ABSTRACT
Numerous studies have proposed biomarkers based on magnetic resonance imaging (MRI) to detect and predict the risk of evolution toward Alzheimer’s disease (AD). While anatomical MRI captures structural alterations, studies demonstrated the ability of diffusion MRI to capture microstructural modifications at an earlier stage. Several methods have focused on hippocampus structure to detect AD. To date, the patch-based grading framework provides the best biomarker based on the hippocampus. However, this structure is complex since the hippocampus is divided into several heterogeneous subfields not equally impacted by AD. Former in-vivo imaging studies only investigated structural alterations of these subfields using volumetric measurements and microstructural modifications with mean diffusivity measurements. The aim of our work is to study the efficiency of hippocampal subfields compared to the whole hippocampus structure with a multimodal patch-based framework that enables to capture subtler structural and microstructural alterations. To this end, we analyze the significance of the different hippocampal subfields for AD diagnosis and prognosis with volumetric, diffusivity measurements and a novel multimodal patch-based grading framework that combines structural and diffusion MRI. The experiments conducted in this work showed that the whole hippocampus provides the most discriminant biomarkers for advanced AD detection while biomarkers applied into subiculum obtain the best results for AD prediction, improving by 2% the accuracy compared to the whole hippocampus.
Introduction
Alzheimer’s disease (AD) is an irreversible neurodegenerative process leading to mental dysfunctions. Subjects presenting Mild Cognitive Impairment (MCI) have higher risk to develop AD1. To study the preclinical phase of the disease, the Alzheimer’s disease neuroimaging initiative (ADNI) was set up based on two MCI definitions: early MCI (eMCI) and late MCI (lMCI). eMCI represents subjects with cognitive impairment milder than lMCI which is composed of amnesic MCI2. Such clinical symptoms caused by changes like synaptic and neuronal losses that lead to structural and micro-structural alterations. Neuroimaging studies performed on AD subjects revealed that brain structure alterations are advanced when diagnosis is established and emphasized the need to study the early stages of the disease.
The improvement of medical imaging techniques such as magnetic resonance imaging (MRI) enabled the development of efficient biomarkers detecting alterations caused by AD3. Over the past years, many methods have been proposed to perform automatic detection of alterations associated with AD. First, studies proposed methods based on specific regions of interest (ROI) capturing alterations at an anatomical scale. Among structures impacted by AD, previous investigations focused on hippocampus (HIPP)4–6, entorhinal cortex (EC)7–9, parahippocampal gyrus, amygdala10 or parietal lobe11, 12. Alterations on these structures are usually estimated using volume13, 14, shape15, 16, or cortical thickness17, 18 measurements. Second, beside ROI-based methods, whole brain analysis performed on structural MRI (s-MRI) have been proposed to detect areas impacted by AD at a voxel scale. These methods are usually based on voxel-based morphometry (VBM) or tensor-based morphometry (TBM) frameworks19. It is interesting to note that both VBM and ROI-based studies confirmed that medial temporal lobe is a key area to detect the first signs of AD20–25. These studies also showed that HIPP is one of the earliest region altered by AD in the medial temporal lobe26. Moreover, HIPP volume is one of the criterion that can be used to confirm the diagnosis of AD in clinical routines27. Recently, advanced methods were proposed to capture subtler structural alterations of HIPP9, 28–30. Those techniques demonstrated an increase of detection and prediction performances at different AD stages compared to volume-based methods30. Among them, patch-based grading (PBG) methods demonstrated competitive results to detect the earliest stages of AD before a clinical diagnosis can be made9, 29, 31. The main idea of this approach is to capture inter-subject pattern similarities via non-local comparisons between two groups of subjects. Such methods have shown their ability to predict AD more than seven years before the conversion to dementia32 and might help for a differential diagnosis33.
Thus, the hippocampus has been one of the most studied structures in order to diagnose AD. However, this structure is complex and not homogeneous. HIPP is subdivided into several subfields, each one having specific characteristics. The terminology differs across segmentation protocol34 but the most recognized definition35 mainly divides HIPP into the subiculum, the cornu ammonis (CA1/2/3/4), and the dentrate gyrus (DG). The CA1 subfield represents the biggest area in the hippocampus. It is composed by different layers called the stratum radiatum (SR), the stratum lacunosum (SL), the stratum molecular (SM), and the stratum pyramidale (SP). Furthermore, hippocampal subfields are not equally impacted by AD36–42. Indeed, several MRI studies demonstrated that subfields are impacted differently according to AD stages. Postmortem, and in vivo imaging studies showed that CA1SR-L-M are the subfields impacted with the greatest atrophy in advanced AD38, 39, 41. Recently, it has been shown that subiculum is the earliest affected hippocampal region42, 43. These studies indicate that a subfield analysis of HIPP alterations at finer scale could provide better tool for AD detection and prediction.
Although structural MRI is a valuable imaging technique to measure global structural modifications, such modality is not able to capture microstructural degradation. However, the microstructural modifications caused by AD are considered to occur before the atrophy measured by structural MRI. Therefore, diffusion MRI (d-MRI) appears as a potential candidate to detect the earliest sign of AD. Several diffusion tensor imaging (DTI) studies proposed automatic methods to detect modifications of diffusion parameters into the whole white matter using machine learning44–46. Others studies showed modifications of diffusion parameters for AD patients into specific white matter structures such as corpus callosum47, 48, fornix49, cingulum47 and also in gray matter tissue such as hippocampus50. More advanced d-MRI studies using brain connectivity and fiber tracking have been proposed to extract features describing axonal fibers alterations49, 51, 52. Finally, it has been shown that hippocampal mean diffusivity (MD) is correlated to pathology progression and thus could be used as an efficient biomarker of AD53. Moreover, it was demonstrated that MD increases with the development of AD in the gray matter54–56. In a previous work, we showed that patch-based features applied on DTI demonstrated competitive performances to classify the early stages of AD57. Methods proposing to fuse d-MRI and s-MRI biomarkers was developed using the complementarity of these two MRI modalities58, 59. Recently, a study combining volumetric measurements and mean diffusivity of HIPP subfields demonstrated that CA1 and subiculum are the most impacted in late AD stage43. These studies showed the complementarity of s-MRI and d-MRI to capture early alteration led by AD.
All these elements indicate that multimodal analysis of hippocampal subfields using an advanced image analysis framework could provide valuable tool to improve AD detection and prediction. Consequently, in this paper, we propose to study hippocampal subfields efficiencies using s-MRI and d-MRI modalities. We have developed a novel multimodal patch-based grading fusion scheme to better capture such structural and microstructural alterations. First, we compare the performance of our novel method with volume and MD within the whole hippocampus. Second, we demonstrate state-of-the-art performances compared to more advanced d-MRI based methods. Finally, we study the efficiency of hippocampal subfields to improve AD detection and prediction with volume, MD and our multimodal patch-based grading method. Our results demonstrate that the study of hippocampus at finner scale improves AD prediction. Indeed, the experiments show that biomarkers based on whole hippocampus obtain best results for AD detection but biomarkers based on subiculum obtain best results for AD prediction.
Materials
Dataset
Data used in this work were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset (http://adni.loni.ucla.edu). ADNI is a North American campaign launched in 2003 with aims to provide MRI, positron emission tomography scans, clinical neurological measures and other biomarkers. This dataset includes AD patients, MCI and control normal (CN) subjects. The group of MCI is composed of subjects who have abnormal memory dysfunctions. In this work we used data from the ADNI-2 campaign that proposes eMCI and lMCI stages. The eMCI and lMCI subgroups were obtained with the Wechsler Scale-Revised Logistical Memory I and II tests in accordance with the education levels of each subject. ADNI-2 provides T1-weighted (T1w) MRI and DTI scans for 54 CN, 79 eMCI, 39 lMCI and 47 AD subjects. Only patients whose have T1w and DTI were selected in our work. Hence, in this work we used 52 CN, 99 MCI composed of 65 eMCI, 34 lMCI and 38 AD instead of the whole initial ADNI-2 dataset. Table 1 shows the distribution of the data for each group. The s-MRI and d-MRI scans used for all considered subjects in this study were acquired with the same protocol (https://adni.loni.usc.edu/wp-content/uploads/2010/05/ADNI2_GE_3T_22.0_T2.pdf). T1w MRI acquisition protocol had been done with the 3D accelerated sagittal IR-SPGR, according to the ADNI protocol60. The d-MRI were composed of 46 separate angles, 5 T2-weighted images with no diffusion sensitization (b0 images) and 41 directions (b=1000s/mm2). The d-MRI protocol was chosen to optimize the signal-to-noise ratio in a fixed scan time61. The native resolution of s-MRI and d-MRI was set to 1mm3 and 2mm3, respectively.
MRI processing
T1w images were processed using the volBrain system62 (http://volbrain.upv.es). This system is based on an advanced pipeline providing automatic segmentation of different brain structures from T1w MRI. The preprocessing is based on: (a) a denoising step with an adaptive non-local mean filter63, (b) an affine registration in the MNI space64, (c) a correction of the image inhomogeneities65 and (d) an intensity normalization.
Afterwards, segmentation of hippocampal subfields was performed with HIPS66 based on a combination of non-linear registration and patch-based label fusion67. This method uses a training library based on a dataset composed of high resolution T1w images manually labeled according to the protocol proposed by Winterburn et al. (2013)35. To perform the segmentation, the ADNI images are up-sampled with a local adaptive super resolution method to fit in the training image resolution68. The method provides automatic segmentation of hippocampal subfields gathered into 5 labels: Subiculum, CA1SP, CA1SR-L-M, CA2-3 and CA4/DG (see Figure 1). Finally, an estimation of the total intra-cranial volume is performed69.
DTI processing
The preprocessing of the diffusion weighted images is based on: (a) a denoising with a LPCA filter70, (b) a correction of the head motion using an affine registration and (c) an affine and a non-rigid registration to the T1w MRI in the MNI space64. Afterwards, a diffusion tensor model71 is fitted at each voxel using Dipy library72. To analyze microstructural modifications, the MD is estimated within each hippocampal subfield and the whole HIPP structure with the segmentation described in the previous section. MD is defined as where λ1, λ2, λ3 are the three eigenvalues of the fitted tensor.
Finally, a quality control was proceeded to exclude data presenting miss-segmentation or miss-registration after image preprocessing step.
Methods
Patch-based grading
Patch-based grading was firstly proposed for s-MRI9. The main idea of this exemplar-based method is to use the capability of patch-based techniques to capture subtle signal modifications related to anatomical degradations caused by AD. To date, PBG methods demonstrate state-of-the-art performances to detect the earliest stage of AD73. To determine the pathological status of the subject under study, PBG method estimates at each voxel the state of cerebral tissues by a similarity measurement. This measurement is performed between the anatomical pattern of the subject under study and those extracted from two training populations, one healthy and another one unhealthy.
First, a training library T composed of two datasets of images is built: one with images from CN subjects and the other one from AD patients. Next, for each voxel xi of the region of interest in the considered subject x, PBG method produces a weak classifier denoted gxi. This weak classifier provides a surrogate of the pathological grading at the considered position. The weak classifier is computed using a measurement of the similarity between the patch Pxi surrounding the voxel xi belonging to the image under study and a set Kxi of the closest patches extracted from the library T. The most similar patches are found using an approximative nearest neighbor method74. The grading value gxi at xi is defined as: where Ptj is the patch surrounding the voxel j belonging to the training template t ∈ T. w(xi, tj) is the weight assigned to the pathological status pt of the training image t. We estimate w such as: where and ε → 0. The pathological status pt is set to −1 for patches extracted from AD patient and to 1 for patches extracted from CN subject. Therefore, PBG method provides at each voxel a score representing an estimation of the alterations caused by AD. Consequently, cerebral tissues strongly altered by AD have grading values close to −1 contrary to healthy one with scores close to 1.
Multimodal patch-based grading fusion
Patch-based method presented in the previous section was firstly designed to capture structural alterations in T1w MRI. Recently, we proposed to extent this method to DTI modality in order to detect microstructural modifications57. We showed the efficiency of MD grading to improve the classification of the early stages of AD.
In this study, we propose a new framework to perform multimodal patch-based grading (MPBG). To this end, we developed adaptive fusion of grading maps derived from different modalities (see example of grading maps on Fig. 2). As shown in the following, this fusion provides more robust and accurate biomarkers compared to monomodal PBG biomarkers.
First, as in previous section, for each modality a training library of CN and AD subjects is built. Next, at each voxel within the ROI of the considered subject and for each modality, a set K of most similar patches is extracted. This step provides one set K of patches per modality m ∈ M, where M corresponds to the set of the different modalities provided. Nevertheless, at each voxel the quality of the grading estimation is not the same for all the modalities. Therefore, the degree of confidence is estimated with the function α defined as: that reflects the confidence of the grading value gxi for the modality m at the voxel xi. This confidence measure is derived from multi-feature fusion75. Thus, each modality provides a weak classifier at each voxel that is weighted with its degree of confidence α(xi,m). The multimodal grading, denoted gxi, is given by:
In other words, the weight w and Kxi,m are estimated independently for each modality and combined afterwards. Therefore, the proposed combination framework is spatially adaptive and takes advantage of having access to a local degree of confidence αxi,m for each modality m. Basically, when the set of patches found for a modality in the training library is composed of good candidates (i.e., patches very similar to the patch from the subject under study), our confidence αxi,m in the grading estimation for this modality is high. At the end, this modality will have a high weight in the mixing procedure described in (4).
Features estimation
Features were estimated in each hippocampal subfield and over the whole hippocampus as the union of all hippocampal subfields masks. To reduce the inter-individual variability, all volumes are normalized by the total intra-cranial volume76. Afterwards, we aggregate local weak classifiers of the grading map into a single feature for each considered structure (i.e., hippocampal subfields and whole HIPP) by averaging them. Therefore, patch-based grading features are computed by an unweighted vote of the weak classifiers using the segmentation masks (see Fig. 3). Finally, to prevent the bias introduced as the structure alterations due to aging, all the features (i.e., volume, mean of MD and MPBG) are age corrected with a linear regression based on the CN group77.
Implementation
To find the most similar patches in the training library, we use the OPAL method78. OPAL is a fast approximate nearest neighbor patch search technique. This method enables to process each modality in about 4 seconds on a standard computer. The training library is equally composed of 37 images for both CN and AD subjects, leading to |T| = 76. The number of patches extracted from each the training library is K = 160 (i.e., 80 from CN subjects and 80 from AD patients) and the patch size is 5 × 5 × 5 voxels. Furthermore, as done in our PBG DTI study57, we used zero normalized sum of squared differences for T1w to compute the L2 norm (see Equation (2)). On the other hand, d-MRI is a quantitative imaging technique. Therefore, to preserve the quantitative information, a straight sum of squared differences is used for MD in Equation (2),
Validation
To evaluate the efficiency of each considered biomarker to detect AD alterations, CN group is compared to AD patients group. In addition, to discriminate the impairment severity of MCI group, eMCI versus lMCI classification is conducted. The classification step is performed with a linear discriminant analysis (LDA) within a repeated stratified 5-fold cross-validation iterated 200 times. Mean area under the curve (AUC) and mean accuracy (ACC) are computed to compare performance for each biomarker over the 200 iterations.
Results
In this section, the results are presented in three parts. In the first part, we compare the different approach applied within the entire HIPP structure to evaluate the performance of our new MPBG compared to usual biomarkers such as volume and average MD. Afterwards, in the second part, we compare the accuracy of each considered biomarkers within hippocampal subfields to investigate the potential of hippocampal subfields analysis to improve result of AD detection and prediction. Finally, in a last part, we compare the results of our proposed multimodal biomarker with state-of-the-art methods based on d-MRI to show the competitive performance of our approach.
Whole hippocampus
Results of the comparisons over the whole HIPP are represented in Table 2. In this experiment, we compared the results of volume, mean of MD and PBG applied with both modality and MPBG over the whole hippocampus.
First, the hippocampus volume and its average of MD were compared. For CN versus AD classification the volume obtains 86.6% of AUC and the average of MD obtains 80.6%. For eMCI versus lMCI classification the volume and the average of MD obtain 59.4% and 55.6% of AUC, respectively. Experiments demonstrate that the hippocampus the volume obtains better classification results than the average of MD for all comparison, especially for CN versus AD. Second, PBG biomarkers applied with T1w and MD were compared. The results showed that T1w PBG provides better results than MD PBG with 92.6% of AUC for CN versus AD classification. However, for eMCI versus lMCI classification MD grading provides the best results with 69.5% of AUC. MPBG methods combining both modalities reaches the best results for CN versus AD and eMCI versus lMCI with 92.1% and 69.5% of AUC, respectively. Finally, the proposed MPBG biomarker provides results similar to the best modalities for all considered comparisons. Compared to volume, MPBG improves CN versus AD comparison result by 5.5% of AUC and by over 10% of AUC for eMCI versus lMCI comparison. Thus, MBPG biomarker has a good capability to capture modifications caused by AD at different severity stages (see Figure 2).
Hippocampal subfields
Figure 4 shows the distribution of volumes (A), average of MD (B) and MPBG (C) for each hippocampal subfield at each different AD stages. For each comparison a p-value was estimated with a multi-comparison test79. We can note that for all hippocampal subfields, alterations caused by the disease are related to a volume and MPBG decrease with a MD increase. Subiculum subfield presents the most significant differences for several comparisons. More importantly, it is the only subfield providing a p-value inferior than 0.05 for the comparison CN versus eMCI with volume, a p-value inferior than 0.01 to lMCI versus AD with MD and a p-value inferior to 0.001 to eMCI versus lMCI with MPBG, which are the most challenging comparisons. The distribution of MPBG shows a better discrimination between each group for all hippocampal subfields. Indeed, MPBG applied within CA1SP, and CA1SR-L-M provides p-values inferior than 0.01 for eMCI versus lMCI. Moreover, MPBG applied within the subiculum provides p-value inferior than 0.001 for the same comparison. Thus, MPBG enables to perform a detection of AD with each subfield with an advantage for subiculum for the comparison of eMCI versus lMCI.
To estimate the efficiency of the considered biomarkers for AD detection, we also performed a classification experiment. Figures 5 shows the results of two comparisons, CN versus AD (part noted A in the figure) and eMCI versus lMCI (part noted B). First, for AD diagnosis (i.e., CN versus AD classification), the subfield providing the most discriminant volume is the CA1S-R-L-M with an AUC of 86.0%. Moreover, the most discriminant MD biomarker is given by the subiculum with an AUC of 88.1%. For this comparison, MD of subiculum is the only biomarker performing better results than whole hippocampus. The best results obtained by MPBG feature is provided by the CA1SP with an AUC of 92.1% followed by CA1S-R-L-M and subiculum. Second, for eMCI versus lMCI classification, the subiculum provides the best results for each considered feature. Indeed, subiculum obtained an AUC of 66.1% for the volume, 62.4% for the average of MD, and 71.8% for MPBG. Moreover, subiculum provided better results than whole hippocampus for each feature. Thus, the experiments conducted with three different biomarkers showed that the use of hippocampal subfields, especially the subiculum, enables to obtain better results for AD prediction than the whole hippocampal analysis.
Comparison with state-of-the-art methods
To evaluate the performance of the proposed MPBG, we compared it with state-of-the-art multimodal methods using d-MRI. To this end, we used the ACC values published by the authors. Table 3 shows the comparison of our proposed biomarkers within the hippocampal area providing the best results (i.e. the whole HIPP and the subiculum) with the state-of-the-arts methods using similar dataset based on ADNI-2. We compared these biomarkers with a method using features based on tractography80, a method based on a connectivity network of the different brain structures52, and a voxel-based method that analyzes alterations of white matter81. The results of comparison show that MPBG over whole HIPP obtains the best score for AD versus CN with 88.1% of accuracy while the best result is achieved by a voxel-based method with a feature selection81 that obtained 87.0% on similar ADNI2 dataset. For the best of our knowledge, the only work providing eMCI and lMCI comparison52 using d-MRI from similar ADNI2 dataset is based on a connectivity network and obtained 63.4%. These comparisons demonstrate the relevance of MPBG biomarkers for AD detection and prediction. Indeed, our method provides similar results than the best methods with similar dataset for CN versus AD classification and provides the best results for eMCI versus lMCI classification.
Moreover, the proposed MPBG method based on subiculum improves the performance for eMCI versus lMCI classification with an accuracy of 70.8%, that increases by 2% the accuracy based the whole HIPP and over 6% compared to a connectivity network based method.
Discussion
In this work, a multimodal analysis of the hippocampal subfields alterations caused by AD is proposed. First, the structural and micro-structural alterations were captured from two MRI modalities with different methods. Thus, the use of volume, MD, and the proposed MPBG methods were investigated to achieve this analysis. In this section, the efficiency of these different methods applied into the whole hippocampus and each hippocampal subfield are discussed.
Whole hippocampus biomarkers
We first compared the performance of different methods applied to the whole hippocampus (see Table 2). The experiments showed that volume and mean of MD within a structure as the hippocampus does not provides discriminant biomarkers to detect early stages of AD. The MPBG method based on s-MRI and d-MRI obtains best results compared to the volume and the average of MD. Moreover, compared to recent methods proposed for AD detection82 (see Table 3), proposed MPBG demonstrates state-of-the-art performances for AD detection and prediction. These results emphasize the relevance of using more accurate biomarker, such MPBG to study the effectiveness of hippocampal subfields for AD detection and prediction.
Hippocampal subfield biomarkers
The main contribution of this study is the multimodal analysis of hippocampal subfields. Indeed, most of the proposed biomarkers based on hippocampus focused only on the whole structure or study alterations of hippocampal subfields with methods that do not provide sensitive biomarkers to detect early modification caused by AD. The lack of work studying alterations of hippocampal subfields with advanced biomarkers could be explained by the fact that automatic segmentation of the hippocampal subfields is a complex task due to subtle borders dividing each area.
In this work, we compared the efficiency of diffusion MRI and multimodal patch-based biomarkers for AD detection and prediction over the hippocampal subfields. Comparisons based on MD, volume and multimodal patch-based biomarkers showed that the subiculum is the most discriminant structure in the earliest stage of AD providing the best results for AD prediction (see Figure 4 and 5). However, whole hippocampus structure, followed by CA1SR-L-M, obtains best results for AD detection.
These results are in accordance with literature studies based on animal model and in vivo imaging combining volume and MD demonstrating that subiculum is the earliest hippocampal region affected by AD42, 43. Moreover, postmortem studies showed that the hippocampal degeneration in early stages of AD is not uniform. After the apparition of alterations in the EC, the pathology spreads to the subiculum, CA1, CA2-3 and finally the CA4 and DG subfields36, 37, 42, 83. It is interesting to note that the results of our experiments using volume-based biomarkers are also coherent with the previous in-vivo imaging studies that analyzed the atrophy of each hippocampal subfield at advanced stage of AD. These studies showed that CA1 is the subfield impacted by the strongest atrophy38, 39, 84, 85. Furthermore, studies using ultra-high field at 7T enabling CA1 layers discrimination showed that CA1SR-L-M is the subfields showing the greater atrophy at advanced stages of AD40, 41.
Conclusion
In this paper, we analyzed hippocampal subfield alterations with a multimodal framework based on structural and diffusion MRI. In addition, to study tenuous modifications occurring into each hippocampal subfield, we developed a new multimodal patch-based framework using T1w and DTI. Our novel MPBG method were compared to the volume and the average of MD over the whole hippocampus. This comparison demonstrated that MPBG method improves performances for AD detection and prediction. In addition, a comparison with state-of-the-art diffusion-based methods showed the competitive performance of MPBG biomarkers. Finally, an analysis of the hippocampal subfields with the volume, the average of MD and MBPG methods was conducted. Although CA1 is the subfields having the greater atrophy in the late stage of AD, the experiments demonstrated that whole hippocampus provides best biomarker for AD detection while subiculum provides best biomarker for AD prediction.
Acknowledgements
This study has been carried out with financial support from the French State, managed by the French National Research Agency (ANR) in the frame of the Investments for the future Program IdEx Bordeaux (HL-MRI ANR-10-IDEX-03-02), Cluster of excellence CPU and TRAIL (HR-DTI ANR-10-LABX-57) and the CNRS multidisciplinary project “Défi imag’In”.
Data collection and sharing for this project was funded by the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904) and by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: AbbVie, Alzheimer’s Biogen; Bristol-Myes Squibb Company; CereSpir, Inc.; Cogstate; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffman-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Pharmaceutical Research & Development LLC.; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organization is the Northern California Institute of Research and Education, and the study is coordinated by the Alzheimer’s Therapeutic Research Institute at the University of Southern California. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California.
Footnotes
↵* Data used in preparation of this article were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in analysis or writing of this report. A complete listing of ADNI investigators can be found at: http://adni.loni.usc.edu/wp-content/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf.
References
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.
- 6.↵
- 7.↵
- 8.
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.
- 22.
- 23.
- 24.
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵
- 84.↵
- 85.↵