Machine learning of neuroimaging for assisted diagnosis of cognitive impairment and dementia: A systematic review

Statistical methods for dementia risk prediction and recommendations for future work: A systematic review

Introduction

Numerous dementia risk prediction models have been developed in the past decade. However, methodological limitations of the analytical tools used may hamper their ability to generate reliable dementia risk scores. We aim to review the used methodologies.

Methods

We systematically reviewed the literature from March 2014 to September 2018 for publications presenting a dementia risk prediction model. We critically discuss the analytical techniques used in the literature.

Results

In total 137 publications were included in the qualitative synthesis. Three techniques were identified as the most commonly used methodologies: machine learning, logistic regression, and Cox regression.

Discussion

We identified three major methodological weaknesses: (1) over-reliance on one data source, (2) poor verification of statistical assumptions of Cox and logistic regression, and (3) lack of validation. The use of larger and more diverse data sets is recommended. Assumptions should be tested thoroughly, and actions should be taken if deviations are detected.

Quantifying Neurodegenerative Progression With DeepSymNet, an End-to-End Data-Driven Approach

Alzheimer's disease (AD) is the most common neurodegenerative disorder worldwide and is one of the leading sources of morbidity and mortality in the aging population. There is a long preclinical period followed by mild cognitive impairment (MCI). Clinical diagnosis and the rate of decline is variable. Progression monitoring remains a challenge in AD, and it is imperative to create better tools to quantify this progression. Brain magnetic resonance imaging (MRI) is commonly used for patient assessment. However, current approaches for analysis require strong a priori assumptions about regions of interest used and complex preprocessing pipelines including computationally expensive non-linear registrations and iterative surface deformations. These preprocessing steps are composed of many stacked processing layers. Any error or bias in an upstream layer will be propagated throughout the pipeline. Failures or biases in the non-linear subject registration and the subjective choice of atlases of specific regions are common in medical neuroimaging analysis and may hinder the translation of many approaches to the clinical practice. Here we propose a data-driven method based on an extension of a deep learning architecture, DeepSymNet, that identifies longitudinal changes without relying on prior brain regions of interest, an atlas, or non-linear registration steps. Our approach is trained end-to-end and learns how a patient's brain structure dynamically changes between two-time points directly from the raw voxels. We compare our approach with Freesurfer longitudinal pipelines and voxel-based methods using the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. Our model can identify AD progression with comparable results to existing Freesurfer longitudinal pipelines without the need of predefined regions of interest, non-rigid registration algorithms, or iterative surface deformation at a fraction of the processing time. When compared to other voxel-based methods which share some of the same benefits, our model showed a statistically significant performance improvement. Additionally, we show that our model can differentiate between healthy subjects and patients with MCI. The model's decision was investigated using the epsilon layer-wise propagation algorithm. We found that the predictions were driven by the pallidum, putamen, and the superior temporal gyrus. Our novel longitudinal based, deep learning approach has the potential to diagnose patients earlier and enable new computational tools to monitor neurodegeneration in clinical practice.

Cortical graph neural network for AD and MCI diagnosis and transfer learning across populations

Combining machine learning with neuroimaging data has a great potential for early diagnosis of mild cognitive impairment (MCI) and Alzheimer's disease (AD). However, it remains unclear how well the classifiers built on one population can predict MCI/AD diagnosis of other populations. This study aimed to employ a spectral graph convolutional neural network (graph-CNN), that incorporated cortical thickness and geometry, to identify MCI and AD based on 3089 T₁-weighted MRI data of the ADNI-2 cohort, and to evaluate its feasibility to predict AD in the ADNI-1 cohort (n = 3602) and an Asian cohort (n = 347). For the ADNI-2 cohort, the graph-CNN showed classification accuracy of controls (CN) vs. AD at 85.8% and early MCI (EMCI) vs. AD at 79.2%, followed by CN vs. late MCI (LMCI) (69.3%), LMCI vs. AD (65.2%), EMCI vs. LMCI (60.9%), and CN vs. EMCI (51.8%). We demonstrated the robustness of the graph-CNN among the existing deep learning approaches, such as Euclidean-domain-based multilayer network and 1D CNN on cortical thickness, and 2D and 3D CNNs on T₁-weighted MR images of the ADNI-2 cohort. The graph-CNN also achieved the prediction on the conversion of EMCI to AD at 75% and that of LMCI to AD at 92%. The find-tuned graph-CNN further provided a promising CN vs. AD classification accuracy of 89.4% on the ADNI-1 cohort and >90% on the Asian cohort. Our study demonstrated the feasibility to transfer AD/MCI classifiers learned from one population to the other. Notably, incorporating cortical geometry in CNN has the potential to improve classification performance.

•

Graph CNN incorporates cortical thickness and geometry.

•

Graph CNN is more robust than other image-based CNN.

•

Graph CNN well identified MCI and AD based on 3089 ADNI-2 MRI data.

•

Graph CNN built on ADNI-2 was transferable to the ADNI-1 and Asian cohorts.

Codex (Cognitive Disorders Examination) Decision Tree Modified for the Detection of Dementia and MCI

Many cognitive screening instruments are available to assess patients with cognitive symptoms in whom a diagnosis of dementia or mild cognitive impairment is being considered. Most are quantitative scales with specified cut-off values. In contrast, the cognitive disorders examination or Codex is a two-step decision tree which incorporates components from the Mini-Mental State Examination (MMSE) (three word recall, spatial orientation) along with a simplified clock drawing test to produce categorical outcomes defining the probability of dementia diagnosis and, by implication, directing clinician response (reassurance, monitoring, further investigation, immediate treatment). Codex has been shown to have high sensitivity and specificity for dementia diagnosis but is less sensitive for the diagnosis of mild cognitive impairment (MCI). We examined minor modifications to the Codex decision tree to try to improve its sensitivity for the diagnosis of MCI, based on data extracted from studies of two other cognitive screening instruments, the Montreal Cognitive Assessment and Free-Cog, which are more stringent than MMSE in their tests of delayed recall. Neither modification proved of diagnostic value for mild cognitive impairment. Possible explanations for this failure are considered.

Diagnosis and prognosis of Alzheimer's disease using brain morphometry and white matter connectomes

Accurate, reliable prediction of risk for Alzheimer's disease (AD) is essential for early, disease-modifying therapeutics. Multimodal MRI, such as structural and diffusion MRI, is likely to contain complementary information of neurodegenerative processes in AD. Here we tested the utility of the multimodal MRI (T1-weighted structure and diffusion MRI), combined with high-throughput brain phenotyping—morphometry and structural connectomics—and machine learning, as a diagnostic tool for AD. We used, firstly, a clinical cohort at a dementia clinic (National Health Insurance Service-Ilsan Hospital [NHIS-IH]; N = 211; 110 AD, 64 mild cognitive impairment [MCI], and 37 cognitively normal with subjective memory complaints [SMC]) to test the diagnostic models; and, secondly, Alzheimer's Disease Neuroimaging Initiative (ADNI)-2 to test the generalizability. Our machine learning models trained on the morphometric and connectome estimates (number of features = 34,646) showed optimal classification accuracy (AD/SMC: 97% accuracy, MCI/SMC: 83% accuracy; AD/MCI: 97% accuracy) in NHIS-IH cohort, outperforming a benchmark model (FLAIR-based white matter hyperintensity volumes). In ADNI-2 data, the combined connectome and morphometry model showed similar or superior accuracies (AD/HC: 96%; MCI/HC: 70%; AD/MCI: 75% accuracy) compared with the CSF biomarker model (t-tau, p-tau, and Amyloid β, and ratios). In predicting MCI to AD progression in a smaller cohort of ADNI-2 (n = 60), the morphometry model showed similar performance with 69% accuracy compared with CSF biomarker model with 70% accuracy. Our comparisons of the classifiers trained on structural MRI, diffusion MRI, FLAIR, and CSF biomarkers showed the promising utility of the white matter structural connectomes in classifying AD and MCI in addition to the widely used structural MRI-based morphometry, when combined with machine learning.

•

This study tests the utility of multimodal brain MRI and machine learning in diagnosis and prognosis of AD.

•

Models trained on connectomes and morphometry beast classify elders with AD or MCI from cognitively normal elders.

•

Models trained on morphometry best predict MCI to AD progression.

Identification of the presence of ischaemic stroke lesions by means of texture analysis on brain magnetic resonance images

•

Radiomics in conventionally segmented tissues can identify MRI scans that had a stroke.

•

Patient’s advanced age can negatively influence classification results.

•

Feature selection and stroke subtype influence but do not determine accuracy.

•

Stroke subtype cannot be ascertained from texture analysis in brain tissues.

Background

The differential quantification of brain atrophy, white matter hyperintensities (WMH) and stroke lesions is important in studies of stroke and dementia. However, the presence of stroke lesions is usually overlooked by automatic neuroimage processing methods and the-state-of-the-art deep learning schemes, which lack sufficient annotated data. We explore the use of radiomics in identifying whether a brain magnetic resonance imaging (MRI) scan belongs to an individual that had a stroke or not.

Materials and methods

We used 1800 3D sets of MRI data from three prospective studies: one of stroke mechanisms and two of cognitive ageing, evaluated 114 textural features in WMH, cerebrospinal fluid, deep grey and normal-appearing white matter, and attempted to classify the scans using a random forest and support vector machine classifiers with and without feature selection. We evaluated the discriminatory power of each feature independently in each population and corrected the result against Type 1 errors. We also evaluated the influence of clinical parameters in the classification results.

Results

Subtypes of ischaemic strokes (i.e. lacunar vs. cortical) cannot be discerned using radiomics, but the presence of a stroke-type lesion can be ascertained with accuracies ranging from 0.7 < AUC < 0.83. Feature selection, tissue type, stroke subtype and MRI sequence did not seem to determine the classification results. From all clinical variables evaluated, age correlated with the proportion of images classified correctly using either different or the same descriptors (Pearson r = 0.31 and 0.39 respectively, p < 0.001).

Conclusions

Texture features in conventionally automatically segmented tissues may help in the identification of the presence of previous stroke lesions on an MRI scan, and should be taken into account in transfer learning strategies of the-state-of-the-art deep learning schemes.

Prediction of future cognitive impairment among the community elderly: A machine-learning based approach

The early detection of cognitive impairment is a key issue among the elderly. Although neuroimaging, genetic, and cerebrospinal measurements show promising results, high costs and invasiveness hinder their widespread use. Predicting cognitive impairment using easy-to-collect variables by non-invasive methods for community-dwelling elderly is useful prior to conducting such a comprehensive evaluation. This study aimed to develop a machine learning-based predictive model for future cognitive impairment. A total of 3424 community elderly without cognitive impairment were included from the nationwide dataset. The gradient boosting machine (GBM) was exploited to predict cognitive impairment after 2 years. The GBM performance was good (sensitivity = 0.967; specificity = 0.825; and AUC = 0.921). This study demonstrated that a machine learning-based predictive model might be used to screen future cognitive impairment using variables, which are commonly collected in community health care institutions. With efforts of enhancing the predictive performance, such a machine learning-based approach can further contribute to the improvement of the cognitive function in community elderly.

Automated detection of age-related macular degeneration in color fundus photography: a systematic review

The rising prevalence of age-related eye diseases, particularly age-related macular degeneration, places an ever-increasing burden on health care providers. As new treatments emerge, it is necessary to develop methods for reliably assessing patients' disease status and stratifying risk of progression. The presence of drusen in the retina represents a key early feature in which size, number, and morphology are thought to correlate significantly with the risk of progression to sight-threatening age-related macular degeneration. Manual labeling of drusen on color fundus photographs by a human is labor intensive and is where automatic computerized detection would appreciably aid patient care. We review and evaluate current artificial intelligence methods and developments for the automated detection of drusen in the context of age-related macular degeneration.