Ultrasound Tissue Characterization of Breast Biopsy Specimens: Expanded Study

Abstract

Tissue classification by examining sets of ultrasound parameters is an elusive goal. We report analysis of measurements of ultrasound speed, attenuation and backscatter in the range 3 to 8 MHz in breast tissues at 37 C. Statistical discriminant analysis and neural net analysis were employed. Data were acquired from 24 biopsy and 7 mastectomy specimens. Best separation of the classes normal, benign, and malignant occurred in the 18 cases where two tissue classes were present in the same specimen and parameters were corrected for within-patient mean; then 85-90% of cases in test sets were correctly classified. Most errors comprised misclassified benign cases. The neural net was comparable to discriminant analysis and slightly superior in separating normal and malignant classes.

A Quantitative Ultrasound-Based Multi-Parameter Classifier for Breast Masses

“…Further, they demonstrated that a 2-parameter analysis (attenuation and BSC) was sufficient to separate the 3 distinct tissue types they studied. Mortensen et al (1996) used an artificial neural network and a feature set consisting of sound speed, attenuation and backscatter parameters to differentiate ex vivo breast tissue samples as either 'normal', 'benign', or 'malignant'. They achieved 0.93 accuracy, far exceeding the 0.729 found by Stavros et al (1995) for standard clinical sonography, even though their data acquisition system did not provide image guidance for region of interest (ROI) selection.…”

Differentiation of Cancerous Lesions in Excised Human Breast Specimens Using Multiband Attenuation Profiles From Ultrasonic Transmission Tomography

“…Ever since ultrasound was introduced into medical imaging, tissue characterization/differentiation has been a major goal due to the rich information that is generated by ultrasound in its interaction with soft tissues. Because of its potential for assisting clinical diagnosis, there have been extensive efforts [12][13][14][15][16][17][18] to develop computerized methods for automatic differentiation of benign from malignant lesions in the hope of improved cancer detection. These efforts have primarily used the conventional echo mode of ultrasonic systems (called B-scan), which uses backscattered information to provide a qualitative map- C ping of tissue borders and geometry based on acoustic impedance differences (which cause echoes).…”

“…Our approach is entirely different from these previous studies using echo mode B-scans [12][13][14][15][16][17][18] because it uses characteristic frequencydependent attenuation signatures (MBPs) of C individual pixels that are obtained from multiband tomographic HUTT images in the transmission mode. These MBPs comprise "relative attenuation indices" at various frequency bands obtained from broadband analysis of the received first-arrival snippet and represent measures of acoustic attenuation through each specific tissue type at various frequency bands.…”

“…Nonetheless, the performance measures reported in this study may not be directly comparable with those of previous studies because the latter were typically executed via inspection of echo signals or multiple image features in small patches of the B-scan images. 12 On the other hand, the evaluation of the tissue differentiation results in this study was based on the reference standard provided by the pathologic diagnoses and must be viewed with the caveat of the huge difference in the thickness of the tissue volume that each method probes. Specifically, the pathologic slides are only a few micrometers thick, but the "virtual slice" volume probed and characterized by HUTT multiband analysis is on the order of 1 mm (≈2 orders of magnitude difference).…”

“…The performance of the classification method was finally assessed by computing the commonly used parameters of sensitivity, specificity, and accuracy. 12 …”