Volume Measurement Using Tissue Characterization of Three-Dimensional Endoscopic Ultrasonographic Images

Background and Study Aims: Although it is time-consuming to measure the volume of lesions using three-dimensional endoscopic ultrasonography (3D-EUS), the technique is suitable for tissue characterization, as it allows images of multiple areas to be obtained simultaneously in uniform conditions. The present study tested automatic volume measurement using tissue characterization based on 3D-EUS.

Materials and Methods: Nine polygonal sections of resected spleen (volume 0.66 ± 0.19 cm³) were immersed in water, and 40 radial 3D-EUS images were obtained. For tissue characterization, the methods of co-occurrence matrix and gray-level difference and discriminant analysis were used. Each spleen section was also measured using 3D-EUS. The volume of tissue identified as spleen using tissue characterization and the volume calculated on the basis of the 3D-EUS images were both compared with the actual volume measured beforehand. Measurements using tissue characterization and 3D-EUS were carried out for every third image. In three clinical cases of cancer the volume of the lesion was measured using tissue characterization and 3D-EUS.

Results: The mean volume of the nine splenic sections estimated using tissue characterization was 1.2 ± 0.41 cm³ (mean ± SD), while the mean volume estimated with 3D-EUS imaging was 1.1 ± 0.30 cm³ (mean ± SD). The volumes measured using tissue characterization were on average 13 % larger than those obtained with 3D-EUS. Linear regression analysis showed a high degree of correlation between the two sets of measurements (r = 0.97, P < 0.00005), and also showed a high correlation between the volumes obtained using tissue characterization and the actual volume (r = 0.93, P < 0.0005). However, the volumes calculated using 3D-EUS images were larger than the actual volume (61 % on average), and the volumes estimated using tissue characterization were also greater than the actual volume. The overestimation reflected the fact that measurement was only carried out in every third 3D-EUS image. In the clinical cases, the mean value for „true” tumor tissue as determined on EUS imaging represented 73 % of the volume interpreted as cancer using tissue characterization.

Conclusions: There was a good correlation between the volume measured with 3D-EUS and the volume obtained using tissue characterization. The tissue characterization volumes were only relatively slightly larger than the volumes measured using 3D-EUS, suggesting that there may be some promise for this application of tissue characterization.

References

• 1 Kanemaki N, Nakazawa S, Inui K, et al. Three-dimensional intraductal ultrasonography: preliminary results of a new technique for the diagnosis of disease of the pancreatobiliary systems.  Endoscopy. 1997;  29 726-731
• 2 Harada K, Nakazawa S, Yoshino J, et al. A study of volume measurement for digestive tract tumors by three-dimensional ultrasound system [in Japanese with English abstract].  Gastroenterol Endosc. 1997;  39 1210-1218
• 3 Yoshino J, Nakazawa S, Inui K, et al. A trial of tissue characterization by endoscopic ultrasonography.  Endoscopy. 1998;  30 (Suppl) 144
• 4 Finette S, Bleier A, Swindell W. Breast tissue classification using diagnostic ultrasound and pattern recognition techniques, 1: methods of pattern recognition.  Ultrason Imaging. 1983;  5 55-70
• 5 Loch T, Gouge J, Bertermann H. Rechnergestützte Realtime-Farb-Sonographie zur objektiven Bildanalyse.  Ultraschall Klin Prax. 1987;  2 242-244
• 6 Oosterveld BJ, Thijssen JM, Hartman PC, et al. Detection of diffuse liver disease by quantitative echography: dependence on a priori of parameters.  Ultrasound Med Biol. 1993;  19 21-25
• 7 Huynex AL, Giesen RJB, de la Rosette JJMC, et al. Analysis of ultrasonographic prostate images for the detection of prostatic carcinoma: the automated urologic diagnosis expert system.  Ultrasound Med Biol. 1994;  20 1-10
• 8 Sujina H, Swarnamani S, Suresh S. Application of artificial neural networks for the classification of liver lesions by image texture parameters.  Ultrasound Med Biol. 1996;  22 1177-1181
• 9 Raeth U, Sclaps D, Limberg B, et al. Diagnostic accuracy of computerized B-scan texture analysis and conventional ultrasonography in diffuse parenchymal and malignant liver disease.  J Clin Ultrasound. 1985;  13 87-99
• 10 Nicholas D, Nassiri DK, Garbutt P, et al. Tissue characterization from ultrasound B-scan data.  Ultrasound Med Biol. 1986;  12 135-143
• 11 Hashioto H, Saito K, Kubota E, et al. Quantitative diagnosis of endoscopic ultrasonography by texture analysis: 1st report [in Japanese with English abstract].  Gastroenterol Endosc. 1990;  32 1363-1368
• 12 Boudinet M, Laugier P, Chiche P, et al. Texture analysis of endoscopic ultrasonographic images of pancreas and peri-digestive lymph nodes.  Eur J Ultrasound. 1994;  1 251-261
• 13 Terada S. Studies on texture analysis of submucosal tumor of stomach by endoscopic ultrasonography [in Japanese with English abstract].  J Sappolo Med Coll. 1995;  64 13-27
• 14 Cortet B, Bourel P, Dubois P, et al. CT scan texture analysis of the distal radius: influence of age and menopausal status.  Rev Rhum Engl Ed. 1998;  65 109-118
• 15 Haralick RM, Shanmugam K, Dinstein I. Texture features for image characterization.  IEEE Trans Syst Man Cybern. 1973;  3 610-621
• 16 Weszka J, Dyer CR, Rosenfeld A. A comparative study of texture measures for terrain classification.  IEEE Trans Syst Man Cybern. 1976;  6 269-285
• 17 Conners RW, Harlow CA. A theoretical comparison of texture algorithms.  IEEE Trans Pattern Anal Mach Intell. 1980;  2 204-222