TL;DR
This paper introduces a novel time-domain near-infrared spectroscopy (TD-NIRS) system for noninvasively monitoring fetal oxygenation levels. By leveraging time-domain information, the system can effectively separate maternal and fetal signals, overcoming limitations of existing methods and paving the way for safer, more accurate fetal health assessments.
What Problem Does This Paper Address?
Monitoring fetal oxygenation is crucial for assessing fetal well-being during pregnancy and labor. Current clinical methods, like cardiotocography (CTG), are indirect and have limited correlation with actual fetal oxygen saturation, leading to a high rate of unnecessary C-sections. Invasive techniques, such as transvaginal fetal pulse oximetry, carry risks of infection and complications. While near-infrared spectroscopy (NIRS) has shown promise, existing continuous-wave (CW-NIRS) approaches struggle with the low signal-to-background ratio from maternal interference and require calibration curves that are ethically challenging to obtain for fetuses. This paper aims to address these issues by developing a noninvasive TD-NIRS method that can accurately measure deep tissue oxygenation without the need for invasive procedures or adult calibration data.
Methods
The core of this work is a Time-Domain Near-Infrared Spectroscopy (TD-NIRS) approach designed to overcome the challenges of transabdominal fetal monitoring.
- Two-Layer Model: The system models the maternal and fetal tissues as a two-layer structure. This simplification allows for the estimation of optical properties, specifically the absorption coefficient, of the deeper fetal layer.
- Time-Domain Information: TD-NIRS measures the time it takes for light photons to travel through tissue. This "time-of-flight" information provides depth sensitivity, enabling the separation of signals originating from different tissue layers. By "time gating" – focusing on later arriving photons – the system can effectively filter out superficial maternal signals and enhance the fetal signal.
- Analytical Diffusion Equation Solutions: Instead of computationally intensive Monte Carlo simulations for forward modeling, the researchers utilize analytical solutions of the diffusion equation for layered media. This significantly speeds up the inverse fitting process, making real-time or near-real-time analysis more feasible.
- Fiber-Based TD-NIRS Prototype: A portable, fiber-based TD-NIRS system was developed. It uses a white-light laser source and spectral filters to acquire data at two wavelengths (740 nm and 850 nm), chosen for their distinct absorption characteristics of oxygenated and deoxygenated hemoglobin. The system employs Galvo mirrors to rapidly switch between wavelengths and a fiber bundle for flexible source-detector arrangements. A single-pixel SPAD (Single-Photon Avalanche Diode) was used for phantom and arm experiments, while a SPAD array was employed for the ewe experiment to allow for multichannel measurements and optimal source-detector separation selection.
- Inverse Fitting: The system uses a nonlinear fitting algorithm (Levenberg-Marquardt) to estimate the optical properties (absorption and scattering coefficients) of both layers from the time-of-flight data. These estimated absorption coefficients are then used to calculate the relative concentrations of oxygenated and deoxygenated hemoglobin, ultimately determining the tissue oxygenation index (TOI).
Datasets
The paper does not explicitly mention training datasets. Instead, it relies on:
- Previously Reported Optical Properties: The optical properties for maternal and fetal tissue used in simulations and phantom experiments were based on values reported in existing literature.
- Simulated Data: Monte Carlo simulations were used to validate the TD-NIRS method and demonstrate its advantages in separating signals and improving accuracy.
- Phantom Experiments: Two-layer tissue phantoms with precisely determined optical properties were used to characterize the prototype system's performance in estimating deep-layer absorption coefficients.
- Human Forearm Emulation: An on-arm occlusion experiment using a tissue phantom over a human forearm was conducted to emulate fetal ischemia beneath a maternal layer.
- In Vivo Ewe Experiment: A pregnant ewe model was used for preclinical validation, with fetal oxygenation levels controlled via umbilical cord occlusion and reference measurements obtained from arterial blood gas (ABG) analysis.
Key Insights and Ideas
- Time Gating is Key for Depth: The ability to "time gate" the detected photons is crucial. By focusing on later arriving photons, the system effectively filters out the strong, superficial maternal signal, allowing the weaker, deeper fetal signal to be resolved. This is a significant advantage over CW-NIRS.
- Analytical Solutions Accelerate Modeling: Moving from computationally intensive Monte Carlo simulations to analytical solutions of the diffusion equation for layered media dramatically speeds up the forward modeling process, making the system more practical for real-time applications.
- TD-NIRS Offers Calibration-Free Potential: Unlike traditional pulse oximetry, TD-NIRS can estimate absolute absorption coefficients, which can be converted to oxygenation levels without requiring specific calibration curves for different populations (like fetuses). This bypasses a major ethical and practical hurdle.
- Two-Wavelength Advantage: Using wavelengths on either side of the isosbestic point of hemoglobin (740 nm and 850 nm) provides a good separation of absorption spectra, leading to more accurate estimation of oxygenated and deoxygenated hemoglobin concentrations.
- Fiber Probe Enhances Flexibility: The portable, fiber-based probe design allows for adaptable source-detector arrangements, making the system more versatile for different measurement scenarios.
Key Results
- Improved Signal-to-Background: Monte Carlo simulations demonstrated that time gating significantly increases the fetal signal-to-background ratio compared to measurements without gating.
- More Accurate Absorption Estimation: TD-NIRS showed a lower relative error in estimating the fetal layer's absorption coefficient compared to CW-NIRS across various fetal oxygenation levels.
- Phantom Accuracy: The prototype system achieved accurate deep-layer absorption coefficient estimation in phantom experiments with a relative error as low as 4%.
- Human Emulation Success: The on-arm occlusion experiment successfully demonstrated the system's ability to detect changes in tissue oxygenation (TOI) corresponding to induced ischemia, with clear differences observed in time-of-flight transients.
- In Vivo Ewe Validation: The in vivo experiment on a pregnant ewe showed a strong correlation (linear correlation coefficient of 0.95) between the TD-NIRS measured TOI and reference ABG measurements, correctly indicating fetal oxygenation trends during induced hypoxia.
Limitations and Future Work
- System Complexity and Form Factor: The current TD-NIRS system, requiring an external laser, is described as more complex and bulkier, limiting its portability for daily monitoring and making it more suitable for bedside use.
- Maternal Layer Thickness: The accuracy of measurements might be limited by a thick maternal layer, which could lead to noisy signals and challenges in inverse fitting. The current work primarily tested on a thin maternal layer.
- Absolute TOI Bias: The in vivo ewe experiments showed a potential positive bias in absolute TOI values, possibly due to low SNR from the SPAD device and variations in fetal depth.
- Future Work:
- Multimodal Integration: Combining TD-NIRS with ultrasound (for fetal positioning) and electrocardiogram (for heart rate monitoring) could provide a more comprehensive fetal assessment.
- System Optimization: Improving light collection efficiency with more detector fibers, exploring optimal time gating strategies, and integrating light sources to reduce form factor are suggested.
- Advanced Data Processing: Leveraging multichannel measurements and exploring advanced data processing methods to utilize spatial information from multiple source-detector separations could further enhance accuracy.
- Thicker Maternal Layer Characterization: Further studies are needed to assess performance with thicker maternal layers to accommodate a wider range of patients.
Summary updated