CAJ | 학술논문

目的：实用化电阻抗断层成像（EIT）系统的测量分辨率（MR）和信噪比（SNR）等性能通常与EIT方法学研究中的理想仿真条件相差甚远。针对MR和SNR对EIT成像的影响开展仿真研究。方法基于对理想仿真条件下相邻激励-相邻测量的边界电压测量数据，分别模拟MR为0.1 mV和0.01 mV时，SNR分别为40~80 dB的实用化EIT系统和无噪声系统，采用Tikhonov-Noser组合正则化算法对不同位置目标A、B、C成像，引入图像重建误差（ER）函数和图像结构相似度（SSIM）函数定量评价成像结果。结果在本研究条件下，MR为0.01 mV时可成像的实用化EIT系统SNR至少要40~50 dB。对目标A、B、C的高质量成像SNR分别需要达到80、70、60 dB。 MR为0.1 mV时系统SNR分别要达到60、50、40 dB方可对A、B、C成像，且提高SNR对成像质量的改善不如MR为0.01 mV时明显。结论盲目追求更高的SNR除了增加系统构建难度，对系统整体性能和成像质量改善的意义将不明显。建议对MR为0.01 mV和0.1 mV的系统，与之相匹配的系统SNR应分别选择为50~60 dB和60~70 dB。
목적：실용화전조항단층성상（EIT）계통적측량분변솔（MR）화신조비（SNR）등성능통상여EIT방법학연구중적이상방진조건상차심원。침대MR화SNR대EIT성상적영향개전방진연구。방법기우대이상방진조건하상린격려-상린측량적변계전압측량수거，분별모의MR위0.1 mV화0.01 mV시，SNR분별위40~80 dB적실용화EIT계통화무조성계통，채용Tikhonov-Noser조합정칙화산법대불동위치목표A、B、C성상，인입도상중건오차（ER）함수화도상결구상사도（SSIM）함수정량평개성상결과。결과재본연구조건하，MR위0.01 mV시가성상적실용화EIT계통SNR지소요40~50 dB。대목표A、B、C적고질량성상SNR분별수요체도80、70、60 dB。 MR위0.1 mV시계통SNR분별요체도60、50、40 dB방가대A、B、C성상，차제고SNR대성상질량적개선불여MR위0.01 mV시명현。결론맹목추구경고적SNR제료증가계통구건난도，대계통정체성능화성상질량개선적의의장불명현。건의대MR위0.01 mV화0.1 mV적계통，여지상필배적계통SNR응분별선택위50~60 dB화60~70 dB。
Objective Practical electrical impedance tomography (EIT) system with certain measurement resolution (MR) and signal-to-noise ratio (SNR) differs a lot from ideal simulation conditions in EIT methodology research. The aim of this paper was to study the impact of practical system with different MR and SNR on EIT. Methods Based on the ideal simulation boundary voltages of adjacent excitation and adjacent measurement pattern, practical systems whose MR were 0.1 mV and 0.01 mV, SNR ranged from 40-80 dB and no noises were simulated, simulation study for three imaging models A, B and C with different positions in the to-be imaged field under practical system conditions above was carried out using the combined Tikhonov-NOSER regularization algorithm, and error of reconstruction (ER) function and structure similarity (SSIM) function were adopted for quantitative evaluation of image effect. Results There are differences between images obtained under different MR and SNR for different image models. In order to obtain images for three models, SNR of system should be at least 40-50 dB when MR was 0.01 mV. For the goal to obtain high quality images of A, B and C, SNR should be 80, 70 and 60 dB respectively. When MR was 0.1 mV, SNR for obtaining images of A, B, C were 60, 50 and 40 dB. Moreover, the improvement of images with increasing SNR under MR of 0.1 mV was not as obvious as that of MR at 0.01 mV. Conclusions Blind pursuit of single high SNR is of no help to improve system performance and image reconstruction effect except increasing the difficulties in hardware design. The extremity SNR is 50-60 dB for system with MR of 0.01 mV and 60-70 dB for system with MR of 0.1 mV.