研究目的
To develop a system for high-speed depth estimation of a light source embedded in a scattering medium using spatially-resolved diffuse imaging and an empirical polynomial model.
研究成果
The polynomial model provides accurate and high-speed depth estimation with an RMS error of 0.12 mm in simulations and within 2 mm in experiments up to 15 mm depth. It outperforms the diffusion model in accuracy and computational efficiency, making it suitable for real-time applications like jet injection monitoring or robotic needle steering. Future work could extend to multi-layer tissues.
研究不足
The model assumes homogeneous optical properties, which may not hold in biological tissues with layered structures. The presence of an opaque needle in experiments introduces discrepancies. Signal-to-noise ratio limits depth estimation at greater depths, especially in high-absorption media. The scaling factor between experimental and Monte Carlo data adds potential error.
1：Experimental Design and Method Selection:
The study uses a polynomial model fitted to Monte Carlo simulated reflectance data for depth estimation, compared against a diffusion model. The design involves generating a dataset via Monte Carlo simulations and experimental validation with a custom fibre optic probe.
2：Sample Selection and Data Sources:
Monte Carlo simulations are performed with 17 absorption coefficients and 17 reduced scattering coefficients over a range, using 10^7 photons per simulation. Experimental data is acquired from tissue-mimicking phantoms made of Intralipid and India Ink solutions with specified optical properties.
3：List of Experimental Equipment and Materials:
Fibre optic probe with four silica-core optical fibres (200 μm core diameter, NA=0.22) mounted in a black acetal disk; avalanche photodiodes (Hamamatsu C12703-01 and C5460-01); voice-coil motor for depth control; 808 nm laser diode; linear potentiometer; real-time controller (cRIO-9022, National Instruments); optical phantoms (Intralipid 20% and Black India Ink).
4：22) mounted in a black acetal disk; avalanche photodiodes (Hamamatsu C12703-01 and C5460-01); voice-coil motor for depth control; 808 nm laser diode; linear potentiometer; real-time controller (cRIO-9022, National Instruments); optical phantoms (Intralipid 20% and Black India Ink).
Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: The probe measures diffuse reflectance profiles at four radial positions. A calibration step normalizes detector sensitivities. The light source is positioned at controlled depths using a voice-coil motor, and reflectance data is acquired at 250 kHz sampling rate. Depth estimation is performed using the polynomial and diffusion models.
5：Data Analysis Methods:
Cross-validation (leave-one-out) is used for the polynomial model. The diffusion model employs a nonlinear least squares fitting procedure with the Levenberg–Marquardt algorithm. Errors are quantified using root mean square error.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
