内窥式近红外频域漫射层析成像图像重构算法

Image Reconstruction Method for Endoscopic Near-Infrared Frequency-Domain Diffuse Optical Tomography

应用内窥式近红外频域漫射层析成像技术可望对早期宫颈癌进行诊断.研究了同时利用频域测量值中幅值和相位信息的光学参数图像重构算法.为解决计算时间长的问题,雅可比矩阵采用伴随源法和修正的广义脉冲谱技术构建,迭代更新因子采用广义最小余量Krylov方法求解.为验证所发展的算法,对吸收系数和约化散射系数分别进行了重构.在无限和有限2种吸收对比度情况下,验证目标深度对目标定位精度的影响.重构结果表明,所发展的算法具有较好的重构精度,吸收系数和约化散射系数的保真度可达80%;并且结果未表现出定位精度受目标深度影响.所发展的算法可用于内窥式近红外频域漫射层析成像.

The endoscopic near-infrared frequency-domain diffuse optical tomography is hopefully to be applied to the diagnosis of early cervical cancer. A reconstruction method using both the measured amplitude and phase informa- tion on frequency domain was developed. To reduce the computing time, the Jacobian matrix was built by combining the adjoint method with the modified Generalized Pulse Spectrum Technique. The Generalized Minimal Residual Kry- lov method was used for obtaining the iterative update factor. The absorption coefficient （μa ） and the reduced scatter- ing coefficient （μs ） were reconstructed respectively for evaluating the developed algorithm. The influence of the depth of target on the accuracy of target location was verified on both infinite and finite μa contrast levels with respect to the background μa value. The results show that μa and μs can be reconstructed with high accuracy and the fidelity can reach up to 80%. Furthermore, the reconstruction results show that the determination of target location has not been affected by the target depth. The developed algorithm is feasible in endoscopic near-infrared frequency-domain diffuse optical tomography.