Validating CT-assisted Region of Interest Acquisition Method in 99mTc-DTPA Scintigraphy for Patients with Low Renal Function

Abstract

Glomerular Filtration Rate (GFR) is an important merit in evaluating the renal function. A common way of GFR determination is utilizing the Gamma camera scintigraphy with 99mTc-DTPA, manually delineating the region of interest (ROI) and calculating GFR via the Gates method. However, the manual ROI delineation could be impossible due to the poor quality of the scintigraphic image for patients with low renal function. In this work we present a CT-assisted ROI acquisition method for patients with at least a severely damaged kidney, who usually undergo the scintigraphy and the CT examination separately or jointly. Nineteen patients (10 men, 9 women, age range 21-87y, mean 58.0 ± 20.2y) were involved. First of all, the scintigraphic image was summed from 2-min to 3-min after the bolus injection to form a composite image of scintigraphy (SIm), on which the ROI of the relative better kidney was delineated manually. Then a binary image (SBIm) was formed only containing the ROI of the better kidney. Secondly, slices of the CT image containing the kidney structure were summed to obtain the projection image (CIm), on which two kidney ROIs were also drawn manually. A structural binary image (CBIm) was then formed accordingly. Thirdly, the CBIm was resized and matched to the SBIm, and used as a mask to filter the SIm to produce the ROI of the poorer kidney that was not drawn on the SIm. After that, sub renal background ROIs were extracted on the SIm proportionally and automatically. Finally, the GFR was calculated using the Gates method with the kidney ROIs and the background ROIs. To evaluate our method, two clinicians with about 5-year clinical experience drew kidney ROIs manually and also employed our CT-assisted method to detect ROIs. We compared the consistence of the method of manual delineation and ours in terms of the GFR. The results showed that the CT-assisted method was more reproducible when applied to the patient data.