Breast microcalcifications are a potential indicator of cancerous tumors. Current visualization methods are either uncomfortable or impractical. Impedance measurement studies have been performed, but not in a clinical setting due to a low sensitivity and specificity. We are hoping to overcome this challenge with the development of a highly accurate impedance probe on a biopsy needle. With this technique, microcalcifications and the surrounding tissue could be differentiated in an efficient and comfortable manner than current techniques for biopsy procedures. We have developed and tested a functioning prototype for a biopsy needle using bioimpedance sensors to detect microcalcifications in the human body. In the final prototype a waveform generator sends a sin wave at a relatively low frequency(<1KHz) into the pre-amplifier, which both stabilizes and amplifies the signal. A modified howland bridge is then used to achieve a steady AC current through the electrodes. The voltage difference across the electrodes is then used to calculate the impedance being experienced between the electrodes. In our testing, the microcalcifications we are looking for have a noticeably higher impedance than the surrounding breast tissue, this spike in impedance is used to signal the presence of the calcifications, which are then sampled for examination by radiology.

ABSTRACT BACKGROUND AND PURPOSE: Sinonasal inverted papilloma (IP) can harbor squamous cell carcinoma (SCC). Consequently, differentiating these tumors is important. The objective of this study was to determine if MRI-based texture analysis can differentiate SCC from IP and provide supplementary information to the radiologist. MATERIALS AND METHODS: Adult patients who had IP or SCC resected were eligible (coexistent IP and SCC were excluded). Inclusion required tumor size greater than 1.5 cm and a pre-operative MRI with axial T1, axial T2, and axial T1 post-contrast sequences. Five well- established texture analysis algorithms were applied to an ROI from the largest tumor cross- section. For a training dataset, machine-learning algorithms were used to identify the most accurate model, and performance was also evaluated in a validation dataset. Based on three separate blinded reviews of the ROI, isolated tumor, and entire images, two neuroradiologists predicted tumor type in consensus. RESULTS: The IP and SCC cohorts were matched for age and gender, while SCC tumor volume was larger (p=0.001). The best classification model achieved similar accuracies for training (17 SCC, 16 IP) and validation (7 SCC, 6 IP) datasets of 90.9% and 84.6% respectively (p=0.537). The machine-learning accuracy for the entire cohort (89.1%) was better than that of the neuroradiologist ROI review (56.5%, p=0.0004) but not significantly different from the neuroradiologist review of the tumors (73.9%, p=0.060) or entire images (87.0%, p=0.748). CONCLUSION: MRI-based texture analysis has potential to differentiate SCC from IP and may provide incremental information to the neuroradiologist, particularly for small or heterogeneous tumors.