P1‐351: Aβ Related Memory Decline and Hippocampal Volume Loss in Mild Cognitive Impairment

tivity and specificity for each of the monocular 3D-perception cut-offs to discriminate amyloid-positive from amyloid-negative MCI patients. A logistic regression model was calculated with behavioral performance on each 3D-shape test as predictor and amyloid status as dependent variable. Results: Age and MMSE did not differ between the amyloid-negative and amyloid-positive MCI patients. Among the 3D-shape tests, only 3D-from-texture scores differed significantly between groups (pcorrected1⁄40.008; 0.40 6 0.02 (amyloid-negative); 0.88 6 0.21 (amyloid-positive)). Based on the binary cut-off for 3D-from-texture, 17 patients were classified concordantly with amyloid status (5/9 amyloid-positive, 12/12 amyloidnegative). 3D-from-motion and 3D-from-shading classified only 12 out of 21 patients concordantly with amyloid status. The logistic regression model for 3D-from-texture (chi(1)1⁄4 8.4, p1⁄40.004) classified 81% of patients correctly. 3D-frommotion and 3D-from-shading could not significantly discriminate between patients. Conclusions: In MCI, a computerized monocular 3D-shape test for texture showed a high specificity for amyloid-positivity, be it with limited sensitivity. The network for 3D-from-texture processing is already affected even in the prodromal stage of AD.