Diagnosis of cerebral microbleed via VGG and extreme learning machine trained by Gaussian map bat algorithm

Area of Science:

• Neurology
• Medical Imaging
• Artificial Intelligence

Background:

• Cerebral microbleeds (CMBs) are a significant public health concern linked to dementia.
• Detecting CMBs via brain magnetic resonance imaging (MRI) is crucial but manual inspection is time-consuming and lacks reproducibility.
• CMBs appear as small, scattered dots on MRI scans, complicating manual analysis.

Purpose of the Study:

• To develop and evaluate a novel, automated method for diagnosing cerebral microbleeds (CMBs) using deep learning and optimization algorithms.
• To improve the efficiency and reproducibility of CMB detection in brain MRI scans.
• To compare the performance of the proposed method against existing state-of-the-art approaches.

Main Methods:

• A dataset was generated from brain MRIs using sliding window processing.
• A pre-trained VGG network was utilized for extracting image features.
• An Extreme Learning Machine (ELM) was trained using the Gaussian-map bat algorithm (GBA) for CMB identification.

Main Results:

• The proposed VGG-ELM-GBA method demonstrated superior generalization performance.
• The automated approach significantly outperformed several state-of-the-art methods in CMB diagnosis.
• The study successfully automated the detection of CMBs and non-CMBs from brain MRI data.

Conclusions:

• The VGG-ELM-GBA method presents a promising automated solution for cerebral microbleed diagnosis.
• This deep learning-based approach enhances the accuracy and efficiency of CMB detection in clinical settings.
• The findings suggest a potential for improved dementia risk assessment through automated MRI analysis.