PURPOSE: To evaluate the diagnostic value of positron emission tomography (PET) using fluorine-18 fluorodeoxyglucose (FDG) for the diagnosis of primary breast cancer. PATIENTS AND METHODS: Preoperatively, 144 patients with masses suggestive of breast cancer underwent PET imaging of the breast. To identify breast cancer by increased metabolic activity, parametric FDG-PET images were analyzed for increased tracer uptake applying conventional image reading (CIR) and sensitive image reading (SIR). One hundred eighty-five breast tumors were evaluated by histology, revealing 132 breast carcinomas and 53 benign masses. RESULTS: Breast carcinomas were identified with an overall sensitivity of 64.4% (CIR) and 80.3% (SIR). The increase in sensitivity (SIR) resulted in a noticeable decrease in specificity, from 94.3% (CIR) to 75.5% (SIR). At stage pT1, only 30 (68.2%) of 44 breast carcinomas were detected, compared with 57 (91.9%) of 62 at stage pT2. A higher percentage of invasive lobular carcinomas were false-negative (65.2%) compared with invasive ductal carcinomas (23.7%). Nevertheless, positive PET scans provided a high positive-predictive value (96.6%) for breast cancer. CONCLUSION: Partial volume effects and varying metabolic activity (dependent on tumor type) seem to represent the most significant limitations for the routine diagnostic application of PET. The number of invasive procedures is therefore unlikely to be significantly reduced by PET imaging in patients presenting with abnormal mammography. However, the high positive-predictive value, resulting from the increased metabolic activity of malignant tissue, may be used with carefully selected subsets of patients as well as to determine the extent of disease or to assess therapy response.
Fuzzy Statistical Unsupervised Learning Based Total Lesion Metabolic Activity Estimation in Positron Emission Tomography Images