234540S
STUDYING MS IS CHALLENGING AS ITS PATHOLOGY INVOLVES DIFFUSE INFLAMMATION, FOCAL CORTICAL AND WHITE MATTER DAMAGE AS WELL AS NEURODEGENERATIVE CHANGES, INCLUDING AXONAL DEGENERATION, SYNAPTIC LOSS AND PRIMARY NEURONAL DAMAGE, ALL OF WHICH COULD DISRUPT NEURAL NETWORKS AND AFFECT COGNITIVE PROCESSES. COGNITIVE IMPAIRMENT IS HIGHLY PREVALENT IN MS, IMPACTING BETWEEN 40-70% OF PERSONS WITH MS. INFORMATION PROCESSING SPEED AND EPISODIC MEMORY ARE THE COGNITIVE DOMAINS MOST FREQUENTLY AND SIGNIFICANTLY IMPACTED. COGNITIVE IMPAIRMENT HAS A DEVASTATING IMPACT ON EVERYDAY FUNCTIONING GIVEN THE ASSOCIATION WITH UNEMPLOYMENT, DIFFICULTY DRIVING, WORSE COMMUNITY INTEGRATION, AND DIFFICULTY MANAGING ACTIVITIES OF DAILY LIVING. COGNITIVE IMPAIRMENT WORSENS OVER TIME. IMPORTANTLY, DUE TO THE HETEROGENOUS NATURE OF MS, THE COGNITIVE DOMAIN AFFECTED, AND TYPE OF THERAPY OR TREATMENT NEEDED IS NOT UNIFORM. UNFORTUNATELY, TREATMENT OPTIONS FOR MS-RELATED COGNITIVE IMPAIRMENT ARE LIMITED AND ARE NOT TYPICALLY NEUROBIOLOGICALLY BASED OR PERSONALIZED; TO DATE, THERE ARE NO APPROVED PHARMACEUTICAL TREATMENTS FOR COGNITIVE SYMPTOMS. THIS STUDY AIMS TO IDENTIFY THE NEUROBIOLOGICAL MECHANISMS UNDERLYING COGNITIVE DECLINE IN MS AT A MICROSCOPIC AND MACROSCOPIC SCALE. WE BELIEVE THAT A TEAM SCIENCE APPROACH IS NECESSARY, AND THAT THE PROPOSED STUDY REQUIRES OVERSIGHT AND EXPERTISE FROM LEADERS IN DISTINCT AREAS OF EXPERTISE: AAcAcAEURSAA'Ac DR. SUSAN GAUTHIER (PI) WILL CO-DIRECT ALL ASPECTS OF THIS RESEARCH WITH CO-PI, INCLUDING STUDY DESIGN, PATIENT RECRUITMENT, PET AND MRI PROCESSING/STORAGE, DATA ACQUISITION, DATA MANAGEMENT, DATA ANALYSIS, DATA INTERPRETATION, PREPARING THE REPORTS, AND PRESENTING RESULTS. DR. GAUTHIER IS AN ASSOCIATE PROFESSOR IN CLINICAL NEUROLOGY AND THE DIRECTOR OF CLINICAL RESEARCH AT THE JUDITH JAFFE MULTIPLE SCLEROSIS CENTER (JJMSC) FOR WHICH SHE HAS SUCCESSFULLY DEVELOPED AN INFRASTRUCTURE FOR CLINICAL TRIALS AND LARGE COHORT OBSERVATIONAL STUDIES IN MULTIPLE SCLEROSIS (MS). SHE IS CURRENTLY LEADING AND COLLABORATING ON MULTIPLE IMAGING PROJECTS IN MS, WHICH HAS LED TO MULTIPLE PUBLICATIONS IN BOTH MRI AND PET. HER KEY FOCUS IS THE CLINICAL TRANSLATION NOVEL QUANTITATIVE IMAGING TOOLS IMPROVE OUR UNDERSTANDING OF PATHOLOGICAL MECHANISMS OF DISEASE IN MULTIPLE SCLEROSIS WITH AN OVERALL GOAL TO IMPROVE PATIENT CARE. SHE DIRECTS THE DESIGN, IMPLEMENTATION, AND ANALYSIS OF ALL THE CLINICAL IMAGING STUDIES IN MS AT THE JJMSC. DR GAUTHIER WILL FUNCTION AS THE CONTACT PI FOR THIS PROPOSAL. AAcAcAEURSAA'Ac DR. AMY KUCEYESKI (CO-PI) WILL CO-DIRECT ALL ASPECTS OF THIS RESEARCH WITH CO-PI, INCLUDING STUDY DESIGN, PET AND MRI PROCESSING/STORAGE, DATA ACQUISITION, DATA MANAGEMENT, DATA ANALYSIS, DATA INTERPRETATION, PREPARING THE REPORTS, AND PRESENTING RESULTS. DR. KUCEYESKI IS AN ASSOCIATE PROFESSOR OF MATHEMATICS AND NEUROSCIENCE AND DIRECTS THE COMPUTATIONAL CONNECTOMICS (COCO) LABORATORY WITHIN THE RADIOLOGY DEPARTMENT AT WEILL CORNELL MEDICINE AND THE DEPARTMENTS OF STATISTICS AND COMPUTATIONAL BIOLOGY AT CORNELL UNIVERSITY. DR. KUCEYESKI HAS DEDICATED HER CAREER TO UNDERSTANDING DOCUSIGN ENVELOPE ID: 3871C02A-ABED-4445-AF29-709C9A2C8631 THE BRAIN, SPECIFICALLY FOCUSED ON THE BRAINAAcAcAEURSAAcAEURZAcS CONNECTOMES, WITH THE GOAL OF IDENTIFYING NOVEL TREATMENTS FOR PATIENTS SUFFERING FROM NEUROLOGICAL DISEASE. SHE HAS EXTENSIVE EXPERIENCE IN DATA SCIENCE, MACHINE LEARNING AND THE APPLICATION OF QUANTITATIVE METHODS TO IDENTIFY ASSOCIATIONS BETWEEN NEUROIMAGING BIOMARKERS AND VARIOUS NEUROLOGICAL DISEASES. SHE HAS BEEN ANALYZING NEUROIMAGING IN MULTIPLE SCLEROSIS COHORTS OVER THE PAST DECADE AAcAcAEURSAA'Ac DR. ABHISHEK JAYWANT (CO-I) ASSISTANT PROFESSOR OF NEUROPSYCHOLOGY, DEPARTMENTS OF PSYCHIATRY & REHABILITATION MEDICINE, WEILL CORNELL MEDICINE. DR. JAYWANT WILL BE RESPONSIBLE FOR THE SELECTION OF NEUROPSYCHOLOGICAL OUTCOME MEASURES, TRAINING/OVERSEEING THE RESEARCH ASSISTANT CONDUCTING THE NEUROPSYCHOLOGICAL EVALUATIONS, ENSURING THE FIDELITY OF TEST SCORING, AND PARTICIPATING IN THE INTERPRETATION OF THESE OUTCOMES AS THEY RELATE TO STRUCTURAL MRI, FUNCTIONAL MRI, AND PET IMAGING. DR. JAYWANT HAS EXTENSIVE CLINICAL EXPERIENCE ASSESSING AND TREATING THE NEUROPSYCHOLOGICAL SEQUELAE OF VARIOUS NEUROLOGICAL DISORDERS, INCLUDING MULTIPLE SCLEROSIS. IN ADDITION TO HIS CLINICAL EXPERTISE, DR. JAYWANT HAS EXPERIENCE WITH IMAGE ANALYSIS, SPECIFICALLY HE HAS CREATED PIPELINE FOR MULTI-ECHO FMRI ANALYSIS. IN ADDITION, HE HAS COLLABORATED WITH BOTH PIS ON MULTIPLE RESEARCH PROJECTS RELATED TO MULTIPLE SCLEROSIS AND TRAUMATIC BRAIN INJURY SINCE 2018 AND PUBLISHED MULTIPLE MANUSCRIPTS WITH DR. KUCEYESKI REGARDING THE IMPACT OF WHITE MATTER HYPERINTENSITIES ON STRUCTURAL AND FUNCTIONAL CONNECTIVITY AND COGNITIVE FUNCTION IN AGING AND DEPRESSION. AAcAcAEURSAA'Ac DR. HURTADO RAAE'A'AoA IS AN ASSOCIATE PROFESSOR OF STATISTICS AT CLEVELAND STATE UNIVERSITY. SHE WILL BE PROVIDING STATISTICAL GUIDANCE AND ANALYSIS FOR ALL AIMS OF THIS PROPOSAL. B. AIMS AIM 1: TO DETERMINE THE LONGITUDINAL ASSOCIATION BETWEEN COGNITIVE DECLINE AND MULTI-MODAL, MULTI-SCALE MEASURES OF BRAIN STRUCTURE, FUNCTION AND NEURONAL INTEGRITY IN PATIENTS WITH MS. AIM 2: TO DETERMINE THE DYNAMIC RELATIONSHIP BETWEEN BRAIN STRUCTURE, FUNCTION AND NEURONAL INTEGRITY, AND HOW THIS RELATIONSHIP VARIES WITH COGNITIVE IMPAIRMENT STAGE IN PATIENTS WITH MS. AIM 3. TO BUILD A MODEL THAT CAN IDENTIFY MS PATIENTS AT RISK FOR FUTURE COGNITIVE DECLINE USING ADVANCED MRI MEASURES, AND, FURTHER, TO PROVIDE A CLINICALLY FEASIBLE ALTERNATIVE USING CLINICAL MRI. C. SIGNIFICANCE OF EXPECTED RESEARCH CONTRIBUTION THE PROPOSED STUDY IS EXPECTED TO RESULT IN A BETTER UNDERSTANDING OF THE LONGITUDINAL MACRO- AND MICRO-SCALE BRAIN MECHANISMS OF COGNITIVE RESILIENCE/DECLINE IN INDIVIDUALS WITH MS, AND, IN ADDITION, PROVIDE A TOOL THAT CAN PREDICT MS PATIENTS AT RISK FOR COGNITIVE DECLINE USING MRI-BASED MEASURES OF BRAIN ANATOMICAL AND PHYSIOLOGICAL NETWORKS. THE CURRENT CONTRADICTORY FINDINGS IN THE LITERATURE SHOWING BOTH FC INCREASES AND DECREASES WITH COGNITIVE DECLINE (AAcAcAEURSAA...AEUROEADAPTIVEAAcAcAEURSAAAA VS AAcAcAEURSAA...AEUROEADAPTIVE EXHAUSTIONAAcAcAEURSAAAA) ARE INDICATIVE OF THE CONFUSION IN THIS AREA; DEEPLYSAMPLED, LONGITUDINAL STUDIES OF THE SAME INDIVIDUALS WITH MS WILL HELP DISENTANGLE THESE RELATIONSHIPS. THE BROADER IMPACT OF THIS WORK WILL RESULT IN MORE ACCURATE, PERSONALIZED TREATMENT PLANS AND OPEN NEW HORIZONS FOR NOVEL THERAPIES TO PROMOTE COGNITIVE RESILIENCE IN MS. FOR EXAMPLE, IF IT TURNS OUT THAT FMZ-PET BPND IS POSITIVELY CORRELATED WITH FC (AS OUR PRELIMINARY DATA IN TBI SUPPORT), IMPACTING THE GABAA RECEPTORS USING PHARMACOLOGICAL AGENTS101-103 COULD POTENTIALLY IMPACT THE ADAPTIVE OR ADAPTIVE EXHAUSTION FC MECHANISMS TO SUPPORT COGNITIVE RESILIENCE. ALTERNATIVELY, UNDERSTANDING HOW FC IN SPECIFIC NETWORKS OR REGIONS RELATES TO COGNITIVE DECLINE MAY ALLOW IDENTIFICATION OF SITES THAT MAY BE RICH THERAPEUTIC TARGETS FOR NONINVASIVE BRAIN STIMULATION. PATIENTS AT RISK FOR COGNITIVE DECLINE MAY BE IDENTIFIED AND MORE AGGRESSIVE TREATMENT PLANS CREATED FOR THOSE PATIENTS. IT IS NOT ENTIRELY UNDERSTOOD HOW BRAIN STIMULATION TECHNIQUES INFLUENCE THE BRAIN, BUT, FOR EXAMPLE, INCREASES IN FC BETWEEN REGIONS HAVE BEEN SHOWN IN HIGH-FREQUENCY REPETITIVE TRANSCRANIAL MAGNETIC STIMULATION (RTMS)104. IN MS, NON-INVASIVE BRAIN STIMULATION HAS BEEN SHOWN TO MODIFY THE BRAINAAcAcAEURSAAcAEURZAcS FC NETWORKS TO MAINTAIN COGNITION105. CURRENTLY, HOWEVER, THE CHOICE OF TARGETS FOR BRAIN STIMULATION IS NOT WELL DEFINED; IT OFTEN DEPENDS ON POPULATION-LEVEL OBSERVATIONS RATHER THAN INDIVIDUAL PATIENT OR LESION CHARACTERISTICS. IN THIS WORK, REGIONS WHOSE FUNCTIONAL REORGANIZATION (EITHER INCREASES OR DECREASES) OVER TIME ARE IDENTIFIED AS BEING ASSOCIATED WITH IMPROVEMENTS IN COGNITIVE FUNCTION WOULD BE PRIME, CLINICALLY RELEVANT TARGETS FOR FUTURE CLINICAL DOCUSIGN ENVELOPE ID: 3871C02A-ABED-4445-AF29-709C9A2C8631 TRIALS OF NON-INVASIVE BRAIN STIMULATION TO SUPPORT COGNITIVE RESILIENCE IN MS. IMPORTANTLY, IF THE THREE-STAGE HYPOTHESIS OF AAcAcAEURSAA...AEUROEADAPTIVEAAcAcAEURSAAAA, AAcAcAEURSAA...AEUROEADAPTIVE EXHAUSTIONAAcAcAEURSAAAA AND AAcAcAEURSAA...AEUROEDECLINEAAcAcAEURSAAAA STAGES IS VERIFIED, THERE WILL BE INCREASED CLARITY ON WHETHER TREATMENTS SHOULD AIM TO INCREASE FC OR DECREASE FC DEPENDING ON WHERE AN INDIVIDUAL IS ALONG THIS TRAJECTORY; MAKING THE OPPOSITE CHOICE MAY IN FACT CAUSE MORE RAPID COGNITIVE DECLINE. THIS CONTRIBUTION WILL BE SIGNIFICANT BECAUSE IT IS EXPECTED TO HAVE BROAD TRANSLATIONAL IMPACT ON THE DEVELOPMENT OF MORE ACCURATE PROGNOSES AS WELL AS TARGETED, PERSONALIZED TREATMENTS FOR PATIENTS WITH MS. D. INNOVATION WE PRESENT SEVERAL INNOVATIONS IN THIS PROPOSAL. IN AIM 1 AND 2, WE WILL BE FIRST TO EXAMINE THE LONGITUDINAL TRAJECTORY OF COGNITIVE CHANGE IN MS UTILIZING A MULTI-SCALE, MULTI-MODAL IMAGING APPROACH. OUR DENSELY SAMPLED LONGITUDINAL DESIGN WHEREIN IMAGING AND COGNITION IS COLLECTED OVER 4 YEARS IS INNOVATIVE, AND NECESSARY, TO CLOSELY TRACK THE EVOLVING DYNAMICS BETWEEN IMAGING METRICS AND COGNITIVE DECLINE. OUR HYPOTHESES AND STATISTICAL APPROACH TO TEST THE DIRECTIONALITY SWITCH OCCURRING BETWEEN FC AND COGNITION ACROSS THE ADAPTIVE AND ADAPTIVE EXHAUSTION STAGES IS INNOVATIVE. IN AIM 2, OUR APPROACH RELATING MACRO-SCALE ANATOMICAL AND PHYSIOLOGICAL BRAIN NETWORKS MEASURED VIA MRI AND MICRO-SCALE MEASURES OF NEURONAL INTEGRITY MEASURED VIA PET HAS NOT BEEN DONE CROSS-SECTIONALLY IN PATIENTS WITH MS (OR IN CONTROLS, EXCEPT IN OUR PRELIMINARY DATA ABOVE), LET ALONE LONGITUDINALLY. OUR NOVEL MULTI-SCALE, MULTI-MODAL APPROACH WILL ALLOW US TO INTERROGATE THE RELATIONSHIP BETWEEN NEURONAL RECEPTOR-LEVEL CHANGE AND BRAIN ACTIVITY, WHICH WILL ALLOW, FOR THE FIRST TIME, TO DESCRIBE THE NEUROBIOLOGICAL MECHANISMS UNDERPINNING COGNITIVE CHANGE IN MS. THE TRANSLATIONAL ASPECT OF THE MULTI-MODAL APPROACH IS INNOVATIVE, IN THAT FUTURE WORK MAY EXAMINE PHARMACOLOGICAL AGENTS THAT MODULATE FC THROUGH GABAA RECEPTOR MANIPULATION AND SUPPORT COGNITIVE RESILIENCE. AIM 3 WILL PROVIDE THE FIRST ACCURATE MODEL, SUPPORTED BY THE NEUROBIOLOGICAL DATA ACQUIRED IN AIM 1 AND 2, TO PREDICT COGNITIVE DECLINE UTILIZING ONLY CLINICALLY-ACQUIRED IMAGING IN MS. OUR PROPOSED MODEL IN AIM 3 USES METRICS DERIVED FROM HUMAN CONNECTOME PROJECT (HCP) DATA TO IMPROVE CLINICAL CARE OF PATIENTS WITH NEUROLOGICAL DISEASE. THE CLINICAL TRANSLATION OF OUR PREDICTIVE MODEL PERFECTLY REPRESENTS THE CLINICAL MISSION FOR DATA USE THAT WAS HOPED FOR WHEN THE EXPENSIVE, AMBITIOUS HCP PROJECT WAS CONCEIVED. SUCCESS OF THIS PROPOSAL WILL PROVIDE A MAJOR CLINICAL INNOVATION TOWARD MORE ACCURATE PROGNOSIS AND THERAPEUTIC DEVELOPMENT OF NOVEL, NEUROBIOLOGICALLY-BASED THERAPIES TO ENCOURAGE COGNITIVE PRESERVATION IN PATIENTS WITH MS. E. APPROACH . AIM 1: TO DETERMINE THE LONGITUDINAL ASSOCIATION BETWEEN COGNITIVE DECLINE AND MULTI-MODAL, MULTI-SCALE MEASURES OF BRAIN STRUCTURE, FUNCTION AND NEURONAL INTEGRITY IN PATIENTS WITH MS. OUR MAIN HYPOTHESIS, BASED ON THE THREE-STAGE TRAJECTORY OF COGNITIVE DECLINE IN MS22, IS THAT INCREASING FC, INCREASING TO STABLE FMZ-PET AND INTACT SC/GRAY MATTER VOLUME WILL BE RELATED TO PRESERVED COGNITION IN THE AAcAcAEURSAA...AEUROEADAPTIVEAAcAcAEURSAAAA STAGE, WHILE FURTHER INCREASING FC AND DECREASING SC/FMZ-PET/GRAY MATTER VOLUME WILL BE RELATED TO WORSENING COGNITION IN THE AAcAcAEURSAA...AEUROEADAPTIVE EXHAUSTIONAAcAcAEURSAAAA STAGE. STATISTICAL MODELS WILL ASSOCIATE LONGITUDINAL CHANGE IN MRI (AIM 1.1) AND FMZ-PET (AIM 1.2) WITH LONGITUDINAL CHANGE IN COGNITION. AIM 1.1: USING LINEAR MIXED EFFECTS MODELS, WE WILL ASSESS THE RELATIONSHIP BETWEEN ANNUAL COGNITION CHANGE AND ANNUAL CHANGE IN EACH IMAGING MEASURE (FC, SC, AND ATROPHY) AND WILL TEST IF THIS RELATIONSHIP DIFFERS FOR THE THREE GROUPS (CP, CI AND HC). OF NOTE, THIS COMPARISON/MODELING APPROACH ACCOUNTS FOR VARIABILITY/CHANGE IN THE IMAGING/COGNITION MEASURES IN MATCHED HC OVER THE SAME TIME AS MS SUBGROUPS. WE WILL FIT A MODEL WITH THE ANNUAL COGNITION CHANGE AS THE OUTCOME, REGRESSED ON GROUP MEMBERSHIP (HC AS REFERENCE GROUP), ANNUAL CHANGE IN THE IMAGING MEASURE, AND THE INTERACTION BETWEEN THE TWO. AGE, DISEASE DURATION, EDUCATION, SUBSTANCE ABUSE MEASURES, BRAIN LESION VOLUME, EDSS, MENOPAUSAL STATUS (FOR WOMEN OVER 45) AND DMT FOR MS WILL BE ADJUSTED FOR IN THE MODEL. A RANDOM INTERCEPT FOR SUBJECT WILL BE INCLUDED IN THE MODEL TO ACCOUNT FOR CORRELATION IN THE DATA RESULTING FROM MULTIPLE OBSERVATIONS PER SUBJECT. TO TEST OUR HYPOTHESIS, WE WILL ASSESS WHETHER THE INTERACTION TERMS ARE SIGNIFICANT, INDICATING THAT THE RELATIONSHIP BETWEEN THE IMAGING MARKER DIFFERS FOR CP AND CI. WE WILL ALSO EXAMINE THE CONTRAST BETWEEN THE INTERACTION TERMS FOR CP AND CI. A SEPARATE MODEL WILL BE FIT FOR EACH IMAGING MEASURES AND MULTIPLE COMPARISON CORRECTION WILL BE APPLIED. OUR SPECIFIC PRIMARY HYPOTHESIS IS THAT INCREASED FC IN THE DEFAULT MODE133 AND VENTRAL ATTENTION62 NETWORKS WILL RELATE TO BETTER SDMT SCORES IN THE AAcAcAEURSAA...AEUROEADAPTIVEAAcAcAEURSAAAA STAGE (CP) AND WORSE SDMT SCORES IN THE AAcAcAEURSAA...AEUROEADAPTIVE EXHAUSTIONAAcAcAEURSAAAA STAGE (CI). FC WITHIN THE DEFAULT MODE AND VENTRAL ATTENTION NETWORKS WILL BE SUMMARIZED BY FIRST CALCULATING REGIONAL NODE STRENGTH (SUM OF ROWS IN THE FC THAT REPRESENTS HOW STRONGLY CONNECTED A GIVEN REGION IS TO THE REST OF THE NETWORK) AND DOCUSIGN ENVELOPE ID: 3871C02A-ABED-4445-AF29-709C9A2C8631 THEN TAKING THE AVERAGE NODE STRENGTH OVER REGIONS IN THE DEFAULT MODE AND VENTRAL ATTENTION NETWORKS, AS DEFINED USING THE YEO ATLAS134. SECONDARY HYPOTHESES INCLUDE: LARGER DECREASES IN VENTRAL ATTENTION AND DEFAULT MODE NETWORK SC (DEFINED AS THE AVERAGE NODE STRENGTH OF REGIONS WITHIN THESE TWO NETWORKS) AND MORE ATROPHY OF SUBCORTICAL STRUCTURES (THALAMUS AND PUTAMEN)3 AND CORTICAL REGIONS (R CINGULATE, L POST-CENTRAL GYRUS AND R MIDDLE TEMPORAL GYRUS)135 WILL BE ASSOCIATED WITH WORSENING SDMT SCORES OVER ALL STAGES. TO CLARIFY, OUR PRIMARY AND SECONDARY ANALYSES CONSIST OF 9 MODELS, ONE FOR EACH IMAGING MEASURE OF INTEREST: AVERAGE FC/SC NODE STRENGTH IN DEFAULT MODE AND VENTRAL ATTENTION NETWORKS (4 TOTAL) AND ATROPHY IN 2 SUBCORTICAL AND 3 CORTICAL REGIONS (5 TOTAL). REMAINING BRAIN REGIONS/NETWORKS AND COGNITIVE SCORES WILL BE INVESTIGATED IN AN EXPLORATORY MANNER, WITH STRICT CONTROL FOR MULTIPLE COMPARISONS. TO EXPLORE THE RELATIONSHIP BETWEEN LONGITUDINAL TRAJECTORIES OF COGNITION AND IMAGING MEASURES, WE WILL FIT 9 GENERALIZED ADDITIVE MIXED-EFFECTS MODELS136 (ONE FOR EACH IMAGING METRIC, WITH CONTROL FOR MULTIPLE COMPARISONS), WHICH ALLOW FITTING SMOOTH FUNCTIONS OF THE RELATIONSHIP BETWEEN TIME AND COGNITION IN ORDER TO MODEL NON-LINEAR TRENDS IN COGNITION. WE WILL ALSO FIT A SMOOTH FUNCTION FOR THE RELATIONSHIP BETWEEN THE 9 IMAGING METRICS AND COGNITION OVER TIME, TO ACCOUNT FOR THE NONLINEAR RELATIONSHIP BETWEEN THESE (FIG 1). INTERACTIONS WITH CHANGE IN AN INDIVIDUALSAAcAcAEURSAAcAEURZAc STAGE OVER TIME WILL BE EXPLORED IN THESE MODELS (CP-CP, CP-CI, CI-CI, HC-HC). A RANDOM INTERCEPT FOR SUBJECT WILL BE INCLUDED TO ACCOUNT FOR CORRELATION IN THE DATA. INFERENCE WILL BE MADE ON THE RELATIONSHIP BETWEEN THE 9 IMAGING MEASURES AND COGNITION OVER TIME AS WELL AS INTERACTIONS; THE SHAPE OF THE FUNCTIONAL COEFFICIENTS WILL BE EXAMINED. AIM 1.2: WE WILL USE A SIMILAR LINEAR MIXED EFFECT MODEL AS IN AIM 1.1 TO MODEL THE RELATIONSHIP BETWEEN COGNITION AND GABAA BPND IN THE SUB-STUDY OF 30 PATIENTS (15 CP AND 15 CI) AND 15 HC. THIS SUBSET OF SUBJECTS WILL HAVE 3 FMZ-PET IMAGING SESSIONS AT BASELINE, 1 AND 3 YEARS TO CAPTURE SHORT AND LONGER INTERVAL CHANGES. OUR PRIMARY HYPOTHESIS IS THAT DECREASING GABAA RECEPTOR DENSITY (BPND) WITHIN DEFAULT MODE AND VENTRAL ATTENTION NETWORKS134 (THE SAME NETWORKS OF INTEREST IN THE PRIMARY HYPOTHESES OF AIM 1.1) WILL BE ASSOCIATED WITH WORSENING SDMT SCORES. OTHER REGIONS OF THE BRAIN AND COGNITIVE SCORES WILL BE INVESTIGATED AS SECONDARY HYPOTHESES. POTENTIAL PROBLEMS AND ALTERNATIVE APPROACHES. GIVEN OUR INTEREST IN MAPPING IMAGING METRICS TO COGNITIVE TRAJECTORIES, A SINGLE CONTINUOUS PRIMARY OUTCOME MUST BE USED. SDMT IS UNIVERSALLY ACCEPTED AS THE SINGULAR TEST TO ASSESS COGNITION IN MS, BUT IT DOES NOT ENCOMPASS THE EXTENT OF COGNITIVE DISFUNCTION IN MS, SO A COMPREHENSIVE BATTERY OF TESTS WILL BE COMPLETED (BRNB AND SCWT). WE WILL PERFORM PRINCIPAL COMPONENT ANALYSIS (PCA) TO DETERMINE WHICH TESTS EXPLAIN THE MOST POPULATION VARIABILITY AND REPEAT THE PRIMARY ANALYSIS USING THESE AS OUTCOME. GIVEN THAT THE CI GROUP MAY HAVE HETEROGENOUS COGNITIVE PROFILES, WE WILL ALSO EXPLORE COGNITIVE PHENOTYPES (E.G., PROCESSING SPEED IMPAIRMENT ONLY; MEMORY IMPAIRMENT ONLY; MULTIDOMAIN IMPAIRMENT)131,137. OTHER TYPES OF FC ANALYSES MAY ALSO BE APPLIED, I.E. DYNAMIC FC138 OR BRAIN STATE ANALYSIS139. AIM 2: TO DETERMINE THE DYNAMIC RELATIONSHIP BETWEEN BRAIN STRUCTURE, FUNCTION AND NEURONAL INTEGRITY, AND HOW THIS RELATIONSHIP VARIES WITH STAGE OF COGNITIVE IMPAIRMENT IN PATIENTS WITH MS. UTILIZING THE SUBSET OF 15 CP, 15 CI AND 15 HC WITH LONGITUDINAL DMRI, FMRI AND FMZ-PET, WE WILL 1) EXAMINE THE RELATIONSHIP BETWEEN FC, SC, GM ATROPHY AND GABAA BPND OVER FOUR YEARS AND 2) WILL EXAMINE HOW CHANGES IN THE MULTIMODAL IMAGING RELATES TO THE TRAJECTORY OF COGNITIVE SCORES. IT IS OUR MAIN HYPOTHESIS THAT MULTI-MODAL, MULTI-SCALE IMAGING WILL QUANTIFY THE RELATIONSHIP BETWEEN BRAIN ACTIVITY, WHITE MATTER CONNECTIVITY AND NEURONAL INTEGRITY, WHICH WILL FLUCTUATE OVER THE THREE STAGES OF COGNITIVE DECLINE. SPECIFICALLY, IN THE AAcAcAEURSAA...AEUROEADAPTIVEAAcAcAEURSAAAA STAGE, BPND WILL BE POSITIVELY CORRELATED WITH FC AND NEGATIVELY CORRELATED WITH SC, WHILE IN THE AAcAcAEURSAA...AEUROEADAPTIVE EXHAUSTIONAAcAcAEURSAAAA STAGE, BOTH RELATIONSHIPS WILL REVERSE DIRECTION. STATISTICAL ANALYSIS: USING LINEAR MIXED EFFECTS MODELS, WE WILL ASSESS THE RELATIONSHIP BETWEEN REGIONAL GABAA RECEPTOR DENSITY AND REGIONAL SC/FC/ATROPHY, AND, FURTHER, WILL TEST IF THIS RELATIONSHIP DIFFERS ACROSS THE TWO MS STAGES (CP AND CI). THE MEASURES OF REGIONAL SC/FC WILL BE THE SAME 4 AS DEFINED IN AIM 1.1 (AVERAGE SC/FC NODE STRENGTH OVER REGIONS IN TWO NETWORKS - DEFAULT MODE AND VENTRAL ATTENTION), WHILE THE 5 ATROPHY MEASURES DEFINED IN AIM 1.1 WILL BE COLLAPSED INTO TWO VALUES REPRESENTING ATROPHY IN CORTEX OF INTEREST (AVERAGE OVER RIGHT CINGULATE, LEFT POST-CENTRAL GYRUS AND RIGHT MIDDLE TEMPORAL GYRUS)135 AND ATROPHY IN SUBCORTEX OF INTEREST (AVERAGE OVER THALAMUS AND PUTAMEN)3. WE CHOSE TO REDUCE THE ATROPHY MEASURES FROM 5 TO 2 TO ACCOMMODATE THE SMALLER SAMPLE SIZE IN THE PET STUDY (COMPARED TO MRI STUDY). WE WILL FIT MODELS WITH THE GABAA BPND AS THE OUTCOME, REGRESSED ON AN INDICATOR OF BEING CP (CI REFERENCE GROUP), TWO VALUES OF ATROPHY, 4 VALUES OF REGIONAL FC/SC NODE STRENGTH, AND THE INTERACTION BETWEEN THE TWO. ONE IMPORTANT CONFOUND IN THE FC-BPND MODEL IS POSSIBLE DISEASE-RELATED CHANGES IN CEREBRAL BLOOD FLOW (CBF) AND/OR NEUROVASCULAR COUPLING THAT HAVE BEEN SHOWN TO OCCUR IN MS. THUS, WE WILL DERIVE A PROXY OF NEUROVASCULAR COUPLING USING A RECENTLY DEVELOPED METRIC DOCUSIGN ENVELOPE ID: 3871C02A-ABED-4445-AF29-709C9A2C8631 THAT IS ESSENTIALLY THE SPATIAL CORRELATION OF AN ASL-DERIVED MEASURE OF CBF AND AN FMRI-DERIVED SUMMARY MEASURE OF NEURAL ACTIVITY (FALFF). WE WILL USE THIS REGION-LEVEL MEASURE AS A REGRESSOR IN THE FC- BPND MODEL. A RANDOM INTERCEPT FOR SUBJECT WILL BE INCLUDED IN THE MODEL TO ACCOUNT FOR CORRELATION IN THE DATA RESULTING FROM MULTIPLE OBSERVATIONS PER SUBJECT. TO TEST OUR HYPOTHESIS, WE WILL ASSESS WHETHER THE INTERACTION TERM WITH FC/SC AND BPND IS SIGNIFICANT, INDICATING THEIR RELATIONSHIPS DIFFER FOR CP VERSUS CI INDIVIDUALS. A SEPARATE MODEL WILL BE FIT FOR EACH IMAGING MARKER AND MULTIPLE COMPARISON CORRECTION WILL BE APPLIED. OUR PRIMARY HYPOTHESIS IS VENTRAL ATTENTION AND DEFAULT MODE NETWORKS WILL SHOW SIGNIFICANT POSITIVE RELATIONSHIPS BETWEEN BPND AND FC NODE STRENGTH (AND NEGATIVE RELATIONSHIPS BETWEEN BPND AND SC) IN THE AAcAcAEURSAA...AEUROEADAPTIVEAAcAcAEURSAAAA STAGE, HOWEVER IN THE AAcAcAEURSAA...AEUROEADAPTIVE EXHAUSTIONAAcAcAEURSAAAA STAGE, THE RELATIONSHIP BETWEEN BPND, AND FC WILL BECOME NEGATIVE WHILE THE RELATIONSHIP BETWEEN BPND AND SC WILL BECOME POSITIVE. OTHER BRAIN REGIONS/NETWORKS WILL BE INVESTIGATED AS SECONDARY HYPOTHESES. EXPECTED OUTCOMES WE EXPECT TO SEE A DYNAMIC RELATIONSHIP BETWEEN RECEPTOR FUNCTION, SC AND FC AS SHOWN IN FIG 1. WE EXPECT THESE RESULTS TO REVEAL THAT THE SHIFT FROM THE AAcAcAEURSAA...AEUROEADAPTIVEAAcAcAEURSAAAA STAGE RELATES TO INCREASING SC AND NEURONAL RECEPTOR-LEVEL DAMAGE LEADING TO AN AAcAcAEURSAA...AEUROEADAPTIVE EXHAUSTIONAAcAcAEURSAAAA FC INCREASE AND COGNITIVE IMPAIRMENT. POTENTIAL PROBLEMS AND ALTERNATIVE APPROACHES IF NO ASSOCIATION BETWEEN FC AND BPND IS FOUND, WE WILL INSTEAD TRY CORRELATING A DIFFERENT FMRI MARKER, I.E. FALFF, WHICH HAS BEEN SHOWN TO BE CORRELATED WITH BPND IN HC78. AIM 3. TO BUILD A MODEL THAT CAN IDENTIFY MS PATIENTS AT RISK FOR FUTURE COGNITIVE DECLINE USING ADVANCED MRI MEASURES, AND, FURTHER, TO PROVIDE A CLINICALLY-FEASIBLE ALTERNATIVE USING CLINICAL MRI. UTILIZING BASELINE FC, SC AND GM ATROPHY METRICS FROM ADVANCED MRI IN MS PATIENTS, WE WILL BUILD A PREDICTIVE MODEL TO IDENTIFY INDIVIDUALS AT HIGH RISK OF COGNITIVE DECLINE OVER 4 YEARS. WE WILL ALSO USE OUR PREVIOUSLY VALIDATED TOOLS FOR ESTIMATING SC AND FC FROM CLINICAL MRI TO BUILD CLINICALLY-APPLICABLE PREDICTIVE MODELS. OUR MAIN HYPOTHESIS IS THAT MRI-BASED MEASURES OF SC/FC/ATROPHY CAN BE USED TO IDENTIFY PATIENTS AT HIGH RISK FOR FUTURE COGNITIVE DECLINE, AND, FURTHER, ESTIMATED FC AND SC USING ONLY CLINICAL MRI CAN PREDICT FUTURE COGNITIVE DECLINE EQUALLY AS WELL AS THE MODEL BASED ON ADVANCED SC/FC/ATROPHY MEASURES. CONSTRUCTION OF ESTIMATED SC AND FC NETWORKS: FOR MS PATIENTS, LESION MASKS WILL BE CREATED SEMI-AUTOMATICALLY DEFINED USING T1, T2 AND FLAIR SEQUENCES FROM THE CLINICAL MRI140. T1/T2S WILL BE LINEARLY AND NON-LINEARLY COREGISTERED TO MNI SPACE, AND THE TRANSFORMATIONS APPLIED TO THE LESION MASK (WITH NEAREST NEIGHBOR INTERPOLATION). WE WILL USE CONTROL POINTS AND MANUAL CORRECTION (IF NEEDED) TO ACHIEVE ACCURATE COREGISTRATION, AS WE HAVE PERFORMED IN MS POPULATIONS BEFORE3,4. ONCE THE LESION MASK IS IN MNI SPACE, IT WILL AGAIN BE VERIFIED MANUALLY WITH SIMULTANEOUS VIEWING OF NATIVE AND MNI TRANSFORMED SCAN. THE NETWORK MODIFICATION (NEMO) TOOL V2 REQUIRES ONLY AN INDIVIDUALAAcAcAEURSAAcAEURZAcS LESION MASK IN MNI SPACE, AS DESCRIBED ABOVE, TO PRODUCE AN ESTIMATED SC AND FC MATRIX (FIG 7). THIS NEW, IMPROVED VERSION OF THE NEMO TOOL, ORIGINALLY PUBLISHED IN 201399, INCLUDES A REFERENCE DATABASE OF CONNECTIVITY NETWORKS FROM 420 UNRELATED INDIVIDUALS FROM THE HUMAN CONNECTOME PROJECTAAcAcAEURSAAcAEURZAcS (HCP) 1200 RELEASE (~50% FEMALE, AGED 25-35). THE NEMO TOOL BEGINS BY MAPPING THE LESION MASK INTO THIS HEALTHY DATABASEAAcAcAEURSAAcAEURZAcS COLLECTION OF TRACTOGRAPHY STREAMLINES THAT QUANTIFY LIKELY WHITE MATTER PATHWAYS. IT IDENTIFIES STREAMLINES THAT PASS THROUGH THE LESION MASK AND REMOVES THEM BEFORE RECALCULATING THE ESTIMATED SC MATRIX (WHICH DOES NOT INCLUDE THESE REMOVED STREAMLINES). FINALLY, WE RECENTLY DEVELOPED ANOTHER OUTPUT FROM THE NEMO TOOL, SPECIFICALLY, ESTIMATED FC. WE FIRST CREATED A NEURAL NETWORK THAT CAN ESTIMATE FC FROM SC, BASED ON A PREVIOUSLY VALIDATED APPROACH141 AND TRAINED ON DATA FROM THE 420 HCP SUBJECTS IN THE NEMO TOOL. ONCE THE NEMO TOOL CREATES THE ESTIMATED SC, WE THEN USE THIS ARTIFICIAL NEURAL NETWORK TO CALCULATE THE CORRESPONDING ESTIMATED FC. IN THE END, THE NEMO TOOL CREATES AN ESTIMATED SC AND ESTIMATED FC MATRIX DIRECTLY FROM CLINICAL MRI (LESION MASKS). THIS APPROACH WAS USED IN 100 MS PATIENTS, WHERE WE SHOWED ESTIMATED SC/FC MODELS PERFORMED JUST AS WELL AS, AND MOSTLY BETTER THAN, OBSERVED SC/FC MEASURES COLLECTED VIA ADVANCED MRI140. BUILD A MODEL TO PREDICT 4-YEAR COGNITIVE DECLINE FROM ATROPHY AND OBSERVED/ESTIMATED SC AND FC. WE WILL CREATE AND VALIDATE MODELS THAT USE OBSERVED SC/FC, EXTRACTED FROM MS SUBJECTSAAcAcAEURSAAcAEURZAc BASELINE MRIS AS DESCRIBED PREVIOUSLY IN THIS APPROACH SECTION (FIG 7), AND THE NEMO TOOLAAcAcAEURSAAcAEURZAcS ESTIMATED SC/FC, BASED ON LESION MASKS DERIVED FROM THE PATIENTSAAcAcAEURSAAcAEURZAc CLINICAL MRIS AS DESCRIBED ABOVE, TO PREDICT 4-YEAR COGNITIVE SCORES. AS IN AIM 1, WE WILL FOCUS ON CHANGE IN SDMT SCORE, SPECIFICALLY A CLINICALLY RELEVANT CHANGE (4 POINTS) TO DEFINE COGNITIVE DECLINE84. TWO MODELS WILL BE BUILT AAcAcAEURSAAcA'A" ONE FOR FUTURE SDMT SCORE AND ONE FOR A BINARY VARIABLE INDICATING COGNITIVE DECLINE. WE WILL ALSO EXPLORE CREATING MODELS TO PREDICT OTHER CHRONIC COGNITIVE SCORES FOR MAXIMAL CLINICAL UTILITY. FIGURE 7: WORKFLOW FOR ESTIMATING SC AND FC FROM CLINICAL MRI. DOCUSIGN ENVELOPE ID: 3871C02A-ABED-4445-AF29-709C9A2C8631 WE WILL USE KERNEL REGRESSION, AS THIS METHOD HAS BEEN SHOWN TO HAVE SOME OF THE BEST ACCURACY RATES WHEN PREDICTING COGNITIVE SCORES FROM FC142. AN ENSEMBLE MODELING APPROACH WILL BE USED TO COMBINE CONNECTIVITY METRICS AND DEMOGRAPHICS/CLINICAL DATA IN THE PREDICTION MODEL. SPECIFICALLY, WE WILL USE KERNEL REGRESSION TO PREDICT COGNITION (SCORE AND DECLINE) FROM AGE, DISEASE DURATION, SEX, EDSS, EDUCATION, BMI, DMT, MENOPAUSAL STATUS, SUBSTANCE USE, COGNITIVE/PHYSICAL THERAPY INFORMATION, COMORBIDITIES, AND AVERAGE THE PREDICTIONS OF THIS MODEL WITH THE PREDICTIONS FROM THE KERNEL REGRESSION MODEL BASED ON ESTIMATED SC/FC AND OBSERVED SC/FC. CROSS-VALIDATION, INCLUDING TRAINING, VALIDATION AND TESTING SPLITS WILL BE USED IN MODEL BUILDING AND HYPERPARAMETER TUNING TO AVOID OVERFITTING AND INCREASE GENERALIZABILITY. CARE WILL BE TAKEN AS TO NOT HAVE DATA LEAKAGE AND DROPOUT WILL BE PERFORMED TO MINIMIZE OVERFITTING120,143. MODEL ACCURACY WILL BE ASSESSED ON HOLDOUT TEST SETS USING EXPLAINED VARIANCE (R2), MEAN SQUARED ERROR, MEAN ABSOLUTE ERROR (FOR CONTINUOUS COGNITIVE SCORE), AND AREA UNDER THE RECEIVER OPERATING CHARACTERISTIC CURVE (AUC) (FOR BINARIZED COGNITIVE DECLINE). DISTRIBUTION OF PREDICTION ACCURACIES FOR THE FOUR MODELS (OVER 100 OUTER LOOP FOLDS) BASED ON THE ENSEMBLE MODEL USING DEMOGRAPHICS AND ESTIMATED SC/FC AND OBSERVED SC/FC WILL BE COMPARED USING T-TESTS. WE HYPOTHESIZE THAT MODELS BASED ON ESTIMATED SC AND FC WILL PERFORM AT LEAST AS WELL AS THOSE BASED ON OBSERVED SC AND FC. SENSITIVITY ANALYSIS WILL PROVIDE IMPORTANCE OF EACH CONNECTION OR DEMOGRAPHIC IN ORDER TO BETTER UNDERSTAND WHAT IS IMPORTANT IN COGNITIVE RESILIENCE AND POTENTIAL TARGETS FOR NON-INVASIVE BRAIN STIMULATION140. VALIDATION OF THE CLINICAL MRI-BASED PREDICTIVE MODEL WITH A LARGE, OUT-OF-SAMPLE DATASET WE WILL VALIDATE THE PREDICTIVE MODEL USING DATA THAT WE ARE CURRENTLY COLLECTING UNDER AN R01 MECHANISM FROM CO-PI GAUTHIER (RO1 NS104283), CONSISTING OF CLINICAL MRI AND COGNITIVE SCORES AT BASELINE AND AT 4 YEARS FROM 150 INDIVIDUALS WITH MS. MODEL ACCURACY WILL BE ASSESSED WITH R2 VALUES OF PREDICTED AND TRUE 4-YEAR SDMT AND AUC FOR PREDICTING COGNITIVE DECLINE. THIS ESSENTIAL STEP ENSURES THE MODEL IS NOT OVERFITTED AND CAN BE APPLIED TO NOVEL INDIVIDUALS. EXPECTED OUTCOMES: THIS AIM IS EXPECTED TO RESULT IN CONCLUSIVE EVIDENCE THAT ESTIMATES OF SC AND FC, BASED ONLY ON A SUBJECTAAcAcAEURSAAcAEURZAcS LESION MASK, CAN PREDICT 4-YEAR COGNITIVE OUTCOMES JUST AS WELL AS DIRECT MEASURES OF SC AND FC, BASED ON AN INDIVIDUALAAcAcAEURSAAcAEURZAcS ADVANCED MRI. THIS MODEL WOULD HAVE CLINICAL IMPLICATIONS IN THAT, GOING FORWARD, ADVANCED MRI WOULD NOT BE NECESSARY TO PROVIDE ACCURATE PREDICTIONS OF FUTURE COGNITIVE DECLINE IN MS. THIS MODEL, MADE PUBLICLY AVAILABLE VIA THE NEMO TOOL, COULD BE USED BY CLINICIANS TO STRATIFY PATIENTS AND IDENTIFY THOSE THAT MIGHT BE AT RISK FOR COGNITIVE DECLINE AND THUS MAY BENEFIT FROM MORE INTENSIVE INTERVENTION. POTENTIAL PROBLEMS AND ALTERNATIVE APPROACHES: PREVIOUS WORK HAS SHOWN KERNEL REGRESSION PERFORMS WELL IN PREDICTING COGNITIVE SCORES FROM CONNECTIVITY MEASURES, STILL THIS APPROACH MAY NOT RESULT IN A HIGH EXPLAINED VARIANCE. THEREFORE, WE MAY USE BRAINNETCNN144, NETWORK BASED STATISTICS145 OR PARTIAL LEAST SQUARES REGRESSION3,4, WHICH HAVE BEEN USED SUCCESSFULLY ON SIMILAR DATASETS TO MAP CONNECTIVITY TO COGNITIVE SCORES. FUTURE PLANS. WE WILL USE INFORMATION GAINED RELATED TO COGNITIVE RESILIENCE TO INFORM NOVEL THERAPEUTICS. FOR EXAMPLE, WE MAY IDENTIFY BRAIN REGIONS WHOSE INCREASED FC IN THE AAcAcAEURSAA...AEUROEADAPTIVEAAcAcAEURSAAAA STAGE OR DECREASED FC IN THE AAcAcAEURSAA...AEUROEADAPTIVE EXHAUSTIONAAcAcAEURSAAAA STAGE IS ASSOCIATED WITH COGNITIVE RESILIENCE AND TARGET THESE REGIONS IN NON-INVASIVE BRAIN STIMULATION THERAPY (EITHER SUPPRESSING OR ACTIVATING DESIGN) TO MODULATE FC LEVELS. IF IT TURNS OUT THAT FMZ-PET BPND IS POSITIVELY CORRELATED WITH FC IN AIM 2, IMPACTING THE GABAA RECEPTORS USING PHARMACOLOGICAL AGENTS101- 103 MAY IMPACT THE ADAPTIVE OR MALADAPTIVE FC MECHANISMS TO SUPPORT COGNITIVE RESILIENCE.
Cleveland State University
Project Grant R01NS134646
$24.1k 11/5/24