Ei UENO¹, Eriko TOHNO², Kouichi ITOH³

¹Institute of Clinical Medicine, University of Tsukuba, ²Department of Radiology, University of Tsukuba Hospital, ³Department of Clinical Pathology, Jichi Medical School

キーワード : Breast neoplasms, Ultrasound diagnosis, Multivariate statistical analysis

　The authors divided echographic images of breast disease into two categories; non-tumor image forming type and tumor image forming type. The tumor image forming type was further divided into attenuation type, intermediate type, and accentuation type, based on the degree of ultrasonic attenuation. The diagnostic criteria for each type of ultrasonic image was defined using Hayashi's multivariate statistical analysis.
　The breasts of a total of 2,165 symptom-bearing patients were examined using ultrasound. Pathologic confirmation was available on 562 cases, including 135 carcinomas.
　The sensitivity of the multivariate statistical analysis and that of the actual clinical sonographic diagnosis, 89.1% and 89.9% respectively, were quite close, while the specificity of the actual diagnosis exceeded that of the statistical study. The accuracy of the multivariate statistical analysis was 81.7%, while that of the actual diagnosis was 86.9%.
　The significant point regarding malignancy is the dense boundary echoes (halo) showing high intensity at the lateral side in the attenuation type, while those of the accentuation type are heterogeneous internal echoes and with a high L/T ratio. Intermediate type lesions, which are the most difficult to diagnose, showed high numeric value for malignancies of a crab-like or triangular shape, heterogeneous internal echoes and irregular borders. These classification and criteria will be able to decrease the cases of erroneous diagnosis represented by false positive results indicated by an acoustic middle shadow sign and false negative results indicated by the accentuation of posterior echoes.