Cross-disorder magnetic resonance imaging can reveal shared cortical organization, but diagnostic groups are rarely distributed uniformly across acquisition protocols. This study asks whether protocol-level recruitment overlap is sufficient to explain the reported resemblance of cortical thickness and surface-area alteration patterns among schizophrenia, bipolar disorder, major depressive disorder, and autism spectrum disorder. The analytic matrix comprised 5,549 individuals allocated across 31 scanner–protocol groups. The primary case–control audit retained 29 protocols with healthy comparators, representing 5,500 individuals; the remaining two protocol rows were retained in a sensitivity calculation. Protocol-distribution concordance analysis combined normalized recruitment entropy, effective protocol number, positive-support Gini concentration, pairwise Hellinger affinity, leave-one-protocol-out deletion, and an exact \(4!\) disorder-label permutation test. Schizophrenia had the broadest recruitment support (23 protocols; effective number 14.080), whereas bipolar disorder, major depressive disorder, and autism spectrum disorder had effective protocol numbers of 5.568, 6.910, and 5.279, respectively. Pairwise recruitment affinity ranged from 0.123 for bipolar disorder–autism spectrum disorder to 0.766 for bipolar disorder–major depressive disorder. This ordering did not mirror cortical concordance. Recruitment affinity had a weak rank association with cortical-thickness similarity (\(\rho=0.257\), exact \(p=0.667\)) and a near-zero association with surface-area similarity (\(\rho=0.086\), exact \(p=0.958\)). Deleting one protocol at a time changed the thickness association to 0.029–0.600 and the surface-area association to 0.086–0.371, without reversing the central inference. Inclusion of both protocols lacking healthy comparators left both observed rank correlations unchanged. Protocol co-sampling therefore does not provide a sufficient account of the transdiagnostic cortical pattern. The result supports biological interpretation of the concordance while defining a concrete acquisition-allocation uncertainty that future individual-level analyses should estimate directly.
Cross-disorder magnetic resonance imaging can reveal shared cortical organization, but diagnostic groups are rarely distributed uniformly across acquisition protocols. This study asks whether protocol-level recruitment overlap is sufficient to explain the reported resemblance of cortical thickness and surface-area alteration patterns among schizophrenia, bipolar disorder, major depressive disorder, and autism spectrum disorder. The analytic matrix comprised 5,549 individuals allocated across 31 scanner–protocol groups. The primary case–control audit retained 29 protocols with healthy comparators, representing 5,500 individuals; the remaining two protocol rows were retained in a sensitivity calculation. Protocol-distribution concordance analysis combined normalized recruitment entropy, effective protocol number, positive-support Gini concentration, pairwise Hellinger affinity, leave-one-protocol-out deletion, and an exact \(4!\) disorder-label permutation test. Schizophrenia had the broadest recruitment support (23 protocols; effective number 14.080), whereas bipolar disorder, major depressive disorder, and autism spectrum disorder had effective protocol numbers of 5.568, 6.910, and 5.279, respectively. Pairwise recruitment affinity ranged from 0.123 for bipolar disorder–autism spectrum disorder to 0.766 for bipolar disorder–major depressive disorder. This ordering did not mirror cortical concordance. Recruitment affinity had a weak rank association with cortical-thickness similarity (\(\rho=0.257\), exact \(p=0.667\)) and a near-zero association with surface-area similarity (\(\rho=0.086\), exact \(p=0.958\)). Deleting one protocol at a time changed the thickness association to 0.029–0.600 and the surface-area association to 0.086–0.371, without reversing the central inference. Inclusion of both protocols lacking healthy comparators left both observed rank correlations unchanged. Protocol co-sampling therefore does not provide a sufficient account of the transdiagnostic cortical pattern. The result supports biological interpretation of the concordance while defining a concrete acquisition-allocation uncertainty that future individual-level analyses should estimate directly.
