Renogram and deconvolution parameters in diagnosis of renal artery stenosis. Variants of background subtraction and analysis techniques.

AIM: Multivariate statistical methods can be used for objective analysis. The emphasis is on analysing renal function parameters together, not one at a time. The aim is to identify curve parameters useful in making predictions in kidneys with and without renal artery stenosis (RAS).

PATIENTS, METHODS: 68 patients with resistant hypertension were subjected to captopril renography with (99m)Tc-DTPA. Variants of background areas and background subtraction methods were employed. A correction was applied for loss of renal parenchyma. Parameters from time-activity curves and retention curves from deconvolution were calculated. Renal angiography established the presence or absence of RAS. Logistic regression analysis, using age- and kidney size-adjusted models, was performed to assess the capability of renography and deconvolution to differentiate between kidneys with and without RAS.

RESULTS: Discrimination between normal kidneys and RAS was achieved by deconvolution and by renography. Deconvolution was the method of first rank with a sensitivity of 87% and a specificity of 98%. For separation of RAS and kidneys with parenchymal insufficiency deconvolution was the method of first rank with a sensitivity of 80 % and specificity of 89 %, whereas renography produced poor results.

CONCLUSION: The best performance with (99m)Tc-DTPA was based on normalised background subtraction using a rectangular area between the kidneys. Deconvolution produced the most favourable results in the separation of kidneys with and without RAS. For separation of RAS and kidneys with parenchymal insufficiency conventional renography produced poor results. Conceptually, the results of a logistic regression analysis of renal function parameters may raise possibilities in the field of computer-aided diagnosis.