Alzheimer's Disease and Small Vessel Disease Differentially Affect White Matter Microstructure.

OBJECTIVE: Alzheimer's disease (AD) and cerebral small vessel disease (cSVD), the two most common causes of dementia, are characterized by white matter (WM) alterations diverging from the physiological changes occurring in healthy aging. Diffusion tensor imaging (DTI) is a valuable tool to quantify WM integrity non-invasively and identify the determinants of such alterations. Here, we investigated main effects and interactions of AD pathology, APOE-ε4, cSVD, and cardiovascular risk on spatial patterns of WM alterations in non-demented older adults. METHODS: Within the prospective European Prevention of Alzheimer's Dementia study, we selected 606 participants (64.9 ± 7.2 years, 376 females) with baseline cerebrospinal fluid samples of amyloid ß1-42 and p-Tau181 and MRI scans, including DTI scans. Longitudinal scans (mean follow-up time = 1.3 ± 0.5 years) were obtained in a subset (n = 223). WM integrity was assessed by extracting fractional anisotropy and mean diffusivity in relevant tracts. To identify the determinants of WM disruption, we performed a multimodel inference to identify the best linear mixed-effects model for each tract. RESULTS: AD pathology, APOE-ε4, cSVD burden, and cardiovascular risk were all associated with WM integrity within several tracts. While limbic tracts were mainly impacted by AD pathology and APOE-ε4, commissural, associative, and projection tract integrity was more related to cSVD burden and cardiovascular risk. AD pathology and cSVD did not show any significant interaction effect. INTERPRETATION: Our results suggest that AD pathology and cSVD exert independent and spatially different effects on WM microstructure, supporting the role of DTI in disease monitoring and suggesting independent targets for preventive medicine approaches.

Investigating reliable amyloid accumulation in Centiloids: Results from the AMYPAD Prognostic and Natural History Study.

INTRODUCTION: To support clinical trial designs focused on early interventions, our study determined reliable early amyloid-ß (Aß) accumulation based on Centiloids (CL) in pre-dementia populations. METHODS: A total of 1032 participants from the Amyloid Imaging to Prevent Alzheimer's Disease-Prognostic and Natural History Study (AMYPAD-PNHS) and Insight46 who underwent [18F]flutemetamol, [18F]florbetaben or [18F]florbetapir amyloid-PET were included. A normative strategy was used to define reliable accumulation by estimating the 95th percentile of longitudinal measurements in sub-populations (NPNHS = 101/750, NInsight46 = 35/382) expected to remain stable over time. The baseline CL threshold that optimally predicts future accumulation was investigated using precision-recall analyses. Accumulation rates were examined using linear mixed-effect models. RESULTS: Reliable accumulation in the PNHS was estimated to occur at >3.0 CL/year. Baseline CL of 16 [12,19] best predicted future Aß-accumulators. Rates of amyloid accumulation were tracer-independent, lower for APOE ε4 non-carriers, and for subjects with higher levels of education. DISCUSSION: Our results support a 12-20 CL window for inclusion into early secondary prevention studies. Reliable accumulation definition warrants further investigations.

A systematic review of (semi-)automatic quality control of T1-weighted MRI scans.

PURPOSE: Artifacts in magnetic resonance imaging (MRI) scans degrade image quality and thus negatively affect the outcome measures of clinical and research scanning. Considering the time-consuming and subjective nature of visual quality control (QC), multiple (semi-)automatic QC algorithms have been developed. This systematic review presents an overview of the available (semi-)automatic QC algorithms and software packages designed for raw, structural T1-weighted (T1w) MRI datasets. The objective of this review was to identify the differences among these algorithms in terms of their features of interest, performance, and benchmarks. METHODS: We queried PubMed, EMBASE (Ovid), and Web of Science databases on the fifth of January 2023, and cross-checked reference lists of retrieved papers. Bias assessment was performed using PROBAST (Prediction model Risk Of Bias ASsessment Tool). RESULTS: A total of 18 distinct algorithms were identified, demonstrating significant variations in methods, features, datasets, and benchmarks. The algorithms were categorized into rule-based, classical machine learning-based, and deep learning-based approaches. Numerous unique features were defined, which can be roughly divided into features capturing entropy, contrast, and normative measures. CONCLUSION: Due to dataset-specific optimization, it is challenging to draw broad conclusions about comparative performance. Additionally, large variations exist in the used datasets and benchmarks, further hindering direct algorithm comparison. The findings emphasize the need for standardization and comparative studies for advancing QC in MR imaging. Efforts should focus on identifying a dataset-independent measure as well as algorithm-independent methods for assessing the relative performance of different approaches.

The Open-Access European Prevention of Alzheimer's Dementia (EPAD) MRI dataset and processing workflow.

The European Prevention of Alzheimer Dementia (EPAD) is a multi-center study that aims to characterize the preclinical and prodromal stages of Alzheimer's Disease. The EPAD imaging dataset includes core (3D T1w, 3D FLAIR) and advanced (ASL, diffusion MRI, and resting-state fMRI) MRI sequences. Here, we give an overview of the semi-automatic multimodal and multisite pipeline that we developed to curate, preprocess, quality control (QC), and compute image-derived phenotypes (IDPs) from the EPAD MRI dataset. This pipeline harmonizes DICOM data structure across sites and performs standardized MRI preprocessing steps. A semi-automated MRI QC procedure was implemented to visualize and flag MRI images next to site-specific distributions of QC features - i.e. metrics that represent image quality. The value of each of these QC features was evaluated through comparison with visual assessment and step-wise parameter selection based on logistic regression. IDPs were computed from 5 different MRI modalities and their sanity and potential clinical relevance were ascertained by assessing their relationship with biological markers of aging and dementia. The EPAD v1500.0 data release encompassed core structural scans from 1356 participants 842 fMRI, 831 dMRI, and 858 ASL scans. From 1356 3D T1w images, we identified 17 images with poor quality and 61 with moderate quality. Five QC features - Signal to Noise Ratio (SNR), Contrast to Noise Ratio (CNR), Coefficient of Joint Variation (CJV), Foreground-Background energy Ratio (FBER), and Image Quality Rate (IQR) - were selected as the most informative on image quality by comparison with visual assessment. The multimodal IDPs showed greater impairment in associations with age and dementia biomarkers, demonstrating the potential of the dataset for future clinical analyses.

Impact of cerebral blood flow and amyloid load on SUVR bias.

BACKGROUND: Despite its widespread use, the semi-quantitative standardized uptake value ratio (SUVR) may be biased compared with the distribution volume ratio (DVR). This bias may be partially explained by changes in cerebral blood flow (CBF) and is likely to be also dependent on the extent of the underlying amyloid-ß (Aß) burden. This study aimed to compare SUVR with DVR and to evaluate the effects of underlying Aß burden and CBF on bias in SUVR in mainly cognitively unimpaired participants. Participants were scanned according to a dual-time window protocol, with either [18F]flutemetamol (N = 90) or [18F]florbetaben (N = 31). The validated basisfunction-based implementation of the two-step simplified reference tissue model was used to derive DVR and R1 parametric images, and SUVR was calculated from 90 to 110 min post-injection, all with the cerebellar grey matter as reference tissue. First, linear regression and Bland-Altman analyses were used to compare (regional) SUVR with DVR. Then, generalized linear models were applied to evaluate whether (bias in) SUVR relative to DVR could be explained by R1 for the global cortical average (GCA), precuneus, posterior cingulate, and orbitofrontal region. RESULTS: Despite high correlations (GCA: R2 ≥ 0.85), large overestimation and proportional bias of SUVR relative to DVR was observed. Negative associations were observed between both SUVR or SUVRbias and R1, albeit non-significant. CONCLUSION: The present findings demonstrate that bias in SUVR relative to DVR is strongly related to underlying Aß burden. Furthermore, in a cohort consisting mainly of cognitively unimpaired individuals, the effect of relative CBF on bias in SUVR appears limited. EudraCT Number: 2018-002277-22, registered on: 25-06-2018.

An Automatic Gait Analysis Pipeline for Wearable Sensors: A Pilot Study in Parkinson's Disease.

The use of wearable sensors allows continuous recordings of physical activity from participants in free-living or at-home clinical studies. The large amount of data collected demands automatic analysis pipelines to extract gait parameters that can be used as clinical endpoints. We introduce a deep learning-based automatic pipeline for wearables that processes tri-axial accelerometry data and extracts gait events-bout segmentation, initial contact (IC), and final contact (FC)-from a single sensor located at either the lower back (near L5), shin or wrist. The gait events detected are posteriorly used for gait parameter estimation, such as step time, length, and symmetry. We report results from a leave-one-subject-out (LOSO) validation on a pilot study dataset of five participants clinically diagnosed with Parkinson's disease (PD) and six healthy controls (HC). Participants wore sensors at three body locations and walked on a pressure-sensing walkway to obtain reference gait data. Mean absolute errors (MAE) for the IC events ranged from 22.82 to 33.09 milliseconds (msecs) for the lower back sensor while for the shin and wrist sensors, MAE ranges were 28.56-64.66 and 40.19-72.50 msecs, respectively. For the FC-event detection, MAE ranges were 29.06-48.42, 40.19-72.70 and 36.06-60.18 msecs for the lumbar, wrist and shin sensors, respectively. Intraclass correlation coefficients, ICC(2,k), between the estimated parameters and the reference data resulted in good-to-excellent agreement (ICC ≥ 0.84) for the lumbar and shin sensors, excluding the double support time (ICC = 0.37 lumbar and 0.38 shin) and swing time (ICC = 0.55 lumbar and 0.59 shin). The wrist sensor also showed good agreements, but the ICCs were lower overall than for the other two sensors. Our proposed analysis pipeline has the potential to extract up to 100 gait-related parameters, and we expect our contribution will further support developments in the fields of wearable sensors, digital health, and remote monitoring in clinical trials.

Regional Brain and Spinal Cord Volume Loss in Spinocerebellar Ataxia Type 3.

BACKGROUND: Given that new therapeutic options for spinocerebellar ataxias are on the horizon, there is a need for markers that reflect disease-related alterations, in particular, in the preataxic stage, in which clinical scales are lacking sensitivity. OBJECTIVE: The objective of this study was to quantify regional brain volumes and upper cervical spinal cord areas in spinocerebellar ataxia type 3 in vivo across the entire time course of the disease. METHODS: We applied a brain segmentation approach that included a lobular subsegmentation of the cerebellum to magnetic resonance images of 210 ataxic and 48 preataxic spinocerebellar ataxia type 3 mutation carriers and 63 healthy controls. In addition, cervical cord cross-sectional areas were determined at 2 levels. RESULTS: The metrics of cervical spinal cord segments C3 and C2, medulla oblongata, pons, and pallidum, and the cerebellar anterior lobe were reduced in preataxic mutation carriers compared with controls. Those of cervical spinal cord segments C2 and C3, medulla oblongata, pons, midbrain, cerebellar lobules crus II and X, cerebellar white matter, and pallidum were reduced in ataxic compared with nonataxic carriers. Of all metrics studied, pontine volume showed the steepest decline across the disease course. It covaried with ataxia severity, CAG repeat length, and age. The multivariate model derived from this analysis explained 46.33% of the variance of pontine volume. CONCLUSION: Regional brain and spinal cord tissue loss in spinocerebellar ataxia type 3 starts before ataxia onset. Pontine volume appears to be the most promising imaging biomarker candidate for interventional trials that aim at slowing the progression of spinocerebellar ataxia type 3. © 2021 The Authors. Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.

Application of the ATN classification scheme in a population without dementia: Findings from the EPAD cohort.

BACKGROUND: We classified non-demented European Prevention of Alzheimer's Dementia (EPAD) participants through the amyloid/tau/neurodegeneration (ATN) scheme and assessed their neuropsychological and imaging profiles. MATERIALS AND METHODS: From 1500 EPAD participants, 312 were excluded. Cerebrospinal fluid cut-offs of 1000 pg/mL for amyloid beta (Aß)1-42 and 27 pg/mL for p-tau181 were validated using Gaussian mixture models. Given strong correlation of p-tau and t-tau (R2  = 0.98, P < 0.001), neurodegeneration was defined by age-adjusted hippocampal volume. Multinomial regressions were used to test whether neuropsychological tests and regional brain volumes could distinguish ATN stages. RESULTS: Age was 65 ± 7 years, with 58% females and 38% apolipoprotein E (APOE) ε4 carriers; 57.1% were A-T-N-, 32.5% were in the Alzheimer's disease (AD) continuum, and 10.4% suspected non-Alzheimer's pathology. Age and cerebrovascular burden progressed with biomarker positivity (P < 0.001). Cognitive dysfunction appeared with T+. Paradoxically higher regional gray matter volumes were observed in A+T-N- compared to A-T-N- (P < 0.001). DISCUSSION: In non-demented individuals along the AD continuum, p-tau drives cognitive dysfunction. Memory and language domains are affected in the earliest stages.

Imaging markers associated with the development of post-stroke depression and apathy: Results of the Cognition and Affect after Stroke - a Prospective Evaluation of Risks study.

INTRODUCTION: It has been suggested that the development of post-stroke apathy (PSA) and depression (PSD) may be more strongly associated with generalised brain pathology, rather than the stroke lesion itself. The present study aimed to investigate associations between imaging markers of lesion-related and generalised brain pathology and the development of PSA and PSD during a one-year follow-up. PATIENTS AND METHODS: In a prospective cohort study, 188 stroke patients received 3-Tesla MRI at baseline (three months post-stroke) for evaluation of lesion-related, vascular, and degenerative brain pathology. Presence of lacunes, microbleeds, white matter hyperintensities, and enlarged perivascular spaces was summed to provide a measure of total cerebral small vessel disease (cSVD) burden (range 0-4). The Mini International Neuropsychiatric Interview and Apathy Evaluation Scale were administered at baseline and repeated at 6- and 12-month follow-up to define presence of PSD and PSA, respectively. RESULTS: Population-averaged logistic regression models showed that global brain atrophy and severe cSVD burden (score 3-4) were significantly associated with the odds of having PSA (ORGEE 5.33, 95% CI 1.99-14.25 and 3.04, 95% CI 1.20-7.69, respectively), independent of stroke lesion volume and co-morbid PSD. Medium cSVD burden (score 2) was significantly associated with the odds of having PSD (ORGEE 2.92, 95% CI 1.09-7.78), independent of stroke lesion volume, co-morbid PSA, and pre-stroke depression. No associations were found with lesion-related markers. CONCLUSIONS: The results suggest that generalised degenerative and vascular brain pathology, rather than lesion-related pathology, is an important predictor for the development of PSA, and less strongly for PSD.

Quantitative amyloid PET in Alzheimer's disease: the AMYPAD prognostic and natural history study.

INTRODUCTION: The Amyloid Imaging to Prevent Alzheimer's Disease (AMYPAD) Prognostic and Natural History Study (PNHS) aims at understanding the role of amyloid imaging in the earliest stages of Alzheimer's disease (AD). AMYPAD PNHS adds (semi-)quantitative amyloid PET imaging to several European parent cohorts (PCs) to predict AD-related progression as well as address methodological challenges in amyloid PET. METHODS: AMYPAD PNHS is an open-label, prospective, multi-center, cohort study recruiting from multiple PCs. Around 2000 participants will undergo baseline amyloid positron emission tomography (PET), half of whom will be invited for a follow-up PET 12 at least 12 months later. RESULTS: Primary include several amyloid PET measurements (Centiloid, SUVr, BPND , R1 ), and secondary are their changes from baseline, relationship to other amyloid markers (cerebrospinal fluid and visual assessment), and predictive value of AD-related decline. EXPECTED IMPACT: Determining the role of amyloid PET for the understanding of this complex disease and potentially improving secondary prevention trials.

The impact of automated hippocampal volumetry on diagnostic confidence in patients with suspected Alzheimer's disease: A European Alzheimer's Disease Consortium study.

INTRODUCTION: Hippocampal volume is a core biomarker of Alzheimer's disease (AD). However, its contribution over the standard diagnostic workup is unclear. METHODS: Three hundred fifty-six patients, under clinical evaluation for cognitive impairment, with suspected AD and Mini-Mental State Examination ≥20, were recruited across 17 European memory clinics. After the traditional diagnostic workup, diagnostic confidence of AD pathology (DCAD) was estimated by the physicians in charge. The latter were provided with the results of automated hippocampal volumetry in standardized format and DCAD was reassessed. RESULTS: An increment of one interquartile range in hippocampal volume was associated with a mean change of DCAD of -8.0% (95% credible interval: [-11.5, -5.0]). Automated hippocampal volumetry showed a statistically significant impact on DCAD beyond the contributions of neuropsychology, 18F-fluorodeoxyglucose positron emission tomography/single-photon emission computed tomography, and cerebrospinal fluid markers (-8.5, CrI: [-11.5, -5.6]; -14.1, CrI: [-19.3, -8.8]; -10.6, CrI: [-14.6, -6.1], respectively). DISCUSSION: There is a measurable effect of hippocampal volume on DCAD even when used on top of the traditional diagnostic workup.

Enrichment of clinical trials in MCI due to AD using markers of amyloid and neurodegeneration.

OBJECTIVE: To investigate the effect of enriching mild cognitive impairment (MCI) clinical trials using combined markers of amyloid pathology and neurodegeneration. METHODS: We evaluate an implementation of the recent National Institute for Aging-Alzheimer's Association (NIA-AA) diagnostic criteria for MCI due to Alzheimer disease (AD) as inclusion criteria in clinical trials and assess the effect of enrichment with amyloid (A+), neurodegeneration (N+), and their combination (A+N+) on the rate of clinical progression, required sample sizes, and estimates of trial time and cost. RESULTS: Enrichment based on an individual marker (A+ or N+) substantially improves all assessed trial characteristics. Combined enrichment (A+N+) further improves these results with a reduction in required sample sizes by 45% to 60%, depending on the endpoint. CONCLUSIONS: Operationalizing the NIA-AA diagnostic criteria for clinical trial screening has the potential to substantially improve the statistical power of trials in MCI due to AD by identifying a more rapidly progressing patient population.

A novel bone suppression method that improves lung nodule detection : Suppressing dedicated bone shadows in radiographs while preserving the remaining signal.

PURPOSE: Suppressing thoracic bone shadows in chest radiographs has been previously reported to improve the detection rates for solid lung nodules, however at the cost of increased false detection rates. These bone suppression methods are based on an artificial neural network that was trained using dual-energy subtraction images in order to mimic their appearance. METHOD: Here, a novel approach is followed where all bone shadows crossing the lung field are suppressed sequentially leaving the intercostal space unaffected. Given a contour delineating a bone, its image region is spatially transferred to separate normal image gradient components from tangential component. Smoothing the normal partial gradient along the contour results in a reconstruction of the image representing the bone shadow only, because all other overlaid signals tend to cancel out each other in this representation. RESULTS: The method works even with highly contrasted overlaid objects such as a pacemaker. The approach was validated in a reader study with two experienced chest radiologists, and these images helped improving both the sensitivity and the specificity of the readers for the detection and localization of solid lung nodules. The AUC improved significantly from 0.596 to 0.655 on a basis of 146 images from patients and normals with a total of 123 confirmed lung nodules. CONCLUSION: Subtracting all reconstructed bone shadows from the original image results in a soft image where lung nodules are no longer obscured by bone shadows. Both the sensitivity and the specificity of experienced radiologists increased.

Discriminative dictionary learning for abdominal multi-organ segmentation.

An automated segmentation method is presented for multi-organ segmentation in abdominal CT images. Dictionary learning and sparse coding techniques are used in the proposed method to generate target specific priors for segmentation. The method simultaneously learns dictionaries which have reconstructive power and classifiers which have discriminative ability from a set of selected atlases. Based on the learnt dictionaries and classifiers, probabilistic atlases are then generated to provide priors for the segmentation of unseen target images. The final segmentation is obtained by applying a post-processing step based on a graph-cuts method. In addition, this paper proposes a voxel-wise local atlas selection strategy to deal with high inter-subject variation in abdominal CT images. The segmentation performance of the proposed method with different atlas selection strategies are also compared. Our proposed method has been evaluated on a database of 150 abdominal CT images and achieves a promising segmentation performance with Dice overlap values of 94.9%, 93.6%, 71.1%, and 92.5% for liver, kidneys, pancreas, and spleen, respectively.

Geodesic Information Flows: Spatially-Variant Graphs and Their Application to Segmentation and Fusion.

Clinical annotations, such as voxel-wise binary or probabilistic tissue segmentations, structural parcellations, pathological regions-of-interest and anatomical landmarks are key to many clinical studies. However, due to the time consuming nature of manually generating these annotations, they tend to be scarce and limited to small subsets of data. This work explores a novel framework to propagate voxel-wise annotations between morphologically dissimilar images by diffusing and mapping the available examples through intermediate steps. A spatially-variant graph structure connecting morphologically similar subjects is introduced over a database of images, enabling the gradual diffusion of information to all the subjects, even in the presence of large-scale morphological variability. We illustrate the utility of the proposed framework on two example applications: brain parcellation using categorical labels and tissue segmentation using probabilistic features. The application of the proposed method to categorical label fusion showed highly statistically significant improvements when compared to state-of-the-art methodologies. Significant improvements were also observed when applying the proposed framework to probabilistic tissue segmentation of both synthetic and real data, mainly in the presence of large morphological variability.

Predicting progression from cognitive impairment to Alzheimer's disease with the Disease State Index.

We evaluated the performance of the Disease State Index (DSI) method when predicting progression to Alzheimer's disease (AD) in patients with subjective cognitive impairment (SCI), amnestic or non-amnestic mild cognitive impairment (aMCI, naMCI). The DSI model measures patients' similarity to diagnosed cases based on available data, such as cognitive tests, the APOE genotype, CSF biomarkers and MRI. We applied the DSI model to data from the DESCRIPA cohort, where non-demented patients (N=775) with different subtypes of cognitive impairment were followed for 1 to 5 years. Classification accuracies for the subgroups were calculated with the DSI using leave-one-out crossvalidation. The DSI's classification accuracy in predicting progression to AD was 0.75 (AUC=0.83) in the total population, 0.70 (AUC=0.77) for aMCI and 0.71 (AUC=0.76) for naMCI. For a subset of approximately half of the patients with high or low DSI values, accuracy reached 0.86 (all), 0.78 (aMCI), and 0.85 (naMCI). For patients with MRI or CSF biomarker data available, theywere 0.78 (all), 0.76 (aMCI) and 0.76 (naMCI), while for clear cases the accuracies rose to 0.90 (all), 0.83 (aMCI) and 0.91 (naMCI). The results show that the DSI model can distinguish between clear and ambiguous cases, assess the severity of the disease and also provide information on the effectiveness of different biomarkers. While a specific test or biomarker may confound analysis for an individual patient, combining several different types of tests and biomarkers could be able to reveal the trajectory of the disease and improve the prediction of AD progression.

Coalition Against Major Diseases/European Medicines Agency biomarker qualification of hippocampal volume for enrichment of clinical trials in predementia stages of Alzheimer's disease.

BACKGROUND: Regulatory qualification of a biomarker for a defined context of use provides scientifically robust assurances to sponsors and regulators that accelerate appropriate adoption of biomarkers into drug development. METHODS: The Coalition Against Major Diseases submitted a dossier to the Scientific Advice Working Party of the European Medicines Agency requesting a qualification opinion on the use of hippocampal volume as a biomarker for enriching clinical trials in subjects with mild cognitive impairment, incorporating a scientific rationale, a literature review and a de novo analysis of Alzheimer's Disease Neuroimaging Initiative data. RESULTS: The literature review and de novo analysis were consistent with the proposed context of use, and the Committee for Medicinal Products for Human Use released an opinion in November 2011. CONCLUSIONS: We summarize the scientific rationale and the data that supported the first qualification of an imaging biomarker by the European Medicines Agency.

Robustness of automated hippocampal volumetry across magnetic resonance field strengths and repeat images.

BACKGROUND: Low HCV has recently been qualified by the European Medicines Agency as a biomarker for enrichment of clinical trials in predementia stages of Alzheimer's disease. For automated methods to meet the necessary regulatory requirements, it is essential they be standardized and their performance be well characterized. METHODS: The within-image and between-field strength reproducibility of automated hippocampal volumetry using the Learning Embeddings for Atlas Propagation (or LEAP) algorithm was assessed on 153 Alzheimer's Disease Neuroimaging Initiative subjects. RESULTS: Tests/retests at 1.5 T and 3 T, and a comparison between 1.5 T and 3 T, yielded average unsigned variabilities in HCVs of 1.51%, 1.52%, and 2.68%. A small bias between field strengths (mean signed difference, 1.17%; standard deviation, 3.07%) was observed. CONCLUSIONS: The measured reproducibility characteristics confirm the suitability of using automated magnetic resonance imaging analyses to assess HCVs quantitatively and to represent a fundamental characterization that is critical to meet the regulatory requirements for using hippocampal volumetry in clinical trials and health care.

Multiple instance learning for classification of dementia in brain MRI.

Machine learning techniques have been widely used to detect morphological abnormalities from structural brain magnetic resonance imaging data and to support the diagnosis of neurological diseases such as dementia. In this paper, we propose to use a multiple instance learning (MIL) method in an application for the detection of Alzheimer's disease (AD) and its prodromal stage mild cognitive impairment (MCI). In our work, local intensity patches are extracted as features. However, not all the patches extracted from patients with dementia are equally affected by the disease and some of them may not be characteristic of morphology associated with the disease. Therefore, there is some ambiguity in assigning disease labels to these patches. The problem of the ambiguous training labels can be addressed by weakly supervised learning techniques such as MIL. A graph is built for each image to exploit the relationships among the patches and then to solve the MIL problem. The constructed graphs contain information about the appearances of patches and the relationships among them, which can reflect the inherent structures of images and aids the classification. Using the baseline MR images of 834 subjects from the ADNI study, the proposed method can achieve a classification accuracy of 89% between AD patients and healthy controls, and 70% between patients defined as stable MCI and progressive MCI in a leave-one-out cross validation. Compared with two state-of-the-art methods using the same dataset, the proposed method can achieve similar or improved results, providing an alternative framework for the detection and prediction of neurodegenerative diseases.

Hierarchical manifold learning for regional image analysis.

We present a novel method of hierarchical manifold learning which aims to automatically discover regional properties of image datasets. While traditional manifold learning methods have become widely used for dimensionality reduction in medical imaging, they suffer from only being able to consider whole images as single data points. We extend conventional techniques by additionally examining local variations, in order to produce spatially-varying manifold embeddings that characterize a given dataset. This involves constructing manifolds in a hierarchy of image patches of increasing granularity, while ensuring consistency between hierarchy levels. We demonstrate the utility of our method in two very different settings: 1) to learn the regional correlations in motion within a sequence of time-resolved MR images of the thoracic cavity; 2) to find discriminative regions of 3-D brain MR images associated with neurodegenerative disease.