ABSTRACTS PRESENTADOS AL 11th INTERNATIONAL SYMPOSIUM "RADIONUCLIDES IN NEPHRO-UROLOGY", REALIZADO EN MONTEREY, CALIFORNIA, 13-17 MAYO 2001.

12. FACTOR ANALYSIS OF DYNAMIC RENAL STUDY REVISITED: DO WE NEED USER-INDEPENDENT FUZZY REGIONS OF INTEREST.

Sámal M.(1), Bergmann H Z (2,3), Staudenherz A (4), Mostbeck A (2), Dudczak R2,(4).

(1)Charles Univ Prague, CZ, (2) Boltzmann Inst Vienna, A, (3) Biomed Eng Phys, (4) Univ Hosp AKH Víenna, A, Dept Nucl Med, Univ Hosp AKH Vienna, A.

samal@cesnet.cz

Factor analysis was introduced to nuclear medicine in 1975 and its potential to separate dynamic structures in scintigraphy demonstrated in 1980's. Poor performance with clinical data established an opinion that the method is good for laboratory but not clinical applications. Factor analysis provides estimates of overlapping 'fuzzy' ROIs (factor images) which are used to extract time-activity (factor) curves with minimum user interaction. In dynamic renal scintigraphy, the method offers a unique possibility to separate automatically the images and curves of renal parenchyma and pelvis inside a whole-kidney ROI. In spite of its past failures, the development of the method continued. The aim of our presentation is a critical assessment of its current performance based on recent developments.

In a previous work, we analyzed 169 renal studies and demonstrated the accuracy and precision of the method for the measurement of relative renal function and parenchymal transit times.

The question for the present experiment was whether the method could provide new information and help to interpret diuretic response. We re-evaluated 55 MAG3 dynamic renal studies in 43 children (0.32-14.5 yrs) before (+20) injection of a diuretic. Based on evaluation of diuretic response, renal outflow was classified as normal (41), delayed (34), partially (16), and completely (13) obstructed. Parameters (time to peak, half time ratio 20/3, etc.) of standard and fuzzy ROI time-activity curves were calculated for each class. Although most parameters correlated significantly with outflow classes, the overlaps of individual values did not allow discrimination. Better discrimination was obtained with weighted combinations of the parameters. Parenchymal curves improved discrimination between normal and pathological kidneys while the pelvic curves improved discrimination between non-obstructed and obstructed kidneys. Factor curves thus provide potentially useful additional information especially when the whole-kidney curves are non-decreasing. The principal advantage of factor to analysis, however, is not in providing new unique information but in a reproducible, accurate and automatic extraction of renal time-activity curves.