The quantitative evaluation of functional neuroimaging experiments: mutual information learning curves.

Learning curves are presented as an unbiased means for evaluating the performance of models for neuroimaging data analysis. The learning curve measures the predictive performance in terms of the generalization or prediction error as a function of the number of independent examples (e.g., subjects) used to determine the parameters in the model. Cross-validation resampling is used to obtain unbiased estimates of a generic multivariate Gaussian classifier, for training set sizes from 2 to 16 subjects. We apply the framework to four different activation experiments, in this case [(15)O]water data sets, although the framework is equally valid for multisubject fMRI studies. We demonstrate how the prediction error can be expressed as the mutual information between the scan and the scan label, measured in units of bits. The mutual information learning curve can be used to evaluate the impact of different methodological choices, e.g., classification label schemes, preprocessing choices. Another application for the learning curve is to examine the model performance using bias/variance considerations enabling the researcher to determine if the model performance is limited by statistical bias or variance. We furthermore present the sensitivity map as a general method for extracting activation maps from statistical models within the probabilistic framework and illustrate relationships between mutual information and pattern reproducibility as derived in the NPAIRS framework described in a companion paper.

Effects of changes in experimental design on PET studies of isometric force.

Based on single-cell recordings in primates, the relationship between neuronal activity and force magnitude is thought to be monotonic, at least for a subset of pyramidal cells in the motor cortex. Functional neuroimaging studies have also suggested a monotonic relationship between cerebral activation and force magnitude. In order to more precisely define this relationship and to characterize the activation pattern(s) associated with the modulation of static force, we studied 40 normal subjects using [(15)O]water PET and a simple visuomotor task-application of static force on a micro force sensor with the thumb and index finger of the right hand. When our experimental design did not produce the expected result (evidence of a relationship between cerebral activation and force magnitude in ten subjects), we made serial changes in the experimental protocol, including the addition of control (baseline) trials, and increased the number of subjects in an effort to increase our sensitivity to variations in force magnitude. We compared univariate and multivariate data-analytic strategies, but we relied on our multivariate results to elucidate the interaction of attentional and motor networks. We found that increasing the number of subjects from 10 to 20 resulted in an increase in statistical power and a more stable (i.e., more replicable) but qualitatively similar result, and that the inclusion of control trials in a 10-subject group did not enhance our ability to discern significant brain-behavior relationships. Our results suggest that sample sizes greater than 20 may be required to detect parametric variation in some instances and that failure to detect such variation may result from unanticipated neurobehavioral effects.

Multivariate predictive relationship between kinematic and functional activation patterns in a PET study of visuomotor learning.

Imaging studies of visuomotor learning have reported practice-related activation in brain regions mediating sensorimotor functions. However, development and testing of functional motor learning models, based on the relationship between imaging and behavioral measures, is complicated by the multidimensional nature of motoric control. In the present study, multivariate techniques were used to analyze [15O]water PET and kinematic correlates of learning in a visuomotor tracing task. Fourteen subjects traced a geometric form over a series of eight tracing trials, preceded and followed by baseline trials in which they passively viewed the geometric form. Simultaneous evaluation of multiple behavioral measures indicated that performance improvement was most strongly associated with a global performance measure and least strongly associated with measures of fine motor control. Results of three independent analytic techniques (i.e., intertrial correlation matrices, power function modeling, iterative canonical variate analysis) indicated that imaging and behavioral measures were most closely related on early learning trials. Performance improvement was associated with covarying increases in normalized activity among superior parietal, postcentral gyrus, and premotor regions and covarying decreases in normalized activity among cerebellar, inferior parietal, pallidal, and medial occipital regions. These findings suggest that performance improvement may be associated with increased activation in neural systems previously implicated in visually guided reaching and decreased activation in neural systems previously implicated in attentive visuospatial processing.

Neurological outcomes in late HIV infection: adverse impact of neurological impairment on survival and protective effect of antiviral therapy. AIDS Clinical Trial Group and Neurological AIDS Research Consortium study team.

OBJECTIVE: In a large multi-center clinical trial of combination reverse transcriptase inhibitors (RTIs), we assessed the impact of antiretroviral therapy on neurological function, the relationship between neurological and systemic benefit, and the prognostic value of neurological performance in late HIV-1 infection. DESIGN: Neurological evaluations incorporated in a randomized, multi-center trial of combination antiretroviral therapy. SETTING: Forty-two AIDS Clinical Trials Group sites and seven National Hemophilia Foundation sites. PATIENTS: Adult HIV-infected patients (n = 1313) with CD4 counts < 50 x 10(6) cells/l. INTERVENTIONS: Four combinations of reverse transcriptase inhibitors consisting of zidovudine (ZDV), alternating monthly with didanosine (ddl), or in combination with zalcitabine (ddC), ddl or ddl and nevirapine. MAIN OUTCOME MEASURES: Mean change from baseline of a four-item quantitative neurological performance battery score, the QNPZ-4, administered to 1031 subjects. RESULTS: Triple therapy and ZDV/ddl combination preserved or improved neurological performance over time compared with the alternating ZDV/ddl and ZDV/ddC regimens (P < 0.001), paralleling their impact on survival in the same trial as previously reported. QNPZ-4 scores were predictive of survival (P < 0.001), after adjusting for CD4 counts and HIV-1 plasma RNA concentrations. CONCLUSIONS: Combination antiretroviral therapy can have a salutary effect on preserving or improving neurological function. Superior systemic treatments may likewise better preserve neurological function. The significant association of poor neurological performance with mortality, independent of CD4 counts and HIV-1 RNA levels indicates that neurological dysfunction is an important cause or a strong marker of poor prognosis in late HIV-1 infection. This study demonstrates the value of adjunctive neurological measures in large therapeutic trials of late HIV-1 infection.

Are brain functions really additive?

Although Positron Emission Tomography (PET) and functional magnetic resonance imaging (fMRI) studies commonly subtract data obtained during two or more experimental conditions to decompose a complex task, there have been few opportunities to evaluate this approach directly. In the present study, PET was used to study three motor speech tasks selected such that two were constituent components of the third, making possible a direct examination of decomposition by subtraction. In Experiment 1, a group of 13 right-handed normal volunteers participated in three activation studies: syllable repetition; phonation; and repetitive lip closure. A scanning session was devoted to a single task, repeated four times. In Experiment 2, six of the original subjects performed the same three activation studies during a single scanning session. Whether tasks were studied in separate scanning sessions or combined within a single session, the results of decomposition by compound subtraction differed significantly from the results obtained when individual tasks were compared to a simple baseline condition. These data failed to demonstrate task additivity, a necessary property if decomposition by subtraction is to provide an accurate characterization of the brain activity accompanying complex behavior.

Generalizable patterns in neuroimaging: how many principal components?

Generalization can be defined quantitatively and can be used to assess the performance of principal component analysis (PCA). The generalizability of PCA depends on the number of principal components retained in the analysis. We provide analytic and test set estimates of generalization. We show how the generalization error can be used to select the number of principal components in two analyses of functional magnetic resonance imaging activation sets.

Stable neurological function in subjects treated with 2'3'-dideoxyinosine.

AIDS Dementia Complex (ADC) is a frequent and devastating complication of HIV infection. There is evidence that zidovudine (ZDV) has an effect in alleviating the symptoms of ADC, and may have a role in its prevention. It is therefore important that new antiretroviral therapies be evaluated not only for the risk of neurologic side effects, but also for their relative efficacy to ZDV in the prevention of ADC. The present study reports the effects of 2'3'-dideoxyinosine (DDI, didanosine, Videx) therapy on neuropsychological performance in the context of several large clinical trials targeting advanced systemic HIV-1 infection. Subjects treated with DDI had stable neurologic performance in quantitative tests over a 1 year period and were similar to zidovudine treated subjects.

Abnormal cerebral glucose metabolism in HIV-1 seropositive subjects with and without dementia.

UNLABELLED: This study was undertaken in order to extend our previous finding of relative basal ganglia hypermetabolism in AIDS dementia complex (ADC) and to develop clinically useful metabolic indices of CNS involvement in HIV-seropositive (HIV+) subjects. METHODS: Twenty-one HIV+ subjects (11 with AIDS) underwent FDG-PET scanning; 12 had a follow-up scan at 6 mo and 4 had a third scan at 12 mo. Forty-three age-matched heterosexual volunteers served as controls. FDG-PET scanning was performed with arterial blood sampling, and scan data were analyzed using the Scaled Subprofile Model (SSM) with principal component analysis. RESULTS: SSM/principal component analysis of the combined (HIV+ and controls) FDG-PET dataset extracted two major disease-related metabolic components: (a) a nonspecific indicator of cerebral dysfunction, which was significantly correlated with age, cerebral atrophy and ADC stage and (b) the striatum, which was heavily weighted (relatively hypermetabolic) and appeared to provide a disease-specific measure of early CNS involvement. CONCLUSION: FDG-PET scans provide quantitative measures of abnormal functional connectivity in HIV-seropositives-with or without AIDS or ADC. These measures, which are robust across centers with respect to instrumentation, scanning technique and disease severity, appear to track the progression of CNS involvement in patients with subclinical neurologic or neuropsychologic dysfunction.

Pilot Study of Didanosine in Patients with HIV Dementia.

In a pilot open-labeled study 10 subjects with AIDS dementia complex (ADC) were treated with didanosine. Only half of the subjects were able to complete the trial as a result of side effects. Five subjects exhibited improved performance on neuropsychological testing, but the mean change in performance in this small group was not statistically significant. The study suggests that this drug may have some value in ADC patients unable to tolerate other therapies, but that further study is needed to establish this firmly.

Principal component analysis and the scaled subprofile model compared to intersubject averaging and statistical parametric mapping: I. "Functional connectivity" of the human motor system studied with [15O]water PET.

Using [15O]water PET and a previously well studied motor activation task, repetitive finger-to-thumb opposition, we compared the spatial activation patterns produced by (1) global normalization and intersubject averaging of paired-image subtractions, (2) the mean differences of ANCOVA-adjusted voxels in Statistical Parametric Mapping, (3) ANCOVA-adjusted voxels followed by principal component analysis (PCA), (4) ANCOVA-adjustment of mean image volumes (mean over subjects at each time point) followed by F-masking and PCA, and (5) PCA with Scaled Subprofile Model pre- and postprocessing. All data analysis techniques identified large positive focal activations in the contralateral sensorimotor cortex and ipsilateral cerebellar cortex, with varying levels of activation in other parts of the motor system, e.g., supplementary motor area, thalamus, putamen; techniques 1-4 also produced extensive negative areas. The activation signal of interest constitutes a very small fraction of the total nonrandom signal in the original dataset, and the exact choice of data preprocessing steps together with a particular analysis procedure have a significant impact on the identification and relative levels of activated regions. The challenge for the future is to identify those preprocessing algorithms and data analysis models that reproducibly optimize the identification and quantification of higher-order sensorimotor and cognitive responses.

Zidovudine treatment of the AIDS dementia complex: results of a placebo-controlled trial. AIDS Clinical Trials Group.

The efficacy of two doses of zidovudine was examined for the treatment of the acquired immunodeficiency syndrome (AIDS) dementia complex in a randomized, double-blinded, placebo-controlled trial conducted at nine study centers. For the initial 16 weeks, 40 subjects with mild to moderate AIDS dementia complex were randomized to one of three treatment arms: 400 mg of zidovudine five times daily, 200 mg of zidovudine five times daily, or placebo five times daily. After week 16, patients initially randomized to the placebo group were rerandomized to one of the two zidovudine treatment arms. The primary efficacy end point was improvement in performance on a battery of seven neuropsychological tests; the secondary end point was improvement on a protocol neurological evaluation directed at the cardinal features of the AIDS dementia complex. For the initial 16-week period, average z scores based on the neuropsychological test battery revealed a significant improvement in the combined treatment groups compared to the placebo group; however, when the two treatment groups were compared separately to the placebo group, only the group receiving the higher zidovudine dose exhibited significant improvement. After rerandomization of the placebo patients to one of the two treatment arms at week 16, this group also showed significant improvement in the average neuropsychological z score by week 32. These results extend previous observations that indicate a therapeutic benefit of zidovudine for the treatment of AIDS dementia complex.

Correspondence between brain ERP and behavioral asymmetries in a dichotic complex tone test.

Electrophysiologic correlates of perceptual asymmetry for dichotic pitch discrimination were investigated in 20 normal subjects. Brain event-related potentials (ERPs) elicited by dichotic pairs and binaural probe tones in the Complex Tone Test (Sidtis, 1981) were recorded from homologous scalp locations over left and right hemispheres (F3, F4; C3, C4; P3, P4; O1, O2). Baseline-to-peak amplitudes were measured for N100, P200, and a late positive complex consisting of P350, P550, and slow wave. A left ear advantage (LEA) was evident in 70% of the subjects, and hemispheric asymmetries related to this behavioral asymmetry were found for P350 and P550 amplitudes to probe stimuli. Subjects with a strong LEA had greater amplitudes over the right hemisphere than the left, whereas subjects with little or no LEA showed a nonsignificant trend toward the opposite hemispheric asymmetry. Hemispheric asymmetry of these late ERPs at parietal and occipital sites was highly correlated with behavioral asymmetry. These findings suggest the utility of electrophysiological measures in assessing hemispheric asymmetries for processing complex pitch information.

The identification of affective-prosodic stimuli by left- and right-hemisphere-damaged subjects: all errors are not created equal.

Impairments in listening tasks that require subjects to match affective-prosodic speech utterances with appropriate facial expressions have been reported after both left- and right-hemisphere damage. In the present study, both left- and right-hemisphere-damaged patients were found to perform poorly compared to a nondamaged control group on a typical affective-prosodic listening task using four emotional types (happy, sad, angry, surprised). To determine if the two brain-damaged groups were exhibiting a similar pattern of performance with respect to their use of acoustic cues, the 16 stimulus utterances were analyzed acoustically, and the results were incorporated into an analysis of the errors made by the patients. A discriminant function analysis using acoustic cues alone indicated that fundamental frequency (FO) variability, mean FO, and syllable durations most successfully distinguished the four emotional sentence types. A similar analysis that incorporated the misclassifications made by the patients revealed that the left-hemisphere-damaged and right-hemisphere-damaged groups were utilizing these acoustic cues differently. The results of this and other studies suggest that rather than being lateralized to a single cerebral hemisphere in a fashion analogous to language, prosodic processes are made up of multiple skills and functions distributed across cerebral systems.

Differential effects of congenital versus acquired unilateral brain injury on dichotic listening performance: evidence for sparing and asymmetric crowding.

We assessed dichotic speech and complex-pitch discrimination in nine young patients with unilateral left-hemisphere injury and eight young patients with unilateral right-hemisphere injury incurred in the pre-perinatal (congenital) period. As in adults with acquired unilateral lesions, both congenital lesion groups demonstrated poor performance on stimuli presented to the ear contralateral to the lesion. In overall performance on speech discrimination, however, the left-hemisphere congenital lesion group performed significantly better than the acquired-lesion group did. On complex-pitch discrimination, the right-hemisphere congenital lesion group performed significantly better than did the acquired-lesion group, but both left- and right-hemisphere congenital lesion groups were significantly worse at complex-pitch discrimination than were their age- and gender-matched normal controls. These results indicate that although congenital damage produces a "lesion effect" in dichotic listening similar to that after damage acquired in adulthood, overall function is relatively spared. To the extent that complex-pitch discrimination is affected by congenital damage to either hemisphere but speech discrimination is not, the present results are consistent with an asymmetric form of crowding during reorganization after congenital unilateral brain damage.

Cerebrospinal fluid beta 2-microglobulin in patients with AIDS dementia complex: an expanded series including response to zidovudine treatment.

OBJECTIVE: To determine the relationship between cerebrospinal fluid (CSF) beta 2-microglobulin (beta 2M) and severity of AIDS dementia complex (ADC), and between CSF beta 2M and response of ADC to zidovudine. DESIGN: A prospective study. SETTING: Tertiary referral hospital. PATIENTS, PARTICIPANTS: Seventy-eight patients with varying stages of ADC were selected from a subgroup of a cohort of HIV-seropositive patients who are being studied prospectively for the neurological complications of HIV-1 infection. To enter our study, patients had to have an ADC stage of at least 0.5 (equivocal symptoms or abnormal neurological signs in the absence of functional impairment). A control group of 11 HIV-1-seropositive, neurologically normal patients was chosen randomly from the patients followed in the Multicenter AIDS Cohort Study. INTERVENTIONS: Patients were assessed neurologically and neuropsychologically and computed tomography of the brain and CSF studies were performed. MAIN OUTCOME MEASURES: Patients were staged according to severity of ADC on clinical criteria. Neuropsychological test scores were converted to an impairment score. CSF beta 2M was quantified in both serum and CSF of all patients and in 10 patients with pre- and post-zidovudine assessments. RESULTS: There was a high correlation between CSF beta 2M concentration and severity of ADC (P less than 0.0001); treatment with zidovudine significantly reduced these concentrations (P = 0.013). CSF beta 2M concentration was independent of CSF white-cell count and blood-brain barrier impairment. Other CSF changes in the same patients (including blood-brain barrier permeability to albumin, intrathecal synthesis of immunoglobulin G and HIV-1-p24-antigen levels) were less useful as objective correlates of ADC severity and response to zidovudine therapy. CONCLUSIONS: CSF beta 2M may be a valuable marker of ADC severity and response to antiviral therapy.

Pursuit eye movement dysfunction in HIV-1 seropositive individuals.

Studies of smooth pursuit eye movements were conducted in 30 ambulatory drug-free HIV-1 seropositive patients who did not yet manifest marked clinical signs of the AIDS Dementia Complex. Seropositive patients demonstrated disturbances in pursuit eye movements that were correlated with extent of immunosuppression, with impairments on neuropsychological tests of fine motor control/speed, and with independent clinical staging of the AIDS Dementia Complex. The results provide quantitative evidence that oculomotor disturbances are present in HIV-1 seropositive individuals before the manifestation of marked AIDS Dementia Complex. For this reason, and because more severe eye movement impairments have been observed in patients with AIDS, quantitative eye movement studies may provide a useful neurobehavioral procedure for characterizing and monitoring progression of CNS involvement associated with HIV-1 infection from early in its course.

The metabolic landscape of cortico-basal ganglionic degeneration: regional asymmetries studied with positron emission tomography.

Regional metabolic rate for glucose (rCMRGlc) was estimated using [18F]fluorodeoxyglucose (FDG) and positron emission tomography (PET) in five patients (four men, one woman; mean age 68; mean disease duration 2.4 years) with clinical findings consistent with the syndrome of cortico-basal ganglionic degeneration (CBGD). Left-right rCMRGlc asymmetry, (L-R)/(L + R) x 100, was calculated for 13 grey matter regions and compared with regional metabolic data from 18 normal volunteers and nine patients with asymmetrical Parkinson's disease (PD). In the CBGD group mean metabolic asymmetry values in the thalamus, inferior parietal lobule and hippocampus were greater than those measured in normal control subjects and patients with asymmetrical PD (p less than 0.02). Parietal lobe asymmetry of 5% or more was evident in all CBGD patients, whereas in PD patients and normal controls, all regional asymmetry measures were less than 5% in absolute value. Measures of frontal, parietal and hemispheric metabolic asymmetry were found to be positively correlated with asymmetries in thalamic rCMRGlc (p less than 0.05). The presence of cortico-thalamic metabolic asymmetry is consistent with the focal neuropathological changes reported in CBGD brains. Our findings suggest that metabolic asymmetries detected with FDG/PET may support a diagnosis of CBGD in life.