Ricardo Guerrero

Standardized evaluation of algorithms for computer-aided diagnosis of dementia based on structural MRI: The CADDementia challenge

Algorithms for computer-aided diagnosis of dementia based on structural MRI have demonstrated high performance in the literature, but are difficult to compare as different data sets and methodology were used for evaluation. In addition, it is unclear how the algorithms would perform on previously unseen data, and thus, how they would perform in clinical practice when there is no real opportunity to adapt the algorithm to the data at hand. To address these comparability, generalizability and clinical applicability issues, we organized a grand challenge that aimed to objectively compare algorithms based on a clinically representative multi-center data set. Using clinical practice as the starting point, the goal was to reproduce the clinical diagnosis. Therefore, we evaluated algorithms for multi-class classification of three diagnostic groups: patients with probable Alzheimers disease, patients with mild cognitive impairment and healthy controls. The diagnosis based on clinical criteria was used as reference standard, as it was the best available reference despite its known limitations. For evaluation, a previously unseen test set was used consisting of 354 T1-weighted MRI scans with the diagnoses blinded. Fifteen research teams participated with a total of 29 algorithms. The algorithms were trained on a small training set (n=30) and optionally on data from other sources (e.g., the Alzheimers Disease Neuroimaging Initiative, the Australian Imaging Biomarkers and Lifestyle flagship study of aging). The best performing algorithm yielded an accuracy of 63.0% and an area under the receiver-operating-characteristic curve (AUC) of 78.8%. In general, the best performances were achieved using feature extraction based on voxel-based morphometry or a combination of features that included volume, cortical thickness, shape and intensity. The challenge is open for new submissions via the web-based framework: http://caddementia.grand-challenge.org.

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 Alzheimers 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.

Multi-input Cardiac Image Super-Resolution Using Convolutional Neural Networks

3D cardiac MR imaging enables accurate analysis of cardiac morphology and physiology. However, due to the requirements for long acquisition and breath-hold, the clinical routine is still dominated by multi-slice 2D imaging, which hamper the visualization of anatomy and quantitative measurements as relatively thick slices are acquired. As a solution, we propose a novel image super-resolution (SR) approach that is based on a residual convolutional neural network (CNN) model. It reconstructs high resolution 3D volumes from 2D image stacks for more accurate image analysis. The proposed model allows the use of multiple input data acquired from different viewing planes for improved performance. Experimental results on 1233 cardiac short and long-axis MR image stacks show that the CNN model outperforms state-of-the-art SR methods in terms of image quality while being computationally efficient. Also, we show that image segmentation and motion tracking benefits more from SR-CNN when it is used as an initial upscaling method than conventional interpolation methods for the subsequent analysis.

Differential diagnosis of neurodegenerative diseases using structural MRI data.

Different neurodegenerative diseases can cause memory disorders and other cognitive impairments. The early detection and the stratification of patients according to the underlying disease are essential for an efficient approach to this healthcare challenge. This emphasizes the importance of differential diagnostics. Most studies compare patients and controls, or Alzheimers disease with one other type of dementia. Such a bilateral comparison does not resemble clinical practice, where a clinician is faced with a number of different possible types of dementia. Here we studied which features in structural magnetic resonance imaging (MRI) scans could best distinguish four types of dementia, Alzheimers disease, frontotemporal dementia, vascular dementia, and dementia with Lewy bodies, and control subjects. We extracted an extensive set of features quantifying volumetric and morphometric characteristics from T1 images, and vascular characteristics from FLAIR images. Classification was performed using a multi-class classifier based on Disease State Index methodology. The classifier provided continuous probability indices for each disease to support clinical decision making. A dataset of 504 individuals was used for evaluation. The cross-validated classification accuracy was 70.6% and balanced accuracy was 69.1% for the five disease groups using only automatically determined MRI features. Vascular dementia patients could be detected with high sensitivity (96%) using features from FLAIR images. Controls (sensitivity 82%) and Alzheimers disease patients (sensitivity 74%) could be accurately classified using T1-based features, whereas the most difficult group was the dementia with Lewy bodies (sensitivity 32%). These results were notable better than the classification accuracies obtained with visual MRI ratings (accuracy 44.6%, balanced accuracy 51.6%). Different quantification methods provided complementary information, and consequently, the best results were obtained by utilizing several quantification methods. The results prove that automatic quantification methods and computerized decision support methods are feasible for clinical practice and provide comprehensive information that may help clinicians in the diagnosis making.

A Novel Grading Biomarker for the Prediction of Conversion From Mild Cognitive Impairment to Alzheimer's Disease

Objective: Identifying mild cognitive impairment (MCI) subjects who will progress to Alzheimers disease (AD) is not only crucial in clinical practice, but also has a significant potential to enrich clinical trials. The purpose of this study is to develop an effective biomarker for an accurate prediction of MCI-to-AD conversion from magnetic resonance images. Methods: We propose a novel grading biomarker for the prediction of MCI-to-AD conversion. First, we comprehensively study the effects of several important factors on the performance in the prediction task including registration accuracy, age correction, feature selection, and the selection of training data. Based on the studies of these factors, a grading biomarker is then calculated for each MCI subject using sparse representation techniques. Finally, the grading biomarker is combined with age and cognitive measures to provide a more accurate prediction of MCI-to-AD conversion. Results: Using the Alzheimers Disease Neuroimaging Initiative (ADNI) dataset, the proposed global grading biomarker achieved an area under the receiver operating characteristic curve (AUC) in the range of 79-81% for the prediction of MCI-to-AD conversion within three years in tenfold cross validations. The classification AUC further increases to 84-92% when age and cognitive measures are combined with the proposed grading biomarker. Conclusion: The obtained accuracy of the proposed biomarker benefits from the contributions of different factors: a tradeoff registration level to align images to the template space, the removal of the normal aging effect, selection of discriminative voxels, the calculation of the grading biomarker using AD and normal control groups, and the integration of sparse representation technique and the combination of cognitive measures. Significance: The evaluation on the ADNI dataset shows the efficacy of the proposed biomarker and demonstrates a significant contribution in accurate prediction of MCI-to-AD conversion.

Manifold population modeling as a neuro-imaging biomarker: application to ADNI and ADNI-GO.

We propose a framework for feature extraction from learned low-dimensional subspaces that represent inter-subject variability. The manifold subspace is built from data-driven regions of interest (ROI). The regions are learned via sparse regression using the mini-mental state examination (MMSE) score as an independent variable which correlates better with the actual disease stage than a discrete class label. The sparse regression is used to perform variable selection along with a re-sampling scheme to reduce sampling bias. We then use the learned manifold coordinates to perform visualization and classification of the subjects. Results of the proposed approach are shown using the ADNI and ADNI-GO datasets. Three types of classification techniques, including a new MRI Disease-State-Score (MRI-DSS) classifier, are tested in conjunction with two learning strategies. In the first case Alzheimers Disease (AD) and progressive mild cognitive impairment (pMCI) subjects were grouped together, while cognitive normal (CN) and stable mild cognitive impaired (sMCI) subjects were also grouped together. In the second approach, the classifiers are learned using the original class labels (with no grouping). We show results that are comparable to other state-of-the-art methods. A classification rate of 71%, of arguably the most clinically relevant subjects, sMCI and pMCI, is shown. Additionally, we present classification accuracies between CN and early MCI (eMCI) subjects, from the ADNI-GO dataset, of 65%. To our knowledge this is the first time classification accuracies for eMCI patients have been reported.

Manifold Alignment and Transfer Learning for Classification of Alzheimer’s Disease

Magnetic resonance (MR) images acquired at different field strengths have different intensity appearance and thus cannot be easily combined into a single manifold space. A framework to learn a joint low-dimensional representation of brain MR images, acquired either at 1.5 or 3 Tesla, is proposed. In this manifold subspace, knowledge can be shared and transfered between the two distinct but related datasets. The joint manifold subspace is built using an adaptation of Laplacian eigenmaps (LE) from a data-driven region of interest (ROI). The ROI is learned using sparse regression to perform simultaneous variable selection at multiple levels of alignment to the MNI152 template. Additionally, a stability selection re-sampling scheme is used to reduce sampling bias while learning the ROI. Knowledge about the intrinsic embedding coordinates of different instances, common to both feature spaces, is used to constrain their alignment in the joint manifold. Alzheimer’s Disease (AD) classification results obtained with the proposed approach are presented using data from more than 1500 subjects from ADNI-1, ADNI-GO and ADNI-2 datasets. Results calculated using the learned joint manifold in general outperform those obtained in each independent manifold. Accuracies calculated on ADNI-1 are comparable to other state-of-the-art approaches. To our knowledge, classification accuracies have not been reported before on the complete ADNI (-1, -GO and -2) cohort combined.

Laplacian eigenmaps manifold learning for landmark localization in brain MR images

The identification of anatomical landmarks in medical images is an important task in registration and morphometry. Manual labeling is time consuming and prone to observer errors. We propose a manifold learning procedure, based on Laplacian Eigenmaps, that learns an embedding from patches drawn from multiple brain MR images. The position of the patches in the manifold can be used to predict the location of the landmarks via regression. New images are embedded in the manifold and the resulting coordinates are used to predict the landmark position in the new image. The output of multiple regressors is fused in a weighted fashion to boost the accuracy and robustness. We demonstrate this framework in 3D brain MR images from the ADNI database. We show an accuracy of -0.5mm, an increase of at least two fold when compared to traditional approaches such as registration or sliding windows.

Stratified Decision Forests for Accurate Anatomical Landmark Localization in Cardiac Images

Accurate localization of anatomical landmarks is an important step in medical imaging, as it provides useful prior information for subsequent image analysis and acquisition methods. It is particularly useful for initialization of automatic image analysis tools (e.g. segmentation and registration) and detection of scan planes for automated image acquisition. Landmark localization has been commonly performed using learning based approaches, such as classifier and/or regressor models. However, trained models may not generalize well in heterogeneous datasets when the images contain large differences due to size, pose and shape variations of organs. To learn more data-adaptive and patient specific models, we propose a novel stratification based training model, and demonstrate its use in a decision forest. The proposed approach does not require any additional training information compared to the standard model training procedure and can be easily integrated into any decision tree framework. The proposed method is evaluated on 1080 3D high-resolution and 90 multi-stack 2D cardiac cine MR images. The experiments show that the proposed method achieves state-of-the-art landmark localization accuracy and outperforms standard regression and classification based approaches. Additionally, the proposed method is used in a multi-atlas segmentation to create a fully automatic segmentation pipeline, and the results show that it achieves state-of-the-art segmentation accuracy.Accurate localization of anatomical landmarks is an important step in medical imaging, as it provides useful prior information for subsequent image analysis and acquisition methods. It is particularly useful for initialization of automatic image analysis tools (e.g. segmentation and registration) and detection of scan planes for automated image acquisition. Landmark localization has been commonly performed using learning based approaches, such as classifier and/or regressor models. However, trained models may not generalize well in heterogeneous datasets when the images contain large differences due to size, pose and shape variations of organs. To learn more data-adaptive and patient specific models, we propose a novel stratification based training model, and demonstrate its use in a decision forest. The proposed approach does not require any additional training information compared to the standard model training procedure and can be easily integrated into any decision tree framework. The proposed method is evaluated on 1080 3D high-resolution and 90 multi-stack 2D cardiac cine MR images. The experiments show that the proposed method achieves state-of-the-art landmark localization accuracy and outperforms standard regression and classification based approaches. Additionally, the proposed method is used in a multi-atlas segmentation to create a fully automatic segmentation pipeline, and the results show that it achieves state-of-the-art segmentation accuracy.

Learning Biomarker Models for Progression Estimation of Alzheimer’s Disease

Being able to estimate a patient’s progress in the course of Alzheimer’s disease and predicting future progression based on a number of observed biomarker values is of great interest for patients, clinicians and researchers alike. In this work, an approach for disease progress estimation is presented. Based on a set of subjects that convert to a more severe disease stage during the study, models that describe typical trajectories of biomarker values in the course of disease are learned using quantile regression. A novel probabilistic method is then derived to estimate the current disease progress as well as the rate of progression of an individual by fitting acquired biomarkers to the models. A particular strength of the method is its ability to naturally handle missing data. This means, it is applicable even if individual biomarker measurements are missing for a subject without requiring a retraining of the model. The functionality of the presented method is demonstrated using synthetic and—employing cognitive scores and image-based biomarkers—real data from the ADNI study. Further, three possible applications for progress estimation are demonstrated to underline the versatility of the approach: classification, construction of a spatio-temporal disease progression atlas and prediction of future disease progression.