3.1. Behavioral data
During the fMRI tasks, performance accuracy decreased with the number of tracked balls (visual attention load: F =17.3, P < 0.0001, repeated measures ANOVA) and was similar for cocaine and control subjects for all task conditions (). Performance accuracy was significantly lower for the more demanding condition (4-ball tracking) as compared to the less demanding condition (2-ball tracking) for cocaine subjects and for controls (P < 0.005 and 0.01, respectively; paired t-test). The load × group interaction effect on performance accuracy was not statistically significant (F = 1.11, df = 2, P = 0.34, repeated measures ANOVA). Performance accuracy during fMRI was not different for cocaine abusers with positive and negative urine results; similarly the load × urine-status interaction effect on performance accuracy was not statistically significant (F = 1.14, df=2, P > 0.12, repeated measures ANOVA). Reaction time increased with the number of tracked balls (F = 3.73, df=2, P < 0.03, repeated measures ANOVA) but was not significantly different across groups or conditions. The load × group interaction effect on reaction time was not statistically significant (F = 1.19, df=2, P = 0.31, repeated measures ANOVA). Reaction time during the three-ball tracking task was shorter for subjects that had positive-cocaine urine screening than for those that had negative-cocaine urine screening (P < 0.02, two-sample t-test). However, the effect of urine and the load × urine interaction effect on RT were not statistically significant (F = 3.54, df=2, P > 0.09, repeated measures ANOVA).
Average performance accuracy and reaction times during fMRI. Sample size: 14 cocaine abusers and 14 controls. Error bars are standard errors.
3.2. Brain activation
In both groups, the VA task activated a bilateral network ( and ) that included the prefrontal cortex (PFC) [caudal dorsal anterior cingulate [Brodmann area (BA) 32], inferior (IFG; BA47), and middle frontal (MFG; BA6 and 9) gyri], inferior (IPC; BA40) and superior (SPC; BA7) parietal cortices, the medial dorsal body of the thalamus (MDTHA), and the cerebellum, in agreement with our previous studies (Chang et al., 2004
; Tomasi et al., 2004
); the tasks deactivated the parahippocampal (PHG; BA30) and cingulate (CG; BA24) gyri, the precuneus (PreCUN; BA31), and the insula (INS; BA13), also in agreement with our previous studies (Tomasi et al., 2006b
Fig 2 Statistical maps of the average BOLD signal across conditions (2-, 3-, and 4-ball tracking tasks) for 14 cocaine abusers (upper panel), 14 control subjects (middle panel), and for the differential activation between the groups (bottom panel). White labels (more ...)
The differential activation pattern between the groups demonstrated widespread differences between cocaine abusers and controls ( and ). This included brain areas that were: (1) activated in controls but activated less or deactivated in cocaine abusers [lateral geniculate body of the thalamus (LGTHA), MDTHA, and the PreCUN BA 19], (2) activated in cocaine abusers more than in controls [MFG, lingual (LG; BA18) and fusiform (FusG; BA37) gyri], and (3) deactivated more in cocaine abusers than in controls [ACG, the left locus coeruleus (LC), and the paracentral lobule (PCL; BA6)]. The average time courses of the fMRI signals in the LG, the rostral anterior cingulate gyrus (rACG; BA32), and the PreCUN in , show the correlation between the stimulus paradigm and signals in regions that demonstrated the largest activation differences between the groups. For cocaine abusers, the correlations were positive in the visual cortex (LG, r = + 0.81, P < 0.0001) and negative in the rACG (r = −0.86, P < 0.0001) and parietal cortices specialized in visuospatial processing (PreCUN, r = −0.82, P < 0.0001). For control subjects, the correlations were not significant in the LG (r = 0.10, P = 0.53), negative in the rACG (r = −0.78, P < 0.0001) and positive in the PreCUN (r = 0.86, P < 0.0001).
Fig 3 Average BOLD signals at specific ROIs () during VA (2-, 3-, and 4-ball conditions combined) for 14 controls (gray) and 14 cocaine (black) subjects. (*) Differential effects between control and cocaine subjects were statistically significant (P (more ...)
Fig 4 Average BOLD responses (symbols) across 14 controls (green) and 14 cocaine abusers (red) exemplifying the time courses of the fMRI signals in three ROIs (27 voxels; 0.73 cc; ) and the fitted hemodynamic responses (lines) elicited by the VA task. (more ...)
3.3. Urine status
These abnormal activations were accentuated in the cocaine subjects with positive urine toxicology screens. Specifically, compared to urine negative cocaine subjects, those with positive urine screens had larger cortical [MFG, precentral gyrus (BA 4 and 6), SPC, IPC, LG] and lower sub cortical [MDTHA and pulvinar thalamus] activations (Pcorr < 0.02; corrected for multiple comparisons), and larger deactivation of the PreCUN, PCG, INS, rACG, ACG, and PCL (Pcorr < 0.001); cocaine abusers with negative urine had larger activation in the postcentral gyrus BA3 (Pcorr < 0.001) and did not have larger deactivation in any brain region compared to those with positive urine.
3.4. VA load
Increased VA load (2 balls to 4 balls) produced increased activation (but not deactivations) bilaterally in the IPC, and in the left dorsolateral PFC (IFG, and MFG) for both groups (; ). For control subjects VA-load additionally activated the posterior lobe of the left cerebellum (tonsil; Pcorr
< 0.001). Thus, there was an interaction (VA-load × group) effect on BOLD responses in the cerebellum (Pcorr
< 0.001). The VA-load × group interaction effect on BOLD responses in the FusG (BA 37), the human homologous of the monkey MT/V5 area (Born and Bradley 2005
), was statistically significant in the ROI analysis ( Right, F = 5.99, df=2, P
= 0.005; two-way repeated measures ANOVA), but the corresponding VA-load × group activation cluster did not survive the SPM2 correction for multiple comparisons (Pcorr
Fig 5 [Left] Statistical maps of VA-load activation for 14 control subjects (upper panel), 14 cocaine abusers (middle panel), and for the differential VA-load activation between the groups (load × group interaction effect on BOLD responses; bottom panel). (more ...)
3.5. Behavior vs. brain activation
For cocaine abusers but not for control subjects, the BOLD signal change from two balls to four balls in the anterior thalamus (Pcorr = 0.02) and in the LG (Pcorr = 0.04) significantly correlated with the corresponding change in performance accuracy. As depicted in , the larger the thalamic and the lower the LG activation decreases, the larger was the decrease in performance accuracy from two balls to four balls. Furthermore for cocaine and control subjects combined, longer RT during the more demanding four-ball tracking task was associated with decreased occipital activation in the LG and the FusG (R < −0.41, P <0.03).
Fig 6 Linear correlations between differential (4 balls vs. 2 balls) accuracy and BOLD responses in the brain for cocaine abusers (full circles) and controls (open circles. Solid lines are linear fits, and r is the Pearson correlation coefficient. Error bars (more ...)