Elvio De Fiori MD, PhD, Abstract Co-Author: Nothing to Disclose

Fausto Maffini, Abstract Co-Author: Nothing to Disclose

Nicoletta Tradati, Abstract Co-Author: Nothing to Disclose

Paola Tredici, Abstract Co-Author: Nothing to Disclose

Letizia Sirica, Abstract Co-Author: Nothing to Disclose

Luke Bonello MD, Presenter: Nothing to Disclose

Massimo Bellomi MD, Abstract Co-Author: Nothing to Disclose

HistoScanningTM is an emerging ultrasound based tissue characterization technology that is specifically aiming at detecting changes induced on the ultrasound signal by the presence of cancerous lesions in the tissue (Lucidarme O. et al. Eur Radiol March 20, 2010). The present study is the first in which HistoScanning is applied to the Thyroid gland. We report the results of matching HistoScanning prediction and whole mount step sectioned histopathology for presence or absence of cancer lesions in the Thyroid gland.

21 patients diagnosed with thyroid cancer scheduled for thyroid surgery had a standardized ultrasound examination of the thyroid. Ultrasound CINE LOOPs of RF/RAW data covering the complete thyroid gland were recorded with hand held transducer (VF13-5) using a Siemens Antares machine equipped with the Axius URI. After surgery, thyroid was fixed in formalin, cut transversally at 3 mm intervals and areas where cancer was present were spatially localized on a 5 X 5 mm grid. The corresponding ultrasounds’ cineloops were divided into training and testing sets. Only good quality ultrasound CineLoops of RF/RAW data, each representing a complete scan of a Thyroid lobe were used for this analysis. The CineLoops’ data was further divided into data sets of 300K & 165K data points for training & testing respectively. A data point corresponds to ultrasound signals originating from 0.001cc of tissue. A non linear classifier for tissue characterization was trained on a data set originating from 5 cancerous, 1 benign and 5 normal lobes. The test data set originated from 4 benign and 9 normal lobes.

The table below provides the classification error for normal and benign tissues at levels ranging from the smallest data point representing 0.001cc of tissue to “grouping together of data points representing 0.1cc of tissue. For calculation of specificity, the denominator was the total number of 0.001 cc volumes analyzed in each of the 4 benign and 9 normal lobes of the test set, and the numerator was the number of 0.001 cc volumes for which the HistoScanning analysis result was negative.

These preliminary data suggest that Thyroid HistoScanning has the potential for accurately characterize normal thyroid tissue.

HistoScanning could become a tool for planning the extent of surgery in patients with thyroid cancer.

De Fiori, E, Maffini, F, Tradati, N, Tredici, P, Sirica, L, Bonello, L, Bellomi, M, Thyroid HISTOSCANNING: A New Tissue Characterization Tool to Aid in the Management of Thyroid Cancer Patients—Preliminary Results.  Radiological Society of North America 2010 Scientific Assembly and Annual Meeting, November 28 - December 3, 2010 ,Chicago IL. http://archive.rsna.org/2010/9010997.html