Empowering Healthcare through Privacy-Preserving MRI Analysis
Authors:
Al Amin,
Kamrul Hasan,
Saleh Zein-Sabatto,
Deo Chimba,
Liang Hong,
Imtiaz Ahmed,
Tariqul Islam
Abstract:
In the healthcare domain, Magnetic Resonance Imaging (MRI) assumes a pivotal role, as it employs Artificial Intelligence (AI) and Machine Learning (ML) methodologies to extract invaluable insights from imaging data. Nonetheless, the imperative need for patient privacy poses significant challenges when collecting data from diverse healthcare sources. Consequently, the Deep Learning (DL) communities…
▽ More
In the healthcare domain, Magnetic Resonance Imaging (MRI) assumes a pivotal role, as it employs Artificial Intelligence (AI) and Machine Learning (ML) methodologies to extract invaluable insights from imaging data. Nonetheless, the imperative need for patient privacy poses significant challenges when collecting data from diverse healthcare sources. Consequently, the Deep Learning (DL) communities occasionally face difficulties detecting rare features. In this research endeavor, we introduce the Ensemble-Based Federated Learning (EBFL) Framework, an innovative solution tailored to address this challenge. The EBFL framework deviates from the conventional approach by emphasizing model features over sharing sensitive patient data. This unique methodology fosters a collaborative and privacy-conscious environment for healthcare institutions, empowering them to harness the capabilities of a centralized server for model refinement while upholding the utmost data privacy standards.Conversely, a robust ensemble architecture boasts potent feature extraction capabilities, distinguishing itself from a single DL model. This quality makes it remarkably dependable for MRI analysis. By harnessing our groundbreaking EBFL methodology, we have achieved remarkable precision in the classification of brain tumors, including glioma, meningioma, pituitary, and non-tumor instances, attaining a precision rate of 94% for the Global model and an impressive 96% for the Ensemble model. Our models underwent rigorous evaluation using conventional performance metrics such as Accuracy, Precision, Recall, and F1 Score. Integrating DL within the Federated Learning (FL) framework has yielded a methodology that offers precise and dependable diagnostics for detecting brain tumors.
△ Less
Submitted 14 March, 2024;
originally announced March 2024.
An Explainable AI Framework for Artificial Intelligence of Medical Things
Authors:
Al Amin,
Kamrul Hasan,
Saleh Zein-Sabatto,
Deo Chimba,
Imtiaz Ahmed,
Tariqul Islam
Abstract:
The healthcare industry has been revolutionized by the convergence of Artificial Intelligence of Medical Things (AIoMT), allowing advanced data-driven solutions to improve healthcare systems. With the increasing complexity of Artificial Intelligence (AI) models, the need for Explainable Artificial Intelligence (XAI) techniques become paramount, particularly in the medical domain, where transparent…
▽ More
The healthcare industry has been revolutionized by the convergence of Artificial Intelligence of Medical Things (AIoMT), allowing advanced data-driven solutions to improve healthcare systems. With the increasing complexity of Artificial Intelligence (AI) models, the need for Explainable Artificial Intelligence (XAI) techniques become paramount, particularly in the medical domain, where transparent and interpretable decision-making becomes crucial. Therefore, in this work, we leverage a custom XAI framework, incorporating techniques such as Local Interpretable Model-Agnostic Explanations (LIME), SHapley Additive exPlanations (SHAP), and Gradient-weighted Class Activation Mapping (Grad-Cam), explicitly designed for the domain of AIoMT. The proposed framework enhances the effectiveness of strategic healthcare methods and aims to instill trust and promote understanding in AI-driven medical applications. Moreover, we utilize a majority voting technique that aggregates predictions from multiple convolutional neural networks (CNNs) and leverages their collective intelligence to make robust and accurate decisions in the healthcare system. Building upon this decision-making process, we apply the XAI framework to brain tumor detection as a use case demonstrating accurate and transparent diagnosis. Evaluation results underscore the exceptional performance of the XAI framework, achieving high precision, recall, and F1 scores with a training accuracy of 99% and a validation accuracy of 98%. Combining advanced XAI techniques with ensemble-based deep-learning (DL) methodologies allows for precise and reliable brain tumor diagnoses as an application of AIoMT.
△ Less
Submitted 6 March, 2024;
originally announced March 2024.
