Magnetic Resonance Imaging Volumetric Measurements Research Articles

Combining neuropsychological assessment and structural neuroimaging to identify early Alzheimer's disease in a memory clinic cohort.

The current study examined the contributions of comprehensive neuropsychological assessment and volumetric assessment of selected mesial temporal subregions on structural magnetic resonance imaging (MRI) to identify patients with amnestic mild cognitive impairment (aMCI) and mild probable Alzheimer's disease (AD) dementia in a memory clinic cohort. Comprehensive neuropsychological assessment and automated entorhinal, transentorhinal, and hippocampal volume measurements were conducted in 40 healthy controls, 38 patients with subjective memory symptoms, 16 patients with aMCI, 16 patients with mild probable AD dementia. Multinomial logistic regression was used to compare the neuropsychological and MRI measures. Combining the neuropsychological and MRI measures improved group membership prediction over the MRI measures alone but did not improve group membership prediction over the neuropsychological measures alone. Comprehensive neuropsychological assessment was an important tool to evaluate cognitive impairment. The mesial temporal volumetric MRI measures contributed no diagnostic value over and above the determinations made through neuropsychological assessment.

The association between cannabis use and neuroimaging measures in older adults: findings from the UK biobank.

Cannabis use has increased in recent years. However, the long-term implications of cannabis use on brain health remain unknown. We explored the associations of cannabis use with volumetric brain magnetic resonance imaging (MRI) measures in dementia-free older adults. This cross-sectional and longitudinal study included dementia-free participants of the UK Biobank aged ≥60years. Linear regression models were used to evaluate the association of cannabis use and patterns of use with volumetric brain MRI measures. The association between cannabis use and change in brain MRI measures over time was also tested. All models were adjusted for potential confounders. The sample included 19,932 participants (mean age 68 ± 5years, 48% men), 3,800 (19%) reported lifetime use of cannabis. Cannabis use was associated with smaller total, white, grey and peripheral cortical grey matter volumes (B= -6,690 ± 1,157; P< 0.001, B= -4,396 ± 766; P< 0.001, B= -2,140 ± 690; P= 0.002 and B= -2,451 ± 606; P< 0.001, respectively). Among cannabis users, longer duration of use was associated with smaller total brain, grey and cortical grey matter volumes (B= -7,878 ± 2,396; P= 0.001, B= -5,411 ± 1,430; P< 0.001, B = -5,396 ± 1,254; P< 0.001, respectively), and with increased white matter hyperintensity volume (B= 0.09 ± 0.03; P= 0.008). Additionally, current vs. former users (B= -10,432 ± 4,395; P= 0.020) and frequent versus non-frequent users (B= -2,274 ± 1,125; P= 0.043) had smaller grey and cortical grey matter volumes, respectively. No significant associations were observed between cannabis use and change in brain MRI measures. Our findings suggest that cannabis use, particularly longer duration and frequent use, may be related to smaller grey and white matter volumes in older ages, but not to late-life changes in these measures over time.

Cross-cultural Diagnostic Validity of the Multilingual Naming Test (MINT) in a Sample of Older Adults.

We aimed to evaluate the psychometric properties and diagnostic accuracy of the 32-item version of the Multilingual Naming Test (MINT) in participants from 2 ethnic groups (European Americans [EA; n= 106] and Hispanic Americans [HA; n= 175]) with 3 diagnostic groups (cognitively normal [CN], n= 94, mild cognitive impairment [MCI], n= 148, and dementia, n= 39). An Item Response Theory model was used to evaluate items across ethnicity and language groups (Spanish and English), resulting in a 24-item version. We analyzed the MINT discriminant and predictive validity across diagnostic groups. A total of 8 items were differentially difficult between languages in the 32-item version of the MINT. EA scored significantly higher than HA, but the difference was not significant when removing those 8 items (controlling for Education). The Receiver Operating Characteristics showed that the MINT had poor accuracy when identifying CN participants and was acceptable in identifying dementia participants but unacceptable in classifying MCI participants. Finally, we tested the association between MINT scores and magnetic resonance imaging volumetric measures of language-related areas in the temporal and frontal lobes. The 32-item MINT in English and Spanish and the 24-item MINT in Spanish were significantly correlated with the bilateral middle temporal gyrus. The left fusiform gyrus correlated with MINT scores regardless of language and MINT version. We also found differential correlations depending on the language of administration. Our results highlight the importance of analyzing cross-cultural samples when implementing clinical neuropsychological tests such as the MINT.

Brain MRI in Progressive Supranuclear Palsy with Richardson's Syndrome and Variant Phenotypes.

Brain magnetic resonance imaging (MRI) is used to support the diagnosis of progressive supranuclear palsy (PSP). However, the value of visual descriptive, manual planimetric, automatic volumetric MRI markers and fully automatic categorization is unclear, particularly regarding PSP predominance types other than Richardson's syndrome (RS). To compare different visual reading strategies and automatic classification of T1-weighted MRI for detection of PSP in a typical clinical cohort including PSP-RS and (non-RS) variant PSP (vPSP) patients. Forty-one patients (21 RS, 20 vPSP) and 46 healthy controls were included. Three readers using three strategies performed MRI analysis: exclusively visual reading using descriptive signs (hummingbird, morning-glory, Mickey-Mouse), visual reading supported by manual planimetry measures, and visual reading supported by automatic volumetry. Fully automatic classification was performed using a pre-trained support vector machine (SVM) on the results of atlas-based volumetry. All tested methods achieved higher specificity than sensitivity. Limited sensitivity was driven to large extent by false negative vPSP cases. Support by automatic volumetry resulted in the highest accuracy (75.1% ± 3.5%) among the visual strategies, but performed not better than the midbrain area (75.9%), the best single planimetric measure. Automatic classification by SVM clearly outperformed all other methods (accuracy, 87.4%), representing the only method to provide clinically useful sensitivity also in vPSP (70.0%). Fully automatic classification of volumetric MRI measures using machine learning methods outperforms visual MRI analysis without and with planimetry or volumetry support, particularly regarding diagnosis of vPSP, suggesting the use in settings with a broad phenotypic PSP spectrum. © 2023 The Authors. Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.

Concordance between US and MRI Two-Dimensional Measurement and Volumetric Segmentation in Fetal Ventriculomegaly.

We provide a study comparison between two-dimensional measurement and volumetric (3D) segmentation of the lateral ventricles and brain structures in fetuses with isolated and non-isolated ventriculomegaly with 3D virtual organ computer-aided analysis (VOCAL) ultrasonography vs. magnetic resonance imaging (MRI) analyzed with 3D-Slicer software. In this cross-sectional study, 40 fetuses between 20 and 38 gestational weeks with various degrees of ventriculomegaly were included. A total of 71 ventricles were measured with ultrasound (US) and with MRI. A total of 64 sonographic ventricular volumes, 80 ventricular and 40 fetal brain MR volumes were segmented and analyzed using both imaging modalities by three observers. Sizes and volumes of the ventricles and brain parenchyma were independently analyzed by two radiologists, and interobserver correlation of the results with 3D fetal ultrasound data was performed. The semiautomated rotational multiplanar 3D VOCAL technique was performed for ultrasound volumetric measurements. Results were compared to manually extracted ventricular and total brain volumes in 3D-Slicer. Segmentation of fetal brain structures (cerebral and cerebellar hemispheres, brainstem, ventricles) performed independently by two radiologists showed high interobserver agreement. An excellent agreement between VOCAL and MRI volumetric and two-dimensional measurements was established, taking into account the intraclass correlation coefficients (ICC), and a Bland-Altman plot was established. US and MRI are valuable tools for performing fetal brain and ventricular volumetry for clinical prognosis and patient counseling. Our datasets could provide the backbone for further construction of quantitative normative trajectories of fetal intracranial structures and support earlier detection of abnormal brain development and ventriculomegaly, its timing and progression during gestation.

Morphometric magnetic resonance imaging cortico-subcortical features in Parkinson's disease with mild cognitive impairment.

Parkinson's disease (PD) patients with cognitive impairment undergo progressive atrophy of several cortical and subcortical areas. The aim was to study the magnetic resonance imaging (MRI) morphometric features of PD patients with mild cognitive impairment (MCI). Patients from the Parkinson's Disease Cognitive Impairment Study (PACOS) cohort with an available structural volumetric brain MRI and morphometric measurements of the midbrain and pons areas, middle cerebellar peduncle, superior cerebellar peduncle width and midbrain anteroposterior diameter (A-Pdiam) were included. MCI was diagnosed according to the Movement Disorder Society level II criteria. Additionally, cortical thickness analysis was performed and correlated with morphometric brainstem measurements. Morphometric measurements were available for 168 subjects, of whom 67 (39.9%) were diagnosed with PD-MCI. The mean age (± standard deviation) of the sample was 64.2 ± 9.8. Amongst patients, 84 (50%) were men with a disease duration of 5.2 ± 5.4 years and a Unified Parkinson's Disease Rating Scale-Motor Examination score of 32.1 ± 12.9. In the univariate and multivariate analysis, after adjusting for age, sex, years of schooling and disease duration, MCI was associated with midbrain area (odds ratio 0.98; 95% confidence interval 0.96-0.99; p=0.048) and A-Pdiam (odds ratio 0.63; 95% confidence interval 0.46-0.86; p=0.005). Furthermore, 121 PD patients underwent cortical thickness analysis, which showed the presence of cortical thinning in lateral orbitofrontal regions of patients with PD-MCI. No correlation was found between cortical thickness and brainstem morphometric measurements. A mild midbrain atrophy and the presence of frontal cortical thickness reduction might be considered a structural MRI feature of PD patients with MCI.

Thalamic atrophy, duration of illness, and years of education are the best predictors of cognitive impairment in multiple sclerosis

Background Cognitive impairment (CI) is common in multiple sclerosis. We assessed the relationship between CI and demographic, motor disability, and volumetric brain atrophy in relapsing remitting multiple sclerosis (RRMS). Methods In all, 43 RRMS patients and 40 age- sex-, and education-matched normal volunteers were enrolled. Assessments included: demographic, Expanded Disability Status Scale (EDSS); Brief International Cognitive Assessment for Multiple Sclerosis (BICAMS); and brain volumetric assessment. Results The mean age of onset of RRMS patients was 25.72 y and the mean number of attacks was 3.30 ± 2.33/year. The total brain volume, cerebral, and cerebellar hemispheres, brainstem as well as thalamic and hippocampal gray matter were significantly reduced in RRMS patients compared with controls. CI was detected in 41.9% of patients. Patients with CI had fewer years of education (P = .001), a higher number of attacks (P = .002), worse EDSS scores (P = .030); reduced gray matter volume, particularly of the thalamus, and hippocampus compared with patients without CI. There were significant negative correlations between CI with duration of illness, number of attacks, time elapsed between onset and diagnosis, and EDSS. Thalamic volume showed strong positive correlations with CI (P < .0001). The best predictors of CI were thalamic atrophy, duration of illness, and years of education (P = .001, .010, and .035 respectively). Conclusion We confirmed the presence of gray matter atrophy of the brain in RRMS patients. Volumetric magnetic resonance imaging measures of thalamus, duration of illness, and years of education could be useful biomarkers of CI in RRMS.

Volumetric MRI is a promising outcome measure of muscle reinnervation

The development of outcome measures that can track the recovery of reinnervated muscle would benefit the clinical investigation of new therapies which hope to enhance peripheral nerve repair. The primary objective of this study was to assess the validity of volumetric Magnetic Resonance Imaging (MRI) as an outcome measure of muscle reinnervation by testing its reproducibility, responsiveness and relationship with clinical indices of muscular function. Over a 3-year period 25 patients who underwent nerve transfer to reinnervate elbow flexor muscles were assessed using intramuscular electromyography (EMG) and MRI (median post-operative assessment time of 258 days, ranging from 86 days pre-operatively to 1698 days post- operatively). Muscle power (Medical Research Council (MRC) grade) and Stanmore Percentage of Normal Elbow Assessment (SPONEA) assessment was also recorded for all patients. Sub-analysis of peak volitional force (PVF), muscular fatigue and co-contraction was performed in those patients with MRC > 3. The responsiveness of each parameter was compared using Pearson or Spearman correlation. A Hierarchical Gaussian Process (HGP) was implemented to determine the ability of volumetric MRI measurements to predict the recovery of muscular function. Reinnervated muscle volume per unit Body Mass Index (BMI) demonstrated good responsiveness (R2 = 0.73, p < 0.001). Using the temporal and muscle volume per unit BMI data, a HGP model was able to predict MRC grade and SPONEA with a mean absolute error (MAE) of 0.73 and 1.7 respectively. Muscle volume per unit BMI demonstrated moderate to good positive correlations with patient reported impairments of reinnervated muscle; co- contraction (R2 = 0.63, p = 0.02) and muscle fatigue (R2 = 0.64, p = 0.04). In summary, volumetric MRI analysis of reinnervated muscle is highly reproducible, responsive to post-operative time and demonstrates correlation with clinical indices of muscle function. This encourages the view that volumetric MRI is a promising outcome measure for muscle reinnervation which will drive advancements in motor recovery therapy.

Data analysis with Shapley values for automatic subject selection in Alzheimer\u2019s disease data sets using interpretable machine learning

BackgroundFor the recruitment and monitoring of subjects for therapy studies, it is important to predict whether mild cognitive impaired (MCI) subjects will prospectively develop Alzheimer’s disease (AD). Machine learning (ML) is suitable to improve early AD prediction. The etiology of AD is heterogeneous, which leads to high variability in disease patterns. Further variability originates from multicentric study designs, varying acquisition protocols, and errors in the preprocessing of magnetic resonance imaging (MRI) scans. The high variability makes the differentiation between signal and noise difficult and may lead to overfitting. This article examines whether an automatic and fair data valuation method based on Shapley values can identify the most informative subjects to improve ML classification.MethodsAn ML workflow was developed and trained for a subset of the Alzheimer’s Disease Neuroimaging Initiative (ADNI) cohort. The validation was executed for an independent ADNI test set and for the Australian Imaging, Biomarker and Lifestyle Flagship Study of Ageing (AIBL) cohort. The workflow included volumetric MRI feature extraction, feature selection, sample selection using Data Shapley, random forest (RF), and eXtreme Gradient Boosting (XGBoost) for model training as well as Kernel SHapley Additive exPlanations (SHAP) values for model interpretation.ResultsThe RF models, which excluded 134 of the 467 training subjects based on their RF Data Shapley values, outperformed the base models that reached a mean accuracy of 62.64% by 5.76% (3.61 percentage points) for the independent ADNI test set. The XGBoost base models reached a mean accuracy of 60.00% for the AIBL data set. The exclusion of those 133 subjects with the smallest RF Data Shapley values could improve the classification accuracy by 2.98% (1.79 percentage points). The cutoff values were calculated using an independent validation set.ConclusionThe Data Shapley method was able to improve the mean accuracies for the test sets. The most informative subjects were associated with the number of ApolipoproteinE ε4 (ApoE ε4) alleles, cognitive test results, and volumetric MRI measurements.

Evaluation of Early Cerebral Atrophy in Mild Cognitive Impairment with Positive Amyloid PET using MRIVolumetric Measurement

Objective: To identify early cerebral atrophy by using magnetic resonance imaging (MRI) volumetric measurement to distinguish mild cognitive impairment (MCI) from normal aging. Materials and Methods: A retrospective study in 29 Alzheimer’s disease (AD) patients (mean 75.6±7.3 years), 11 MCI patients (mean 68.8±3.9 years), and 27 healthy control (HC) subjects (mean 69.3±4.7 years) was performed with analysis of neurological and neuropsychiatric test, and underwent 3T MRI with high-resolution 3D-T1W. Quantitative volumetric analysis of brain including surface area, cortical gray matter volume, cortical thickness, and hippocampal subfields was performed by using FreeSurfer software. Results: The diminishment of cortical gray matter volume and cortical thickness were involved in most of brain regions, predominantly in temporal lobe with statistical significance in AD compared with MCI and HC. Comparison between amyloid positron emission tomography (PET) positive MCI subjects and HC has statistically significant difference in most regions of hippocampal subfield. The highest accuracy of 90.01% with sensitivity of 50% and specificity of 100% were found at subiculum. A comparison between amyloid PET positive MCI subjects and amyloid PET negative MCI subjects revealed significant differences at right molecular layer, right/average GC-ML-DG, right CA2/3, right CA4, and average CA4 with good accuracy, sensitivity, and specificity. Conclusion: The present study confirmed improved sensitivity of MRI volumetric measurement with hippocampal subfield analysis to identify early stage of AD in MCI patients, at least compared with positive amyloid PET MCI. Study with higher number of subjects using this method to discriminate MCI and normal aging control would provide benefits as the screening tool in older population. Keywords: Hippocampal subfield; Volumetric analysis; Alzheimer’s disease; Mild cognitive impairment

Traumatic brain injury: a comparison of diffusion and volumetric magnetic resonance imaging measures.

Cognitive impairment after traumatic brain injury remains hard to predict. This is partly because axonal injury, which is of fundamental importance, is difficult to measure clinically. Advances in MRI allow axonal injury to be detected after traumatic brain injury, but the most sensitive approach is unclear. Here, we compare the performance of diffusion tensor imaging, neurite orientation dispersion and density-imaging and volumetric measures of brain atrophy in the identification of white-matter abnormalities after traumatic brain injury. Thirty patients with moderate–severe traumatic brain injury in the chronic phase and 20 age-matched controls had T1-weighted and diffusion MRI. Neuropsychological tests of processing speed, executive functioning and memory were used to detect cognitive impairment. Extensive abnormalities in neurite density index and orientation dispersion index were observed, with distinct spatial patterns. Fractional anisotropy and mean diffusivity also indicated widespread abnormalities of white-matter structure. Neurite density index was significantly correlated with processing speed. Slower processing speed was also related to higher mean diffusivity in the corticospinal tracts. Lower white-matter volumes were seen after brain injury with greater effect sizes compared to diffusion metrics; however, volume was not sensitive to changes in cognitive performance. Volume was the most sensitive at detecting change between groups but was not specific for determining relationships with cognition. Abnormalities in fractional anisotropy and mean diffusivity were the most sensitive diffusion measures; however, neurite density index and orientation dispersion index may be more spatially specific. Lower neurite density index may be a useful metric for examining slower processing speed.

Relationship Between Interpersonal Depressive Symptoms and Reduced Amygdala Volume in People with Multiple Sclerosis: Considerations for Clinical Practice.

The lifetime prevalence of depression in people with multiple sclerosis (MS) is approximately 50% compared with around 15% in the general population. There is a relationship between depression and quality of life in people with MS and evidence that depression may contribute to disease progression. This cross-sectional pilot study assessed the association between depression and regional brain atrophy, including amygdala and hippocampal volume. Forty-nine participants with MS recruited through a hospital MS clinic were administered the Center for Epidemiological Studies Depression Scale Revised (CESD-R) to investigate whether higher endorsements on the items depressive affect and interpersonal symptoms were associated with volumetric magnetic resonance imaging measurements of hippocampal and amygdala atrophy. Regression analysis revealed an association between depression-related interpersonal symptoms and right amygdala volume. No association was found between depression and hippocampal volume. These results provide preliminary support for a unilateral, biologically based relationship between the right amygdala and characteristic interpersonal depressive symptoms expressed by people with MS and add to the growing body of literature implicating regional brain atrophy in MS-associated depression. Given that the interpersonal subcomponent of the CESD-R measures social functioning, and the neural networks in the amygdala are known to be implicated in processing social stimuli, this research suggests that targeted diagnosis and treatments for depression in people with MS may be particularly beneficial. Further confirmatory research of this relationship is required.

Optical coherence tomography measures correlate with brain and spinal cord atrophy and multiple sclerosis disease-related disability.

Both optical coherence tomography (OCT) and magnetic resonance imaging (MRI) volumetric measures have been postulated as potential biomarkers of multiple sclerosis (MS)-related disability. The aim of the study was to investigate the association between OCT and brain volume and spinal cord area (SCA) parameters in patients with relapsing MS and to assess their independent associations with disability. This was a cross-sectional analysis of 90 patients with MS who underwent OCT and MRI examination. Values of peripapillary retinal nerve fibre layer (pRNFL), ganglion cell/inner plexiform layer (GCIPL) and inner nuclear layer of eyes without previous optic neuritis were obtained. SCA and brain parenchymal fraction (BPF), grey and white matter fractions were obtained. Multivariable regression analyses were conducted with disability as dependent variable. Lower pRNFL thickness and lower GCIPL volume as well as lower BPF, grey matter fraction and SCA were associated with a longer disease duration and a higher Expanded Disability Status Scale score. Lower pRNFL thickness and GCIPL volumes were associated with lower BPF and SCA. In the multivariable logistic regression analyses, pRNFL thickness and GCIPL volume outperformed MRI in predicting disability. The OCT measures correlate with brain and spinal cord atrophy and appear more closely associated with disability than MRI volumetric measures.

Normal hip joint fluid volumes in healthy children of different ages, based on MRI volumetric quantitative measurement.

While magnetic resonance imaging (MRI) of the pelvis and hips is common in pediatric patients, to date there are no data on the quantification of normal hip joint fluid volume in this patient population. We sought to assess the feasibility and reliability of quantitative hip joint fluid measurement in the pediatric population to estimate the normal volume of fluid in a pediatric hip joint. Seventy healthy children ages 8-17years underwent a pelvic MRI including a large field of view coronal T2 fat-saturated sequence where hips were entirely imaged. Following 3 training sessions, 2 readers with experience in musculoskeletal imaging performed volumetric quantitative measurements of hip fluid (140 hips) using semiautomated pixel-based thresholding on custom MATLAB software. The mean processing time per hip was 2min, 41s. The mean volume of fluid in a hip joint was 2.1mL (range: 0.38-5.41mL), increasing slightly with age. Volumes were also greater in boys than in girls (P=0.004). Intra-observer and interobserver agreement were high (intra-class correlation coefficients 0.93 and 0.98, respectively), with mean volume differences of 0.04mL for intra-observer and 0.09mL for interobserver. A semiautomated pixel-based thresholding approach was feasible and reliable for measuring joint fluid in pediatric hip MRI. The average fluid volume of 2.1mL can represent a visually substantial quantity of fluid per MRI slice, particularly in small children, and should not be misinterpreted as a joint effusion.

Brain tissue volumes at term-equivalent age are associated with early motor behavior in very preterm infants.

Preterm birth is associated with a wide range of adverse developmental outcomes, including sensory, motor, cognitive and language impairments, and behavioral or attention problems. Subtle motor deficits that might emerge in premature infants with no evident or with mild brain injury encompass qualitative and quantitative aspects of motor behavior. This prospective cohort study provided an evaluation of the relationship between brain tissue volumes revealed by magnetic resonance imaging (MRI) at term-equivalent age and motor behavior in infancy in very preterm infants (total number=40; mean gestational age=28weeks+4days; mean birth weight=1190g) without evident or with mild brain injury. Infants were recruited at birth and assessed at 12months corrected age using the tool for qualitative and quantitative assessment of motor behavior, infant motor profile. The brain tissue was segmented first using advanced segmentation techniques and the volumes were measured by summing the volumes of all voxels belonging to a particular tissue class. The associations between volumetric brain MRI measures with motor behavior were explored using linear regression analyses. Results showed that larger total brain volumes were associated with higher motor score. Similar relationships were documented for parietal lobe, deep gray matter, and cerebellum volumes. Volumetric quantitative data of brain structures may serve as biomarkers for subtle motor deficits described in very preterm born infants without or with mild brain lesions apparent on MRI.

Diagnosis Across the Spectrum of Progressive Supranuclear Palsy and Corticobasal Syndrome

Atypical parkinsonian syndromes (APS), including progressive supranuclear palsy (PSP), corticobasal syndrome (CBS), and multiple system atrophy (MSA), may be difficult to distinguish in early stages and are often misdiagnosed as Parkinson disease (PD). The diagnostic criteria for PSP have been updated to encompass a range of clinical subtypes but have not been prospectively studied. To define the distinguishing features of PSP and CBS subtypes and to assess their usefulness in facilitating early diagnosis and separation from PD. This cohort study recruited patients with APS and PD from movement disorder clinics across the United Kingdom from September 1, 2015, through December 1, 2018. Patients with APS were stratified into the following groups: those with Richardson syndrome (PSP-RS), PSP-subcortical (including PSP-parkinsonism and progressive gait freezing subtypes), PSP-cortical (including PSP-frontal and PSP-CBS overlap subtypes), MSA-parkinsonism, MSA-cerebellar, CBS-Alzheimer disease (CBS-AD), and CBS-non-AD. Data were analyzed from February 1, through May 1, 2019. Baseline group comparisons used (1) clinical trajectory; (2) cognitive screening scales; (3) serum neurofilament light chain (NF-L) levels; (4) TRIM11, ApoE, and MAPT genotypes; and (5) volumetric magnetic resonance imaging measures. A total of 222 patients with APS (101 with PSP, 55 with MSA, 40 with CBS, and 26 indeterminate) were recruited (129 [58.1%] male; mean [SD] age at recruitment, 68.3 [8.7] years). Age-matched control participants (n = 76) and patients with PD (n = 1967) were included for comparison. Concordance between the antemortem clinical and pathologic diagnoses was achieved in 12 of 13 patients with PSP and CBS (92.3%) undergoing postmortem evaluation. Applying the Movement Disorder Society PSP diagnostic criteria almost doubled the number of patients diagnosed with PSP from 58 to 101. Forty-nine of 101 patients with reclassified PSP (48.5%) did not have the classic PSP-RS subtype. Patients in the PSP-subcortical group had a longer diagnostic latency and a more benign clinical trajectory than those in PSP-RS and PSP-cortical groups. The PSP-subcortical group was distinguished from PSP-cortical and PSP-RS groups by cortical volumetric magnetic resonance imaging measures (area under the curve [AUC], 0.84-0.89), cognitive profile (AUC, 0.80-0.83), serum NF-L level (AUC, 0.75-0.83), and TRIM11 rs564309 genotype. Midbrain atrophy was a common feature of all PSP groups. Eight of 17 patients with CBS (47.1%) undergoing cerebrospinal fluid analysis were identified as having the CBS-AD subtype. Patients in the CBS-AD group had a longer diagnostic latency, relatively benign clinical trajectory, greater cognitive impairment, and higher APOE-ε4 allele frequency than those in the CBS-non-AD group (AUC, 0.80-0.87; P < .05). Serum NF-L levels distinguished PD from all PSP and CBS cases combined (AUC, 0.80; P < .05). These findings suggest that studies focusing on the PSP-RS subtype are likely to miss a large number of patients with underlying PSP tau pathology. Analysis of cerebrospinal fluid defined a distinct CBS-AD subtype. The PSP and CBS subtypes have distinct characteristics that may enhance their early diagnosis.

Optimizing Machine Learning Methods to Improve Predictive Models of Alzheimer's Disease.

Predicting clinical course of cognitive decline can boost clinical trials' power and improve our clinical decision-making. Machine learning (ML) algorithms are specifically designed for the purpose of prediction; however. identifying optimal features or algorithms is still a challenge. To investigate the accuracy of different ML methods and different features to classify cognitively normal (CN) individuals from Alzheimer's disease (AD) and to predict longitudinal outcome in participants with mild cognitive impairment (MCI). A total of 1,329 participants from the Alzheimer's Disease Neuroimaging Initiative (ADNI) were included: 424 CN, 656 MCI, and 249 AD individuals. Four feature-sets at baseline (hippocampal volume and volume of 47 cortical and subcortical regions with and without demographics and APOE4) and six machine learning methods (decision trees, support vector machines, K-nearest neighbor, ensemble linear discriminant, boosted trees, and random forests) were used to classify participants with normal cognition from participants with AD. Subsequently the model with best classification performance was used for predicting clinical outcome of MCI participants. Ensemble linear discriminant models using demographics and all volumetric magnetic resonance imaging measures as feature-set showed the best performance in classification of CN versus AD participants (accuracy = 92.8%, sensitivity = 95.8%, and specificity = 88.3%). Prediction accuracy of future conversion from MCI to AD for this ensemble linear discriminant at 6, 12, 24, 36, and 48 months was 63.8% (sensitivity = 74.4, specificity = 63.1), 68.9% (sensitivity = 75.9, specificity = 67.8), 74.9% (sensitivity = 71.5, specificity = 76.3), 75.3%, (sensitivity = 65.2, specificity = 79.7), and 77.0% (sensitivity = 59.6, specificity = 86.1), respectively. Machine learning models trained for classification of CN versus AD can improve our prediction ability of MCI conversion to AD.

Shared Latent Structures Between Imaging Features and Biomarkers in Early Stages of Alzheimer's Disease: A Predictive Study.

Magnetic resonance imaging (MRI) provides high resolution brain morphological information and is used as a biomarker in neurodegenerative diseases. Population studies of brain morphology often seek to identify pathological structural changes related to different diagnostic categories (e.g.: controls, mild cognitive impairment or dementia) which normally describe highly heterogeneous groups with a single categorical variable. Instead, multiple biomarkers are used as a proxy for pathology and are more powerful in capturing structural variability. Hence, using the joint modeling of brain morphology and biomarkers, we aim at describing structural changes related to any brain condition by means of few underlying processes. In this regard, we use a multivariate approach based on Projection to Latent Structures in its regression variant (PLSR) to study structural changes related to aging and AD pathology. MRI volumetric and cortical thickness measurements are used for brain morphology and cerebrospinal fluid (CSF) biomarkers (t-tau, p-tau and amyloid-beta) are used as a proxy for AD pathology. By relating both sets of measurements, PLSR finds a low-dimensional latent space describing AD pathological effects on brain structure. The proposed framework allows us to separately model aging effects on brain morphology as a confounder variable orthogonal to the pathological effect. The predictive power of the associated latent spaces (i.e., the capacity of predicting biomarker values) is assessed in a cross-validation framework.

Prediction of Alzheimer's disease dementia with MRI beyond the short-term: Implications for the design of predictive models.

Magnetic resonance imaging (MRI) volumetric measures have become a standard tool for the detection of incipient Alzheimer's Disease (AD) dementia in mild cognitive impairment (MCI). Focused on providing an earlier and more accurate diagnosis, sophisticated MRI machine learning algorithms have been developed over the recent years, most of them learning their non-disease patterns from MCI that remained stable over 2–3 years. In this work, we analyzed whether these stable MCI over short-term periods are actually appropriate training examples of non-disease patterns. To this aim, we compared the diagnosis of MCI patients at 2 and 5 years of follow-up and investigated its impact on the predictive performance of baseline volumetric MRI measures primarily involved in AD, i.e., hippocampal and entorhinal cortex volumes. Predictive power was evaluated in terms of the area under the ROC curve (AUC), sensitivity, and specificity in a trial sample of 248 MCI patients followed-up over 5 years. We further compared the sensitivity in those MCI that converted before 2 years and those that converted after 2 years. Our results indicate that 23% of the stable MCI at 2 years progressed in the next three years and that MRI volumetric measures are good predictors of conversion to AD dementia even at the mid-term, showing a better specificity and AUC as follow-up time increases. The combination of hippocampus and entorhinal cortex yielded an AUC that was significantly higher for the 5-year follow-up (AUC = 73% at 2 years vs. AUC = 84% at 5 years), as well as for specificity (56% vs. 71%). Sensitivity showed a non-significant slight decrease (81% vs. 78%). Remarkably, the performance of this model was comparable to machine learning models at the same follow-up times. MRI correctly identified most of the patients that converted after 2 years (with sensitivity >60%), and these patients showed a similar degree of abnormalities to those that converted before 2 years. This implies that most of the MCI patients that remained stable over short periods and subsequently progressed to AD dementia had evident atrophies at baseline. Therefore, machine learning models that use these patients to learn non-disease patterns are including an important fraction of patients with evident pathological changes related to the disease, something that might result in reduced performance and lack of biological interpretability.

Data Augmentation Based on Substituting Regional MRIs Volume Scores.

Due to difficulties in collecting sufficient training data, recent advances in neural-network-based methods have not been fully explored in the analysis of brain Magnetic Resonance Imaging (MRI). A possible solution to the limited-data issue is to augment the training set with synthetically generated data. In this paper, we propose a data augmentation strategy based on regional feature substitution. We demonstrate the advantages of this strategy with respect to training a simple neural-network-based classifier in predicting when individual youth transition from no-to-low to medium-to-heavy alcohol drinkers solely based on their volumetric MRI measurements. Based on 20-fold cross-validation, we generate more than one million synthetic samples from less than 500 subjects for each training run. The classifier achieves an accuracy of 74.1% in correctly distinguishing non-drinkers from drinkers at baseline and a 43.2% weighted accuracy in predicting the transition over a three year period (5-group classification task). Both accuracy scores are significantly better than training the classifier on the original dataset.