Role of Combining Artificial Intelligence (AI) and Neuroimaging in Detecting Dementia with Lewy Bodies (DLB): A Scoping Review
Mahmoud Tarek Hefnawy1, Ibrahim Serag2, Ahmed Y. Azzam3, Mohamed Ahmed Ali4, Amr K. Hassan5, Rehab Adel Diab6, Mohamed Diab7, Ahmed Negida8
1Faculty of Medicine Zagazig University, Egypt, 2Faculty of Medicine, Mansoura University, 3Faculty of Medicine, October 6 University, Giza, Egypt, 4Qena Faculty of Medicine, South Valley University, Qena, Egypt, 5University of California, Irvine, CA, USA, 6Faculty of medicine, Al-Azhar university, Cairo, Egypt, 7Faculty of Medicine Alexandria University, Alexandria, Egypt, 8Virginia Commonwealth University
Objective:
This scoping review systematically maps the literature on AI models in neuroimaging for detecting and differentiating DLB from other neurodegenerative conditions. It evaluates diverse AI approaches and imaging modalities, identifies trends and methodological issues, and provides recommendations to address limitations and improve diagnostic accuracy.
Background:
Dementia with Lewy Bodies (DLB) is a significant neurodegenerative disorder characterized by the presence of Lewy bodies, leading to cognitive and motor impairments. Early and accurate diagnosis of DLB is crucial but challenging due to its overlap with other dementias, such as Alzheimer’s Disease (AD). Recent advancements in artificial intelligence and neuroimaging offer promising avenues for enhancing diagnostic precision.
Design/Methods:
Following PRISMA-ScR guidelines, a comprehensive search was conducted across Scopus, Web of Science, and PubMed. Studies were included if they involved human subjects diagnosed with DLB and utilized AI models applied to neuroimaging data. Data extraction and analysis focused on study characteristics, AI model specifics, neuroimaging techniques, and diagnostic performance metrics.
Results:
Out of 124 identified studies, 22 met the inclusion criteria. For DLB detection, models like Convolutional Neural Networks achieved sensitivities of 86% and specificities of 100%. Differentiation from other dementias has improved, with models showing up to 94% sensitivity and 95.4% specificity for distinguishing DLB from AD and Parkinsonian syndromes. However, limitations include reliance on specific datasets, small sample sizes, and the use of synthetic data, which impact generalizability and diagnostic accuracy. Future research should focus on expanding datasets to include diverse populations, integrating additional clinical data, and validating AI models with independent datasets. Addressing these limitations will be crucial for enhancing diagnostic precision, integrating AI into clinical workflows, and advancing personalized treatment strategies for DLB.
Conclusions:
The integration of AI and neuroimaging presents significant potential for improving DLB diagnosis and differentiation from other dementias. However, challenges such as data limitations, sample size , and model generalizability remain.
Disclaimer: Abstracts were not reviewed by Neurology® and do not reflect the views of Neurology® editors or staff.