Deep learning-based classification of healthy aging controls, mild cognitive impairment and Alzheimer’s disease using fusion of MRI-PET imaging

Rallabandi, V. P. Subramanyam; Seetharaman, K.

doi:10.1016/j.bspc.2022.104312

Cited by 28 publications

(8 citation statements)

References 77 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…RF and multiple linear regression models both revealed indoxyl sulfate, choline, 5-hydroxyindole acetic acid, IPA and kynurenic acid as key early indicators of cognitive decline, with RF presenting an AUC predictive performance of 0.74, strongly supporting a significant link between metabolic perturbations associated with the gut microbiome and preclinical AD progression. Previous studies have predominantly concentrated on binary classification approaches, primarily utilising MRI and PET imaging modalities, to investigate AD progression [55][56][57] . However, in clinical practice, multiclass classification of blood samples of patients with SCI, MCI and healthy controls could provide a useful approach.…”

Section: Discussionmentioning

confidence: 99%

Circulatory dietary and gut-derived metabolites predict preclinical Alzheimer’s disease

Connell,

Sami,

Khondoker

et al. 2024

Preprint

View full text Add to dashboard Cite

A key component of disease prevention is the identification of at-risk individuals. Microbial dysbiosis and microbe-derived metabolites (MDM) can influence the central nervous system, but their role in disease progression and as prognostic indicators is unknown. To identify preclinical factors associated with Alzheimer disease (AD), we compared gut microbiome and metabolome profiles of cognitively healthy subjects, subjective cognitive impairment (SCI) participants and mild cognitive impairment (MCI) participants (n=50 per group, matched for age, BMI and sex), targeting metabolites previously associated with cognitive health (TMAO, bile acids, tryptophan, p-cresol and their derivatives). 16S rRNA bacterial microbiome sequencing and targeted LC-MS/MS were employed for faecal microbiome speciation and serum MDM quantification. Microbiome beta diversity differed between healthy controls and SCI participants. Multiple linear regression modelling highlighted five serum metabolites (indoxyl sulfate, choline, 5-hydroxyindole acetic acid, indole-3-propionic acid (IPA) and kynurenic acid) significantly altered in preclinical AD. Neuroprotective metabolites, including choline, 5-hydroxyindole acetic acid and IPA, exhibited lower concentrations in SCI and MCI in comparison to controls, while the cytotoxic metabolite indoxyl sulfate had higher levels. A Random Forest algorithm with multiclass classification confirmed and extended our results, identifying six metabolites (indoxyl sulfate, choline, 5-hydroxyindole acetic acid, IPA, kynurenic acid, kynurenine) as predictors of early cognitive decline, with an area under the curve of 0.74. In summary, a combined statistical and machine learning approach identified MDM as a novel composite risk factor for the early identification of future dementia risk.

show abstract

Section: Discussionmentioning

confidence: 99%

Circulatory dietary and gut-derived metabolites predict preclinical Alzheimer’s disease

Connell,

Sami,

Khondoker

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…This is then analyzed by a fully connected neural network for classification and prediction. Rallambandi and Seetharaman 27 proposed a deep learning-based Inception-ResNet50 wrapper model for distinguishing Mild Cognitive Impairment (MCI) and AD dementia patients from healthy controls, leveraging both structural MRI for spatial detail and functional PET for temporal resolution, underscoring the value of combining multi-modal imaging modalities.…”

Section: Related Workmentioning

confidence: 99%

Automated detection of Alzheimer’s disease: a multi-modal approach with 3D MRI and amyloid PET

Castellano,

Esposito,

Lella

et al. 2024

Sci Rep

View full text Add to dashboard Cite

Recent advances in deep learning and imaging technologies have revolutionized automated medical image analysis, especially in diagnosing Alzheimer’s disease through neuroimaging. Despite the availability of various imaging modalities for the same patient, the development of multi-modal models leveraging these modalities remains underexplored. This paper addresses this gap by proposing and evaluating classification models using 2D and 3D MRI images and amyloid PET scans in uni-modal and multi-modal frameworks. Our findings demonstrate that models using volumetric data learn more effective representations than those using only 2D images. Furthermore, integrating multiple modalities enhances model performance over single-modality approaches significantly. We achieved state-of-the-art performance on the OASIS-3 cohort. Additionally, explainability analyses with Grad-CAM indicate that our model focuses on crucial AD-related regions for its predictions, underscoring its potential to aid in understanding the disease’s causes.

show abstract

“…When multimodal imaging data was used in the included studies, the workflow of the model usually consisted of feature extraction, feature selection, feature fusion and using multi-source discriminative features for classification [ 53 ]. Convolutional neural network (CNN) was the most widely used technique for feature extraction [ 17 , 42 , 43 , 53 , 57 , 68 , 71 ]. After extraction of biomedical image features, feature selection was used to explore deep common features among different image features and gain information sharing among multiple modal data [ 53 ].…”

Section: Narrative Synthesis Of Relevant Findings From the Evidencementioning

confidence: 99%

“…In terms of the AI methods used, a study published in 2021 reported an accuracy of 97.95% for the classification of controls and PD patients using Support vector machines (SVMs) [ 28 ]. Rallabandi et al [ 68 ] prediction and achieved an accuracy of 98.81% on prediction of MCI-to-AD conversion in 5 years.…”

Section: Narrative Synthesis Of Relevant Findings From the Evidencementioning

confidence: 99%

Multimodal learning of clinically accessible tests to aid diagnosis of neurodegenerative disorders: a scoping review

Huang

Bai

et al. 2023

Health Inf Sci Syst

View full text Add to dashboard Cite

With ageing populations around the world, there is a rapid rise in the number of people with Alzheimer’s disease (AD) and Parkinson’s disease (PD), the two most common types of neurodegenerative disorders. There is an urgent need to find new ways of aiding early diagnosis of these conditions. Multimodal learning of clinically accessible data is a relatively new approach that holds great potential to support early precise diagnosis. This scoping review follows the PRSIMA guidelines and we analysed 46 papers, comprising 11,750 participants, 3569 with AD, 978 with PD, and 2482 healthy controls; the recency of this topic was highlighted by nearly all papers being published in the last 5 years. It highlights the effectiveness of combining different types of data, such as brain scans, cognitive scores, speech and language, gait, hand and eye movements, and genetic assessments for the early detection of AD and PD. The review also outlines the AI methods and the model used in each study, which includes feature extraction, feature selection, feature fusion, and using multi-source discriminative features for classification. The review identifies knowledge gaps around the need to validate findings and address limitations such as small sample sizes. Applying multimodal learning of clinically accessible tests holds strong potential to aid the development of low-cost, reliable, and non-invasive methods for early detection of AD and PD.

show abstract

Deep learning-based classification of healthy aging controls, mild cognitive impairment and Alzheimer’s disease using fusion of MRI-PET imaging

Cited by 28 publications

References 77 publications

Circulatory dietary and gut-derived metabolites predict preclinical Alzheimer’s disease

Circulatory dietary and gut-derived metabolites predict preclinical Alzheimer’s disease

Automated detection of Alzheimer’s disease: a multi-modal approach with 3D MRI and amyloid PET

Multimodal learning of clinically accessible tests to aid diagnosis of neurodegenerative disorders: a scoping review

Contact Info

Product

Resources

About