Differentiation of cancerous lesions in excised human breast specimens using multiband attenuation profiles from ultrasonic transmission tomography

J Ultrasound Med. 2008 Mar;27(3):435-51. doi: 10.7863/jum.2008.27.3.435.


Objective: This study examines the tissue differentiation capability of the recently developed high-resolution ultrasonic transmission tomography (HUTT) system in the context of differentiating between benign and malignant tissue types in mastectomy specimens.

Methods: Eight mastectomy patients provided breast specimens with benign and malignant lesions. The specimens were scanned by the HUTT system with a pair of either 8- or 4-MHz transducers. Multiband HUTT images over the frequency range from 2 to 10 MHz provide characteristic profiles of frequency-dependent attenuation, termed "multiband profiles," at individual pixels. These features are classified through a novel algorithm of "segment-wise classification" that identifies the disjoint segments of various tissue types and subsequently classifies them into respective diagnostic categories using a measure of proximity to the respective multiband profile templates that have been previously obtained from reference data.

Results: We preformed intraspecimen and interspecimen analyses of 108 slices from 8 mastectomy specimens for which "ground truth" was provided by pathology reports. The average performance indices for 2-way classification (malignant versus nonmalignant tissue) in these intraspecimen (interspecimen) specimen studies were found to be sensitivity of 81.9% (89.6%), specificity of 92.9% (92.1%), and accuracy of 89.2% (89.4%), whereas the indices for the 3-way classification were moderately lower.

Conclusions: The results have shown the potential of the HUTT technology for reliable differentiation of cancerous lesions from benign changes and normal tissue in mastectomy specimens using frequency-dependent ultrasound attenuation profiles.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Algorithms
  • Artificial Intelligence
  • Breast Neoplasms / diagnostic imaging*
  • Breast Neoplasms / pathology
  • Diagnosis, Differential
  • Female
  • Humans
  • Image Enhancement / methods*
  • Image Interpretation, Computer-Assisted / methods*
  • In Vitro Techniques
  • Mastectomy
  • Pattern Recognition, Automated
  • Phantoms, Imaging
  • Sensitivity and Specificity
  • Ultrasonography / methods*