Comparison of no-prior and soft-prior regularization in biomedical microwave imaging.
Comparison of no-prior and soft-prior regularization in biomedical microwave imaging.
复制标题
DOI：
10.4103/0971-6203.83482
复制
发表时间：
2011-07
影响因子：
0.9
通讯作者：
Paulsen KD
中科院分区：
文献类型：
--
作者：
Golnabi AH;Meaney PM;Geimer SD;Paulsen KD
用于医学应用的微波成像是有吸引力的，因为不同软组织的介电性质的范围可以是相当大的。由于正常组织和恶性组织之间的对比，乳腺癌的检测和治疗反应的监测是该技术可能很重要的领域。不幸的是，该技术不能实现在组织中的深度处的高空间分辨率，该高空间分辨率可从诸如X射线计算机断层扫描（CT）或磁共振成像（MRI）的其他常规模态获得。我们在微波重建算法中采用了软先验正则化策略，并将其与传统的无先验（Levenberg-Marquardt）正则化获得的图像进行了比较。初始模拟和体模结果表明，恢复的电性能的显着改善。具体而言，微波特性估计的误差改善了95%。一个假夹杂物区域的影响也进行了评估，结果表明，一个小的剩余属性偏差的6%的介电常数和15%的电导率可以发生，否则不会降低属性恢复准确性的夹杂物，实际存在的。这项工作为将来将微波成像与MR集成以提高乳腺的分辨率和功能成像奠定了基础。
Microwave imaging for medical applications is attractive because the range of dielectric properties of different soft tissues can be substantial. Breast cancer detection and monitoring of treatment response are areas where this technology could be important because of the contrast between normal and malignant tissue. Unfortunately, the technique is unable to achieve the high spatial resolution at depth in tissue which is available from other conventional modalities such as x-ray computed tomography (CT) or magnetic resonance imaging (MRI). We have incorporated a soft-prior regularization strategy within our microwave reconstruction algorithm and compared it with the images obtained with traditional no-prior (Levenberg-Marquardt) regularization. Initial simulation and phantom results show a significant improvement of the recovered electrical properties. Specifically, errors in the microwave property estimates were improved by as much as 95%. The effects of a false-inclusion region were also evaluated and the results show that a small residual property bias of 6% in permittivity and 15% in conductivity can occur that does not otherwise degrade the property recovery accuracy of inclusions that actually exist. The work sets the stage for integrating microwave imaging with MR for improved resolution and functional imaging of the breast in the future.