Abstract
Two hundred and fifty-six (256) patients (72% preoperative, 28% pre-Mammotome) were prospectively examined with EIS using the TS 2000 (TransScan Research and Development Center, Israel; temporarily distributed by Siemens, Erlangen) with the “LOS”-software (level of suspicion). All exams were performed with the targeted scan probe, the observer knowing all clinical and imaging facts. The area of the lesions was examined with EIS at least with 5 single scans. The evaluation included a scaling of lesions from 1 (surely benign) up to 5 (highly suggestive for malignancy) as well as the additional notification of spots. Results of EIS were based upon the automatic scaling which is provided by the software and were compared with mammography and histology. Furthermore the influence of the histology, size of lesions, and presence/absence of spots on the EIS results were analyzed.
Histology revealed benign results in 138 lesions and malignant results in 118 lesions (DCIS=61, ID-Ca=51, IL-Ca=5, mucinous Ca=1). Mammography as expected yielded high values with 91% sensitivity and 62% specificity. Overall sensitivity of EIS was 75.4%, specificity 42.03%, negative predictive value 66.7% and positive predictive value 52.7% (89 TP, 58 TN, 80 FP, 29 FN). EIS was false negative in 20 ID-Ca, 3 IL-Ca, 1 IDL-Ca, 4 DCIS, and 1 mucinous carcinoma. Sensitivity and specificity of EIS did not differ for the different histological differentiations neither for the degree of invasion. Also the additional notification of “spots” didn't show a correlation to malignancy. There were significant differences of the sensitivity of EIS regarding the tumor size. While EIS correctly diagnosed 85% of lesions <10 mm in size, only 64% of lesions >10 mm were detected. Most frequent lesion types for false positives were mastopathy (55/80 FP) and fibroadenoma (21/80 FP). Patient acceptance of EIS was perfect and there were no drop outs because of movement artifacts.
In conclusion the “LOS”-software clearly improved the clinical performance of the TS 2000 as compared to the initial software. The high sensitivity of EIS in small cancers which was found in our study may indicate an advantage of this method. However, the overall sensitivity and specificity with this setup of EIS is still far too low. Further improvements especially including the measurement of higher frequencies should be realized.