Quality Control of Ultrasound Transducers using distribution-free Overlapping Coefficients

Ultrasound tomography for breast cancer imaging relies on high quantities of transducers. These should exhibit reliable performance, low variability and high production yield. This work aims for finding a suitable quality control method to identify manufacturing errors of ultrasound transducer arrays. Electromechanical impedance measurements were performed at five states during the array manufacturing process. Ten parameter were derived from each measurement at each state. The relative change of these parameters allows to evaluate their suitability for quality control. Suitable parameters must exhibit high sensitivity to changes in manufacturing, but low sensitivity to parasitic circuit elements. To quantify the separability, overlapping coefficients were calculated for each parameter between the current and the preceding manufacturing state. These coefficients reflect the intersecting area of two probability density functions which were calculated using a normal kernel function to avoid distributional assumptions. The bandwidth of the electrical input power was found the most suitable parameter for quality control. Identification of manufacturing defects is possible with a probability of 0.006 for false negative decisions. Additionally, it shows high overlapping coefficients when adding parasitic circuit elements. These results encouraged us to use the derived classifiers from easy-to-perform electromechanical impedance measurements for quality control of our ultrasonic transducer arrays.