Tsx 034a

To ensure the accuracy of the manual voxel count method, CT scans were performed on six ex vivo formalin-fixed cadaveric kidneys (Alexion, TSX-034A, Canon medical system Europe B.V., Zoetermeer, Netherlands). Imaging protocols were 120 kVp, 150 mAs, 512 × 512 matrix, and 1 rotation time with a 1 mm slice thickness. In total, 367 axial slices from 6 ex vivo kidneys were collected. The numbers of image slices segmented in each kidney are 66, 63, 46, 47, 85, and 60. The kidney volumes were calculated using the manual voxel count method from the CT images and compared with the actual volume which was measured using the water displacement method. The values obtained using the two methods were compared.

A total of 167 pre-contrast CT scans (instruments used were as follows: Alexion, TSX-034A, Canon Medical System Europe B.V. and Zoetermeer, Netherlands; Revolution ACT, GE Healthcare, Milwaukee, WI, USA; and Brivo CT385, GE Healthcare, Milwaukee, WI, USA) of 167 dogs were randomly collected from multiple centers. Among the 167 pre-contrast CT scans, 34,655 transverseimages from 76 CT scans included kidney calculi, and were used for training and validation. The imaging protocols were as follows; 120 kVp, 150 mAs, 512 × 512 matrix and 0.75 rotation time (Alexion); 120 kVp, 84 mAs, 512 × 512 matrix, and 1 rotation time (Revolution ACT); and 120 kVp, 69 mAs, 512 × 512 matrix, and 1 rotation time (Brivo CT385). The slice thickness of the CT scans included in the study varied from 0.75 mm to 2.5 mm. Postcontrast CT scans were not included in the present study.
The precontrast CT images included in this study were divided into training and validation data at a ratio of 80 to 20. Therefore, a total of 61 CT scans were randomly chosen as the training data and 15 CT scans were used as the validation data. The CT scans of dogs without medical records were excluded from the study. In addition, scans with motion artifacts, without volume information, or with an axis smaller than a certain size were excluded.