
Source-Detector Geometry
The spatial sensitivity profile of a PPG sensor is determined by the source-detector separation. Larger separations preferentially sample deeper tissue layers because photons traveling shallow paths are less likely to reach the distant detector. Transabdominal systems therefore use separations of several centimeters to reach fetal tissue.
Multi-distance detector arrays can exploit this depth selectivity. Short-separation channels (1-2 cm) primarily capture maternal hemodynamics and superficial tissue changes, while long-separation channels (5-8 cm) contain both maternal and fetal components. Subtracting or adaptively filtering the short-separation signal from the long-separation signal provides a crude but effective first-stage maternal signal removal.
Signal Extraction Methods
Extracting the fetal heart rate from abdominal PPG recordings requires separating a weak quasi-periodic signal (fetal pulse at 110-160 BPM) from a stronger quasi-periodic interference (maternal pulse at 60-100 BPM) plus noise. Several algorithmic approaches have been investigated.
Adaptive Noise Cancellation
Adaptive filtering using a reference signal is the most widely applied technique. The maternal PPG signal, recorded from a finger or wrist sensor, serves as the reference input. An adaptive filter (LMS, NLMS, or RLS) estimates the maternal component present in the abdominal signal and subtracts it, leaving a residual that contains the fetal signal plus uncorrelated noise.
Gan et al. (2009, doi:10.1109/TBME.2009.2021578) demonstrated this approach with an 890 nm source and a detector 4 cm away on the maternal abdomen, using an adaptive filter with a maternal finger PPG reference. Across 24 recordings from six women at 37 ± 2 weeks' gestation, optical fetal heart rate correlated with ultrasound at r = 0.96, with a maximum error of 4%. Such methods work best when maternal and fetal heart rates are well separated in frequency, and degrade when they overlap.
For a detailed treatment of adaptive filtering algorithms used in PPG signal processing, see our guide on motion artifact removal and our signal processing algorithms overview.
Independent Component Analysis (ICA)
ICA assumes that the recorded abdominal PPG signals are linear mixtures of statistically independent source signals (maternal pulse, fetal pulse, respiration, noise). By recording from multiple spatially distributed sensors on the abdomen, ICA can decompose the mixed signals into independent components, one of which corresponds to the fetal cardiac signal.
Wavelet-Based Decomposition
Continuous wavelet transform (CWT) analysis provides time-frequency localization that can separate maternal and fetal cardiac components even when their frequencies are close. The scalogram (time-scale representation) shows distinct ridges corresponding to each heart rate and its harmonics. Ridge tracking algorithms follow the fetal fundamental frequency across time, providing beat-to-beat heart rate estimation.
Deep Learning Approaches
Recent work has applied convolutional neural networks (CNNs) and recurrent neural networks (RNNs) to fetal heart rate extraction from abdominal optical signals.
Transfer learning from large maternal PPG datasets has also shown promise. Models pre-trained on adult cardiac signal characterization can be fine-tuned with relatively small fetal datasets (n=50-100 recordings), partially addressing the data scarcity problem that plagues fetal PPG research.
Fetal Pulse Oximetry: Beyond Heart Rate
A unique advantage of optical monitoring over Doppler ultrasound is the potential for fetal oxygen saturation (SpO2) estimation. By using two wavelengths -- one where oxyhemoglobin and deoxyhemoglobin have different absorption coefficients (e.g., 735 nm) and one where they are similar or have a known relationship (e.g., 890 nm) -- the ratio of pulsatile signal amplitudes can be mapped to oxygen saturation using the Beer-Lambert relationship.
Transabdominal fetal pulse oximetry has been pursued by several groups. Wavelength choice matters at fetal saturations: Mannheimer et al. (1997, doi:10.1109/10.554761) showed that conventional oximeters lose accuracy below 70% saturation and that 735 and 890 nm emitters perform better than 660 and 900 nm at low saturation. The theoretical framework is identical to adult pulse oximetry, but the practical challenges are far greater. The fetal pulsatile signal must be extracted at both wavelengths simultaneously with sufficient signal-to-noise ratio to compute a reliable ratio.
Diffuse optical spectroscopy studies on pregnant women have demonstrated that fetal hemoglobin concentration and oxygenation can be estimated transabdominally (see Zourabian et al. in the references). However, the accuracy of transabdominal fetal SpO2 estimation remains limited and is not yet sufficient for clinical decision-making, where the critical threshold for fetal hypoxia is approximately 30% SpO2.
For background on how pulse oximetry calculations work, see our pulse oximetry readings guide and the algorithms section.
Clinical Validation Studies
Clinical validation of transabdominal optical fetal heart rate monitoring is still limited, and most published work consists of feasibility studies rather than large comparisons against Doppler CTG.
For context, the closest validated remote alternative is not optical. Mhajna et al. (2020; doi:10.1016/j.ajogmf.2020.100101) evaluated the Invu belt, which uses passive electrical and acoustic sensors rather than PPG, in 147 women at 32 weeks' gestation or later. Its fetal heart rate correlated with cardiotocography at r = 0.92, with 95% limits of agreement of -8.84 to 8.24 BPM.
Challenges and Limitations
Body Mass Index and Tissue Thickness
Maternal body composition is the single most important determinant of abdominal fetal PPG signal quality. Subcutaneous adipose tissue strongly scatters NIR photons, reducing the depth from which useful signal can be obtained. Studies consistently report significant performance degradation in women with BMI above 30-35. Given that obesity rates among pregnant women exceed 30% in many countries, this limitation affects a substantial fraction of the target population.
Potential mitigation strategies include using longer wavelengths (e.g., 1050-1100 nm, near the second NIR window), increasing source power while remaining within safety limits, and employing time-domain or frequency-domain diffuse optical techniques that can better discriminate depth information compared to continuous-wave PPG.
Gestational Age Dependence
Fetal PPG signal quality improves significantly as pregnancy advances. Before 28 weeks, the fetus is small and deep within the uterus, making optical access extremely difficult. Most successful studies have been conducted at 32-40 weeks gestational age. This limits the clinical utility for early third-trimester or second-trimester monitoring, precisely when early detection of growth restriction would be most valuable.
Motion and Ambulatory Use
Maternal motion creates artifacts in abdominal PPG signals through the same mechanisms that affect wrist-worn PPG -- sensor displacement, tissue deformation, and ambient light variation. Ambulatory monitoring during daily activities introduces substantial motion artifact that must be removed before fetal signal extraction. Multi-axis accelerometer references and advanced motion artifact removal algorithms are essential for any practical wearable implementation.
Fetal Position and Movement
The fetus is not stationary. Fetal position changes alter the optical path length and signal strength unpredictably. Fetal movement itself creates transient signal disruptions. Adaptive algorithms must handle these non-stationary conditions without losing track of the fetal heart rate.
Future Directions
Several technological developments could substantially improve abdominal fetal PPG performance. Silicon photomultipliers (SiPMs) and single-photon avalanche diode (SPAD) arrays offer orders-of-magnitude improvement in detector sensitivity compared to conventional photodiodes, potentially recovering fetal signals from deeper tissue layers. Time-resolved diffuse optical techniques, which measure the temporal distribution of photon arrival times, can provide depth-resolved measurements that suppress superficial maternal signal contamination.
Multi-modal sensor fusion combining PPG with abdominal fetal ECG (extracted from maternal abdominal electrodes) and passive acoustic sensing (phonocardiography) offers another path forward. Each modality has different sensitivity to gestational age, BMI, and fetal position, and their combination provides robustness that no single modality achieves alone.
Regulatory pathways for fetal monitoring devices are well-established (FDA Class II for external monitors), and several companies are developing wearable abdominal monitors that incorporate optical sensing. The transition from laboratory feasibility to clinical deployment will require large-scale validation studies (n > 1000) across diverse populations, demonstration of reliability during ambulatory use, and evidence that continuous monitoring improves clinical outcomes beyond standard CTG.
For researchers and engineers working on fetal PPG signal processing, our algorithms reference and conditions database provide additional context on the physiological parameters and clinical conditions relevant to prenatal monitoring.
Conclusion
Abdominal PPG for fetal heart rate monitoring represents a technically demanding but clinically promising application of photoplethysmography. The core signal extraction challenge -- isolating a weak fetal pulse from dominant maternal hemodynamics -- has been addressed through adaptive filtering, ICA, wavelet analysis, and deep learning, with the best systems achieving greater than 90% detection sensitivity and heart rate errors below 3 BPM compared to CTG reference. The potential for simultaneous fetal SpO2 estimation adds unique clinical value beyond what Doppler ultrasound provides. However, performance limitations related to maternal BMI, early gestational age, and ambulatory motion artifacts must be overcome before abdominal PPG can transition from research tool to routine clinical practice. Continued advances in detector technology, multi-modal sensor fusion, and machine learning signal extraction are steadily narrowing this gap.
References
- Zourabian A, Siegel A, Chance B, Ramanujan N, Rode M, Boas DA. Trans-abdominal monitoring of fetal arterial blood oxygenation using pulse oximetry. J Biomed Opt. 2000;5(4):391-405. https://doi.org/10.1117/1.1289359
- Mhajna M, Schwartz N, Levit-Rosen L, et al. Wireless, remote solution for home fetal and maternal heart rate monitoring. Am J Obstet Gynecol MFM. 2020;2(2):100101. https://doi.org/10.1016/j.ajogmf.2020.100101
Related articles
PPG for Cardiac Monitoring: From Heart Rate to Heart Failure Detection
PPG is becoming a first-line cardiac monitoring tool. This guide covers validated applications: heart rate, HRV, atrial fibrillation, heart failure, and emerging ECG-free arrhythmia detection.
PPG vs ECG Accuracy Comparison: Heart Rate, HRV, and Arrhythmia Detection Data
PPG vs ECG accuracy compared for heart rate, HRV metrics, arrhythmia detection, and respiratory rate with published validation data and comparison tables.
PPG Heart Rate Sensor Explained: Optical Modules, Multi-LED Design, and Motion Compensation
How PPG heart rate sensors work inside wearables. Learn about multi-LED arrays, ambient light rejection, accelerometer fusion, and sensor module design.
PPG Heart Rate Measurement: Algorithms, Accuracy, and Practical Implementation
How PPG measures heart rate step by step. Learn peak detection algorithms, beat-to-beat vs averaged HR, and accuracy benchmarks for rest, exercise, and sleep.
Non-Invasive Glucose Monitoring via PPG: Current State of the Art
**Non-invasive glucose monitoring through photoplethysmography remains one of the most pursued and most elusive goals in biosensing.** Despite decades...