ACCURACY OF DIAGNOSTIC ULTRASOUND FOR DETECTION OF CYSTIC LESIONS: DETERMINATION USING RECEIVER OPERATING CHARACTERISTIC CURVE ANALYSIS OF FINDINGS IN PHANTOM STUDIES F.J. McEvou, MVB, DVR, PHD, J. KOCH, DVM, PHD, K.M. PEDERSEN, DVM, D.H. NIELSEN, DVM, PHD, P. MANTIS, DVM zyxwv Anechoic, spherical cystic lesions are important findings on ultrasound examination. In polycystic renal disease of cats, they may be the only ultrasonographic sign of disease. This study assesses the accuracy of ultrasound, as used in a veterinary setting, for the detection of cysts. Using a spherical lesion ultrasound imaging phantom and 7.5- to 8-MHz linear array transducers, images of cysts of 2- and 4-mm diameter were created at various imaging depths from 0 to 6 cm. These were digitized and given to a panel of readers for interpretation. Area under receiver operating characteristic curves were used to evaluate test accuracy. It was shown that test accuracy was different at different imaging depths and differed with machines of differing price category. It is suggested that when setting standards for quality control in ultrasound, criteria used might be better related to imaging outcome studies rather than to aspects of machine specification. zyxwvuts Veterinary Radiology zyxwvu & Ultrasound, Vol. zyxwv 44, No. 4, zyxw 2003, pp 443-449. Key words: Ultrasound Phantom, Diagnostic accuracy, ROC curve, Cyst lesion Introduction YST LESIONS are associated with a variety of pathologic C conditions. They have been reported at locations throughout the body including the liver, gastric wall, ovary, and kidney.’ Cystic lesions of the kidney may have special significance in cats, given the high incidence of polycystic renal disease in that specie^.^-^ This hereditary disease may not be readily detected by alterations in clinical signs, and ultrasound imaging is widely used for the diagnosis. Ultra- sound is used as the sole test to determine the presence or absence of cysts in screening schemes in cats used for breeding.4-6 The ability of an ultrasound imaging system to display cystic lesions with sufficient clarity to allow their detection by experienced observers is critical. It is reasonable to sup- pose that there is a threshold size below which the detection of cysts using ultrasound becomes impossible or at least unreliable. The possibility of false positives and false nega- tives has been acknowledged, and the lack of objective data to quantify the accuracy of ultrasound for cyst detection has been re~ognized.~ From the Royal Veterinary and Agricultural University, Department of Clinical Studies, Copenhagen, Denmark (McEvoy, Koch, Pedersen, Nielsen) and Royal Veterinary College, Department of Veterinary Clinical Sciences, University of London, UK (Mantis). Address correspondence and reprint requests to Dr. Fintan McEvoy, Ridebanevej 12, DK 1870 Frederiksberg C, Denmark. Received June 19, 2002; accepted for publication November 25, 2002. Some organizations have set criteria relating to equip- ment intended for use in screening schemes for the detection of cysts in polycystic kidney disease7 These criteria com- monly relate to transducer specification; for example, the use of a linear array transducer of 7.5 MHz or greater for the detection of renal cysts to a lower limit of 2-mm diameter. The selection of transducer type in published surveys of this disease is variable, depending on the investigators. Some, for example, used a 7.5-MHz sector scanner438 while others’ recommend the use of linear array transducers. The physical principals underlying ultrasound imaging would support the concept of setting minimum transducer frequency requirements, given that transducer frequency has an influence on axial and lateral resolution of the trans- ducer. However, it is also true to say that the ability to detect a lesion will be influenced by other machine parameters including signal-to-noise ratio and its resolution in the other dimension (i.e., elevational resolution). lo It could be hypothesized that since the signal-to-noise ratio and image resolution (other than axial) will differ at different depths in the ultrasound beam, one might expect that the diagnostic accuracy of the imaging system will vary with the depth at which the lesion is situated. Further, as signal-to-noise ratio will differ in different systems (nor- mally greater in more expensive units), one might expect again that the ability to detect lesions will vary with the ultrasound unit used in spite of the use of a linear 7.5-MHz transducer in each patient. If these proposals prove to be valid, one can suggest that when defining minimum stan- 443