Functional Imaging
DIRECTOR: Daniel S. O'Leary Ph.D.
CO-DIRECTORS: Nancy C. Andreasen, M.D., Ph.D., Richard
Hichwa, Ph.D.
1. To examine basic technical and methodical issues
involved in PET and fMR imaging of cognitive function.
2. To apply multiple injection [150]water PET and fMRI
protocols in order to map systems involved in cognitive operations or functions
such as timing, planning, sequencing, attention, language, emotion, and
memory.
3. To apply them to the study of cognition in healthy normal individuals.
4. To apply them in order to explore disturbances in cognitive
function that occur in mental illnesses that are hypothesized to have neurodevelopmental
origins (i.e., schizophrenia and autism), as well as disturbances in cognitive
function that result from focal vascular lesions.
5. To use them to explore disturbances in cognitive function
resulting from acute and chronic use of addictive, psychoactive substances
such as nicotine, marijuana and alcohol.
Background and Rationale
Understanding normal brain function and the nature of abnormalities
in brain function that characterize mental illnesses is a major goal
of research in cognitive neuroscience and neuropsychiatry. PET scanning,
particularly with tracers of short half life that permit repeated studies
in the same individual, provides a powerful opportunity to realize
this goal. The multiple injection [150]water paradigm was developed
by the Washington University group and systematically applied in a
series of pioneering studies. A variety of research studies conducted
in the MH-CRC use this powerful strategy in order to continue the mapping
of normal cognitive functions and to extend this mapping to a systematic
investigation of abnormalities in cognitive function associated with
schizophrenia, autism, and vascular brain lesions. We will also continue
to use PET imaging to explore the effects of psychoactive substances
such as nicotine, marijuana, and alcohol.
As a complement to our PET work, we have recently begun to utilize functional magnetic resonance imaging (fMRI), a new and potentially very powerful blood flow imaging technique. The term functional MRI applies to several methods, but we will use it synonymously with the so-called BOLD technique (blood oxygen level dependent contrast), as introduced by Ogawa et al in 1990. The use of deoxygenated hemoglobin as an endogenous tracer in this technique accounts for its major advantage over PET, i.e., it involves no ionizing radiation, and the same subject can therefore be scanned multiple times. Additionally, spatial resolution of fMRI is theoretically on the level of 1 mm, and registration with anatomical images is straightforward. Like all new techniques, however, there are many potential pitfalls and complications that must be systematically explored. We plan to continue our technical and methodological development of this technique and then to utilize fMRI protocols in addition to PET to explore the neural substrates of normal cognition, and to assess the effects of nicotine and neuroleptic medication in patients with schizophrenia.
The functional imaging research conducted by the MH-CRC explores a variety of substantive areas. Seven sets of projects are described below. As these projects indicate, functional imaging research at the MH-CRC focuses upon the study of normal individuals and individuals with major psychoses and related disorders. As noted above (Overview section), however, the boundaries between schizophrenia and many other major mental illneses are not clear, and we also seek to gain insights about the neural mechanisms of normal and abnormal mental phenomena by studying other patient groups that will provide informative comparison data for our studies of patients suffering from schizophrenia. These include studies of rCBF and emotion in patients with vascular lesions, and studies of language functions and performance on "theory of mind" tasks in adult autistic individuals. We also study the effects of psychoactive drugs on rCBF and cognition, because these substances are frequently abused by patients with mental illness, because of their addictive qualities in both patients and normals, and because of cognitive changes caused by these substances. Finally, as described below and in the Image Analysis and Biostatistics Cores, we have examined, and will continue to examine, a variety of methodological questions that are fundamental to interpreting PET and fMR data.
Back to top
Progress Report
When our previous application was submitted, we had
completed a series of SPECT studies and were initiating our PET effort,
having just completed some preliminary work using PET. Our functional
imaging efforts have expanded substantially during the relatively short
four-year time span since our previous submission, yielding a total
of 26 completed papers (too many to include in the appendix) describing
our findings in normal cognition and in schizophrenia. We have been
very successful in obtaining external funding for our functional imaging
studies. Four of the seven projects in this research unit are supported
by NIH-funded RO1s. Two have non-NIH external support from Scottish
Rite, the Nellie Ball trust, and the CIFRE mechanism. Only one is an
unsupported pilot project. We have initiated collaborative PET projects
with established researchers in the fields of autism (Dr. Piven), marijuana
abuse (Dr. Block), and alcoholism (Drs. Nathan and Cocco). We have
also made considerable progress in implementing fMRI studies.
Most importantly, however, our extensive experience in conducting functional imaging studies during the past four years has substantially increased our sophistication and expertise. Our work has included extensive methodological studies, careful studies of multiple facets of normal cognition (attention, memory, hedonic capacity, and language), and studies of abnormal cognition in patients suffering from schizophrenia. The PET work has been particularly generative in alerting us to the importance of the thalamus and cerebellum in normal cognition, in permitting us to visualize functionally related brain regions that may represent circuits, and in suggesting that patients with schizophrenia may manifest a cognitive dysmetria at the clinical level that reflects dysfunctional connectivity at the neural level.
Progress with Image Analysis
Techniques, Statistics, and Methodology: PET
of our work in the development of image analysis techniques for PET
data is summarized in the Image Analysis Core. Reviewers are referred to
that section for examples of our visualization tools and a detailed description
of the statistical approaches available to us. Many of the papers that
document our progress in PET image analysis, statistics, and methodology
are found in the Appendices to the Image Analysis Core and the Biostatistics
Core. A brief summary of our progress with image analysis and statistical
techniques for PET includes:
Implementation of the Montreal method for conducting
subtraction analyses in order to do between-task comparisons in normals.
Implementation of a precise method for co-registering
MR and PET data, accompanied by a user-interface that permits simultaneous
visualization of peak and t maps in three orthogonal planes, thereby permitting
more precise visualization and localization of functional activity than
occurs when limited to "glass brains" or anatomic atlases.
Comparison of the Montreal method and SPM analysis in
a single large sample.
A study of the effects of the timing of stimulus delivery
on data acquisition, permitting us to identify the appropriate time window
(typically 40 seconds) that is closely tied to bolus arrival in the brain.
Methodological studies of the effects of sample size on
findings and determination of the optimal sample size for the types of
cognitive tasks used in our studies.
Methodological studies of the effects of raising and lowering
the significance threshold in FOI-driven analyses such as the Montreal
method or SPM.
Methodological studies of the comparability of absolute
flow values vs estimates based on counts.
Implementation of methods to obtain local and normalized
flow values in ROIs in an automated a priori way, as an adjunct to exploratory
(FOI) analyses.
Development of a nonparametric method for statistical
analysis, randomization, which requires minimal statistical assumptions
and is robust to unequal variance, thereby making it ideal for between-group
comparisons (e.g., patients and controls).
Development of a statistical technique that permits the
examination of correlations in blood flow data--with level of task performance,
severity of symptoms, etc.
This work addresses many of the methodological questions raised in the review of our previous submission and provides us with a solid foundation for our future PET studies. We have a rich and flexible array of image analysis tools that permit us to visualize accurately and sensitively on co-registered MR/PET images and to measure and report quantitative data.
Progress with the Study of
Functional/Circuit Abnormalities Using PET
We have selected tasks that assess the components of cognitive
systems suspected to be disrupted in schizophrenia: e.g., attention, memory,
language. We first map the functional circuitry of these components in
normals, and subsequently we study patients. Our approach to task design
has typically emphasized the conceptual approaches of experimental cognitive
psychology, as pioneered by Posner and the Washington University group.
Because of our interest in memory, we have consulted frequently with Tulving.
The following are some of our key findings:
Studies of Normal Cognition
Short-term and long-term recognition memory for word lists
engage very similar and broadly distributed circuits, despite a large difference
in retention interval (sixty seconds vs. one week).
Novel recall and recognition memory tasks produce larger
nodes of activation than practiced tasks, suggesting that practiced tasks
are performed more efficiently and have lower blood flow/metabolic requirements.
The prefrontal cortex plays a major role in memory processes,
with the left being more active in tasks that involve encoding and the
right in retrieval.
She cerebellum is activated in a variety of cognitive
tasks, even when motor function is either well-controlled or totally absent,
suggesting that it has a cognitive function in the human brain in addition
to its motor function.
Sustained spatial attention greatly increases blood flow
in modality-specific regions that process stimuli from the attended location.
Visual attention produces more widespread cortical activation
in non-sensory brain regions than does auditory attention.
Regions of the frontal lobe and cerebellum are activated
by visual attention, even when motor activity is minimal and equivalent
in the comparison condition.
Studies of Schizophrenia
During both practiced and novel recall of complex narratives,
medication-free patients (withdrawn for three weeks) show a decrease in
blood flow in frontal, cerebellar, and thalamic regions.
During practiced and novel recall of word lists, a similar
pattern is observed.
During recall of episodic memory, a similar pattern is
observed.
During recognition memory of word lists, a similar pattern is observed.
Drug naive patients studied during random episodic silent
thought (REST) display a different pattern, with decreased flow in frontal
and other cortical regions and increases in thalamic and cerebellar regions.
During an auditory attention condition patients with schizophrenia
show a decrease in blood flow in a number of regions of the frontal lobe.
During an auditory attention condition in two separate
studies, medication-free patients show an abnormal pattern of blood flow
in temporal lobe auditory cortices.
During an auditory/visual attention study, patients had significantly
lower rCBF in right parietal lobe, in the cerebellum, and in temporal lobe
auditory cortices during auditory attention conditions and higher flow
in a large region of occipital cortex during visual attention conditions,
suggesting that they have deficits in both aspects of selective attention
i.e., the ability to enhance the processing of relevant stimulus information
and the ability to inhibit processing of irrelevant stimuli.
Schizophrenic patients tend to perform most tasks more poorly than normal volunteers, raising important questions about the design and interpretation of functional imaging studies. We refer to this as the "chicken and egg problem." If differences in blood flow are observed during poor task performance, is it because the patients are not "in the task" due to disease processes such as auditory hallucinations or avolition, or are they performing as well as possible but unable to perform at normal levels due to a primary brain abnormality?
In our initial PET study, the patient group as a whole did not perform as well as controls on the dichotic listening tasks. To assess the effects of task performance we performed separate ROI analyses in a subset of 5 patients who performed as well as controls. This smaller patient group with good performance showed the same rCBF abnormalities as the larger patient group. Thus, the patients were able to perform the task, but did so using different neural circuitry than controls. At the time that we submitted our first auditory study for publication we did not have an adequate method for direct comparison of patient and control group differences. Following development of the randomization technique, we have directly compared rCBF images of a separate sample of 12 patients and 11 controls performing dichotic listening tasks. This analysis again showed that patients had lower rCBF than controls in a large region of right STG when attending to the left ear for nonsense word stimuli. Regional CBF was also lower in patients in a smaller region of left STG in the attend right condition. Importantly, the randomization analysis revealed that patients had significantly lower rCBF in the thalamus in both attention conditions, as well as lower rCBF in the frontal lobes and thalamus. These data support the cognitive dysmetria model and are consistent with dysfunction in a cortico-cerebellar-thalamic-cortical circuit (CCTCC).
In our more recent PET studies we have used strategies other than post-hoc grouping of patients to equate performance between patients and controls. We have performed several memory studies in which we were able to equate long-term memory performance in patient and control groups by practicing subjects in both groups until performance achieved 100% accuracy. In this way we were able to compare rCBF in patient and control groups when they were performing memory tasks at equivalent accuracy levels, as well as when performing non-practiced memory tasks on which patients had poorer performance. As noted below, many of the studies proposed for the next funding period utilize a new, graded task approach in which a parameter of psychological interest is systematically varied across scans and correlations are computed between the rCBF changes and the parameter manipulated. In addition to the advantages for within-group comparisons noted below, the graded task approach also provides situations in which both controls and patients have task parameters that range from easy to difficult. This will permit conditions to be compared in which controls are performing a task at the same level of accuracy as controls.
The results of our studies of patients with schizophrenia are surprisingly convergent, in that the majority of them indicate flow abnormalities in interconnected CCTCC circuits, which are relatively independent of the task. These observations have led to our hypothesis of a fundamental cognitive deficit, cognitive dysmetria, which is a common thread that unifies the concept of schizophrenia and that may help in creating a more heuristic definition of the phenotype. Many questions remain to be answered, however, such as the significance of relative increases or decreases in flow, the relative contribution of subregions within prefrontal cortex or from other cortical regions, more refined dissection of thalamic and cerebellar contributions, and more elegant dissection of task components. Our future PET work will therefore continue to explore the extent to which CCTCC abnormalities persist across cognitive systems (e.g., memory, attention) and processes (e.g., retrieval, inhibition). In order to avoid the complications of interpretation produced by subtraction analyses (i.e., the meaning of relative increases and decreases in flow), we will expand our design strategy to include some studies involving graded stimuli. We will also follow up on our observations about the contributions of the cerebellum to cognition by conducting cognitive experiments that will dissect cerebellar functions and their anatomic and functional localizations in more detail.
Overall Research Plan
All six PET projects described below utilize the methods developed
by the Iowa MH-CRC for PET and MR imaging data acquisition and analysis.
This implicitly confers a shared orientation within the projects for
a wide variety of methodological and theoretical issues. These range
from an emphasis on the importance of individually co-registering PET
images from every activation condition with each subject's MR image
set, to issues relating to the cognitive functions that may occur in
patient and control subjects during a "resting" baseline condition.
There is also explicit interaction between investigators in the design
and implementation of the PET projects. The result is that studies
with differing patient populations address a related set of issues
concerning structure-function relationships in the normal brain and
in subjects in which normal brain function has been altered by neurodevelopmental
disorders (Dr. Piven's studies of adult autism), by vascular lesions
(Dr. Robinson's project), or by psychoactive substances (Drs. Block's
and O'Leary's studies of chronic and acute marijuana use, Dr. Cocco's
study of alcohol abuse). The fMRI effort at the MH-CRC has developed
with similar methodological and theoretical perspectives as the PET
imaging effort, and there is considerable overlap of personnel involved
in PET and fMRI imaging. We anticipate that fMRI will play a larger
role in our studies of normal cognition and schizophrenia in the next
few years. It is a major advantage to have access to both imaging technologies,
and we plan to implement a number of the PET projects described in
Project 1 for use with fMRI, and to run a subset of control and patient
subjects through both imaging procedures using the same cognitive tasks.
Image Acquisition and Analysis:
Regional cerebral blood flow (rCBF) is measured using
the bolus [15O]water method with a GE4096PLUS Scanner. The subjects
recline on the PET couch and are oriented in the PET scanner with laser
light guides (indicating the center of the lowest slice) aligned at
the orbital-meatal line. Fifteen slices (6.5mm center-to-center) are
acquired with a 10cm axial field of view. Images were acquired as 20
5-sec frames following injection of 50 millicuries of [15O]water into
a catheter in the antecubital vein of the right arm. Images are reconstructed
using a Butterworth filter (cutoff frequency = 0.35 Nyquist) for a
40 sec interval that begins after bolus transit through the brain,
determined by time-activity curves derived from an ROI placed over
a major cerebral artery. Arterial blood is sampled from a catheter
placed in left radial artery to allow calculation of tissue perfusion
in milliliters per minute per 100 grams (mL/min/100g) of tissue using
the autoradiographic method. Injections are repeated at approximately
15 minute intervals. The quantitative PET blood flow images are transferred
to the Image Processing Laboratory of the University of Iowa Mental
Health Clinical Research Center for analysis using locally-developed
software. The MR images are contiguous 1.5 mm thick coronal MR slices
acquired on a 1.5 Tesla General Electric Signa scanner using an SPGR
sequence with flip angle of 40 degrees, TE of 5 ms, TR of 24 ms, with
2 excitations. The outlines of the brain are placed on the MR images
using a neural network followed by hand editing to separate the brain
from surrounding CSF. The outlines of the PET brains are automatically
identified with an edge detection algorithm. Each individual's PET
images for each injection are registered with MR images of their brain
using software that initially performs a least-squares minimization
to fit the surface of the PET and MR images. The output parameters
of this surface registration are then used as input parameters for
a variance minimization program . Brain landmarks identified on the
co-registered MR are then used to place each brain into a standardized
coordinate space. This image normalization currently uses a linear "bounding-box" approach,
but we are developing multi-dimensional morphing techniques. An 18
mm Hanning filter is typically applied to the PET images for each condition
to eliminate residual anatomical variability, although we also sometimes
utilize non-Hanning filtered images. (Further details concerning acquisition
and analysis methods may be found in the Image Analysis Core Unit.)
© The University of Iowa 2005. All rights reserved.
Latest update May 7, 2005 Webpages maintained by Hans J. Johnson. E-mail the webmaster
MB