西村 正利¹, 北畠 顕¹, 西岡 浩嗣¹, 島津 敬², 松山 泰三¹, 尾崎 仁¹, 佐藤 秀幸¹, 楠岡 英雄¹, 堀 正二¹, 武田 裕³, 井上 通敏³, 鎌田 武信¹

Masatoshi NISHIMURA¹, Akira KITABATAKE¹, Hirotsugu NISHIOKA¹, Takashi SHIMAZU², Taizo MATSUYAMA¹, Hitoshi OZAKI¹, Hideyuki SATO¹, Hideo KUSUOKA¹, Masatsugu HORI¹, Hiroshi TAKEDA³, Michitoshi INOUE³, Takenobu KAMADA¹

¹大阪大学医学部第一内科, ²神戸掖済会病院内科, ³大阪大学医学部附属病院医療情報部

¹The First Department of Medicine, Osaka University, Medical School, ²Kobe-Ekisaikai Hospital, ³Medical Information Science, Osaka University

キーワード : Ultrasonic tissue characterization, Quantitative texture analysis, RF signal, Myocardial infarction

We developed a new method for texture analysis of myocardial tissue properties, which is independent of the attenuation effects and the absolute signal intensity. The present method is characterized by (1) the acquistion of RF signal perpendicular to the epicardium at 100(10×10) sites of a 1.6 mm square matrix of the excised myocardium, using 4.0 MHz wide-band ultrasound, (2) the data calculation of the integrated backscatter from RF signal corresponding to myocardium thickness of 0.5 mm, (3) the reconstruction of the analysis plane from the data at the point equidistant from the transducer, and (4) the normalization of backscatter data with the mean value on the plane.
To test the feasibility of the present method, we applied it to normal canine myocardium(N), transmural infarction(TMI) produced by coronary ligation and scattered infarction(SMI) produced by microsphere injection. The gray level difference statistics(GLDs) and the run-length statistics (RLs) were computed along both circumferential(circ; and longitudinal(long) direction of the specimen, and compared with histological findings.
GLDs in SMI, i.e., entropy-circ, mean gray level difference-circ, angular second moment-long, entropy-long were significantly different from those in N or TMI. Gray level nonuniformity-circ of RLs in TMI and SMI significantly decreased compared with that in N. Thus, the combination of GLDs and RLs could differentiate N, TMI and SMI. These results indicate that the proposed texture analysis is useful in noninvasive characterization of the myocardial tissue changes.