This repository is based on the following peer-reviewed publication:

Paper Title: Enhanced ROI guided deep learning model for Alzheimer's detection using 3D MRI images

Journal: Informatics in Medicine Unlocked, Volume 56, 2025, Page 101650

DOI: https://doi.org/10.1016/j.imu.2025.101650

Read on:

• ScienceDirect
• ResearchGate

@article{KHAN2025101650,
title = {Enhanced ROI guided deep learning model for Alzheimer's detection using 3D MRI images},
journal = {Informatics in Medicine Unlocked},
volume = {56},
pages = {101650},
year = {2025},
issn = {2352-9148},
doi = {https://doi.org/10.1016/j.imu.2025.101650},
url = {https://www.sciencedirect.com/science/article/pii/S2352914825000383},
author = {Israt Jahan Khan and Md. Fahim Bin Amin and Md. Delwar Shahadat Deepu and Hazera Khatun Hira and Asif Mahmud and Anas Mashad Chowdhury and Salekul Islam and Md. Saddam Hossain Mukta and Swakkhar Shatabda},
keywords = {Alzheimer's disease, Regions of Interest (ROIs), 3D MRI, Transfer learning, ResNet 3D, Efficient computation},
abstract = {Alzheimer's disease is an incurable condition that predominantly affects the human brain, leading to the shrinkage of various brain regions and the disruption of neuronal connections. Current state-of-the-art methods for detecting Alzheimer's disease using 3D MRI images are resource-intensive and time-consuming. In this paper, we propose a Regions of Interest (ROI)-guided detection paradigm to address these challenges. We employ a 3D ResNet integrated with a Convolutional Block Attention Module (CBAM), demonstrating that emphasising ROIs in brain imaging can substantially reduce both computational expenditure and training time. Our model exhibits robust performance in discriminating Alzheimer's disease from mild cognitive impairment, achieving an accuracy of 88% across the entire brain and 92% within targeted ROIs on the ADNI dataset. The accuracy on the OASIS dataset is even higher, reaching 98% for all regions and 98.33% for the ROIs. When distinguishing Alzheimer's disease from cognitively normal individuals, the accuracy improves further, achieving 93.33% for the ROIs on the ADNI dataset and 97.8% on the OASIS dataset. In differentiating cognitively normal individuals from those with mild cognitive impairment, the model attains an accuracy of 88.2% for the ROIs on the ADNI dataset and 98.6% on the OASIS dataset. These findings highlight a notable enhancement in detection accuracy through the utilisation of fewer, yet more salient brain regions, underscoring the efficacy of our ROI-guided approach.}
}

This project leverages advanced Convolutional Neural Networks (CNNs) for automated segmentation and analysis of 3D MRI brain images to detect Alzheimer's Disease. The core model architecture is a 3D ResNet enhanced with a Convolutional Block Attention Module (CBAM), significantly improving both model accuracy and interpretability.

Our workflow includes:

• Preprocessing (bias correction, skull stripping, CLAHE)
• Image registration and standardization
• FastSurfer-based segmentation
• ROI extraction
• Extensive data augmentation

Datasets were obtained from ADNI and OASIS, and experiments focused on classifying subjects into:

• Alzheimer's Disease (AD)
• Mild Cognitive Impairment (MCI)
• Cognitively Normal (CN)

This method is particularly effective for early-stage detection, guiding clinical decision-making through high-accuracy, ROI-focused models.

Alzheimer's disease is an incurable condition that predominantly affects the human brain, leading to the shrinkage of various brain regions and the disruption of neuronal connections. Current state-of-the-art methods for detecting Alzheimer's disease using 3D MRI images are resource-intensive and time-consuming. In this paper, we propose a Regions of Interest (ROI)-guided detection paradigm to address these challenges. We employ a 3D ResNet integrated with a Convolutional Block Attention Module (CBAM), demonstrating that emphasising ROIs in brain imaging can substantially reduce both computational expenditure and training time. Our model exhibits robust performance in discriminating Alzheimer's disease from mild cognitive impairment, achieving an accuracy of 88% across the entire brain and 92% within targeted ROIs on the ADNI dataset. The accuracy on the OASIS dataset is even higher, reaching 98% for all regions and 98.33% for the ROIs. When distinguishing Alzheimer's disease from cognitively normal individuals, the accuracy improves further, achieving 93.33% for the ROIs on the ADNI dataset and 97.8% on the OASIS dataset. In differentiating cognitively normal individuals from those with mild cognitive impairment, the model attains an accuracy of 88.2% for the ROIs on the ADNI dataset and 98.6% on the OASIS dataset. These findings highlight a notable enhancement in detection accuracy through the utilisation of fewer, yet more salient brain regions, underscoring the efficacy of our ROI-guided approach.

In the first phase, we gathered 3D MRI datasets from ADNI and OASIS, renowned initiatives for advancing the understanding of Alzheimer's disease. We categorized the data into Alzheimer's Disease (AD), Mild Cognitive Impairment (MCI), and Cognitively Normal (CN) classes. Additionally, we obtained CSV files containing detailed patient information for labeling. Strict privacy rules were followed to ensure the security of sensitive medical data.

For our research, we categorized the dataset into three distinct groups: AD (Alzheimer's Disease), CN (Cognitively Normal), and MCI (Mild Cognitive Impairment).

We utilized the ADNI Screening 1.5T dataset, which includes subjects who underwent a screening scan. We collected Neuroimaging Informatics Technology Initiative (NIfTI) images from 2294 subjects.

The dataset includes 416 patients aged 18 to 96. Each person undergoes three or four T1-weighted MRI scans in a single session. Among the participants over sixty, 100 have been diagnosed with Alzheimer's disease. Additionally, there is a reliability dataset with 20 nondemented participants.

This repository includes files to help you recreate the development environment used in this project. Follow the instructions below to set up the environment on your local machine.

• environment.yml: Captures the full Conda environment, including both Conda and pip-installed packages.
• requirements.txt: Lists only pip-installed packages, useful for users who prefer pip-based installations.

• Install Miniconda or Anaconda.

1. Open a terminal and navigate to the project directory.
2. Run the following command to create the environment:

   conda env create -f environment.yml

3. Activate the newly created environment:

   conda activate fydp-old

4. Verify the environment:

   conda list

• Ensure Python and pip are installed on your system.

1. Create a virtual environment:

   python -m venv venv

2. Activate the virtual environment:
   • On Linux/macOS:

     source venv/bin/activate

   • On Windows:

     venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

4. Verify the installation:

   pip list

• Preferred Method: The environment.yml file is recommended as it fully captures both Conda and pip-installed packages, ensuring 100% reproducibility.
• Pip Limitation: The requirements.txt file only includes pip-installed packages and may not recreate Conda-specific dependencies or system-level requirements.

This repository is archived and no further maintenance is planned. If you have any questions regarding the research, the methodology, or the published paper, please contact the team lead:

Md. Fahim Bin Amin

Email: fahimbinamin@gmail.com