Optimizing Windowing of pPancreatic and Kidney Pathology on Dual Energy CT Using the Simplex Algorithm

Purpose: To determine optimal window settings for conspicuity of inflammatory pancreatic and kidney processes on 50keV low-monoenergetic images derived from dual-layer spectral detector CT images.

Methods and Materials: Fifteen patients with either clinically proven pancreatitis or pyelonephritis with foci of inflammation on DECT scans (30 patients total) were selected to serve as reference populations. 50keV low-monoenergetic images in the portal venous phase were iteratively evaluated by 6 abdominal radiologists in multiple different windows (7-350 center ; 120-580 width). Each reader graded the parenchymal hypodense lesions conspicuity on a scale of 1-5 (1-impossible, 2-poor, 3-acceptable ,4-good, 5-optimal), as well as the overall image background quality on a scale of 1-3 (1–poor, 2-acceptable, 3-optimal). Simplex algorithm, an efficient mathematical tool for solving linear programing problems to optimize system parameters was used in order to determine optimal window settings for lesion conspicuity . Three-dimensional contour maps expressing the relationship between window setting and overall reader grade were constructed for both pancreas and kidney, as well as for overall background quality.

Results: Convergence was achieved over 18 iterations over a total of 3780 reads (twenty-one windows setting x 6 readers x 30 cases). Optimal conspicuity (grade >4.5±0.04) of hypodense pancreas parenchymal lesions ranged from 100-145 center over 420-460 width, whereas hypodense pyelonephritis foci had optimal conspicuity with a center ranging 160-260 and width 420-450. Optimal visualization of background structures was optimal at a lower 50-150 center maintaining a 400-540 width. An ideal range for visualizing both pancreas and kidney lesions together were 150/440 (>4±0.18), substantially higher than manufacturer recommended settings of 60/400 (center/width).

Conclusions: 50keV low-monoenergetic images require specific window setting to achieve optimal visualization of pathology in different abdominal organs. Simplex optimization may offer a robust method to to enable better lesion conspicuity compared to default window settings potentially improving diagnostic accuracy.