In an article recently published in the journal Scientific Reports, researchers proposed a quantum machine learning (QML) and deep ensemble learning-based approach for the detection of Alzheimer's disease (AD).
Comparison of individual models using performance metrics. Image Credit: https://www.nature.com/articles/s41598-024-61452-1
Background
Alzheimer's disease is one of the most chronic neurodegenerative disorders threatening global public health. This degenerative brain disorder diminishes critical mental abilities like memory, and patients suffering from AD face diverse challenges, including mobility issues, memory loss, behavioral defects, and cognitive impairment.
Thus, early diagnosis is crucial for timely intervention and medication to improve the quality of life of affected individuals. Structural magnetic resonance imaging (MRI), an asymptomatic diagnostic method, is recognized as a common imaging biomarker in categorizing and identifying AD stages as non-demented, very mild demented, moderate demented, and mild demented.
To this end, deep learning (DL) techniques have displayed substantial improvements in disease detection, classification, and detection tasks. Ensemble learning can further boost the performance of these DL models. Additionally, QML, which is the intersection of quantum computing and classical ML, has also gained significant attention from the scientific community due to its promising speed and scalability, representation power, and flexibility.
Quantum computing provides a more efficient model for various disease classification tasks compared to classical ML approaches, with many studies demonstrating the advantages of QML algorithms over conventional ML algorithms for several applications like healthcare. However, the full quantum computing potential has not been applied to AD classification tasks.
The Study
In this work, researchers proposed an ensemble DL model based on QML classifiers for classifying AD. The proposed model can detect the AD stages from a brain MRI using an ensemble of customized versions of VGG16-ResNet50 DL models as feature extraction and then applying QML algorithms for classification.
The objective of the study was to develop an efficient model by integrating different source data to come up with a better outcome prediction to address the challenge of AD classification. The Alzheimer's Disease Neuroimaging Initiative II (ADNI2) and Alzheimer's Disease Neuroimaging Initiative I (ADNI1) MRI datasets, obtained from Kaggle, were combined for the Alzheimer's disease classification.
Key AD features extracted from the merged images/ADNI MRI image data using the customized version of the ensemble DL algorithm/VGG16 and ResNet50 models were initially combined and fed to the QML classifier.
The QML classifier/quantum support vector machine (QSVM) then classified them as very mild demented, moderate demented, mild demented, and non-demented. Thus, the proposed approach can make decisions in a more varied, thorough, and dependable manner.
Researchers evaluated the proposed model's performance using multiple metrics, including recall, precision, F1-score, the area under the curve, and accuracy, and investigated the effect of QML to effectively improve the model's computational efficiency. They also investigated the effectiveness of the model against other techniques, including convolutional neural networks (CNN), Inception-ResNet-V2, VGG16, AlexNet, and Random Forest and Adaptive Boosting classifiers.
Additionally, the researchers evaluated the performance of ensemble learning methods with the QSVM and classical SVM classifiers. They utilized a five-qubit quantum simulator/hardware and the QSVM model from the Qiskit library. The QSVM model was optimized by adding hyperparameters.
Importance of this Work
The results of individual DL models showed that ResNet had a poor performance based on F1 score and recall, while VGGNet realized outstanding results along with ensemble models. The area under the curve fared better compared to other metrics, with VGGNet, ResNet, and the ensemble model achieving 97.59, 93.70, and 98.84 scores, respectively.
Researchers examined the effects of the SVM-based DL models on the AD dataset. Results showed that the VGGNet+SVM model attained an 85 % F1 score, 85.30 % recall, 85.24 % accuracy, 89.18 % area-under-the-curve score, and 85.00 % precision, while the ResNet+SVM model realized 82.24 % accuracy, 84.73 F1-score, and 85.43 recall. The proposed ensemble (VGGNet+ResNet) with the classical SVM model realized the highest 86.78 % accuracy and 90.53 % area under the curve score.
The results of DL models with QSVM also displayed that the proposed ensemble model+QSVM had a better performance compared to VGGNet+QSVM and ResNet+QSVM, with the ensemble model+QSVM achieving the highest accuracy of 99.89, precision of 99.25, and area under the curve score of 99.99 on the merged ADNI dataset.
These results validated that the proposed ensemble model with SVM and QSVM outperforms DL with classical SVM and DL with QSVM. Moreover, the proposed model also outperformed other techniques for AD detection by achieving the highest accuracy of 99.89 on the merged ADNI1+ADNI2 dataset, surpassing the previous best accuracy of 99.68 achieved by CNN on the OASIS dataset.
To summarize, the proposed approach provides an effective solution to support AD primary care, specifically when the MRI scan is blurred, which makes it difficult for experts to suggest the disease properly.
Journal Reference
Jenber Belay, A., Walle, Y. M., Haile, M. B. (2024). Deep Ensemble learning and quantum machine learning approach for Alzheimer’s disease detection. Scientific Reports, 14(1), 1-10. https://doi.org/10.1038/s41598-024-61452-1, https://www.nature.com/articles/s41598-024-61452-1
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.