Identification of Alzheimer Detection Using Hybrid Machine Learning Approach

Abstract

Early detection of Alzheimer\'s is very important. The early and precise detection of Alzheimer’s disease-associated side effects and fundamental malady pathology by clinicians is essential for the screening, determination, and ensuing administration of Alzheimer’s illness patients. It moreover empowers patients and their caregivers to arrange for a long-standing time and make fitting way-of-life changes that might offer assistance to keep up their quality of life for longer. Tragically, recognizing early-stage Alzheimer’s illness in clinical hone can be challenging and is hindered by a few boundaries counting imperatives on clinicians’ time, trouble precisely diagnosing Alzheimer’s pathology, which patients and healthcare suppliers frequently expel indications as portion of the typical maturing prepare. As the predominance of this infection proceeds to develop, the current show for Alzheimer’s infection conclusion and persistent administration will be got to advance to coordinated care over clinical disciplines and the infection continuum, starting with essential care

Introduction

I. INTRODUCTION

In recent years, machine learning techniques have gained prominence in precise disease diagnosis and prognosis, applied to conditions such as cancer, thyroid disorders, and COVID-19. Alzheimer's disease is anticipated to rank as the sixth leading cause of death and the most significant global socio-economic burden by 2050. Machine learning methods show potential in utilizing voice frequency analysis to monitor the advancement of subjective ailments like Parkinson's disease (PD), offering non-intrusive and rapid diagnostic tools. The emphasis on personalized medicine for PD is critical given the distinctiveness of each case, where customized diagnosis and treatment result in improved clinical results. The capability of machine learning to discern intricate data patterns and automate data processing establishes it as a valuable asset for precision medicine in cases of this disorder.

A range of machine learning algorithms, including K-Nearest Neighbor (KNN), Support Vector Machine (SVM), and Random Forest, have been utilized in the diagnosis and prognosis of this condition. Moreover, deep learning techniques like convolutional neural networks (CNN) display potential in medical imaging diagnostics and forecasts. Nevertheless, the use of unsupervised and incremental learning strategies should be considered to accurately anticipate the progression of the ailment, especially with substantial datasets. Subsequent research should strive to formulate a comprehensive ML model encompassing all symptoms associated with the condition as input parameters, enabling the identification of diverse PD symptoms through a compact, wearable, and washable apparatus.

The automation of medical image interpretation through machine learning holds significance in minimizing diagnostic duration and enhancing accuracy, particularly considering the scarcity of radiologists. The integration of machine learning algorithms into medical device design stands to significantly elevate diagnostic precision and expedite decision-making procedures. This survey endeavors to examine diverse machine learning and deep learning-based methodologies employed in diagnosing this ailment, presenting insights into prevailing research trends. By enhancing the precision and efficiency of digital workflows, machine learning has the potential to propel the diagnosis and management of this condition, ultimately enhancing patient well-being.

II. SYMPTOMS

Alzheimer’s disease is a condition that affects the brain and memory in a worsening manner. People, with this disease often struggle to remember information leading to questioning and forgetfulness about important events or appointments. As the disease progresses language difficulties can arise, making it challenging for individuals to find the words to form sentences or follow conversations. Decision-making, problem-solving, and planning tasks may also become more difficult causing confusion and frustration. Additionally, behavioral changes such as mood swings, social withdrawal, and heightened anxiety can occur due to situations. In stages of Alzheimer’s disease, individuals may lose the ability to recognize loved ones and familiar places and increasingly rely on others for care and support. The wide range of symptoms associated with Alzheimer’s highlights the need for research into treatments and interventions that address cognitive abilities as well, as emotional well-being. to the growing body of knowledge on inclusive leadership and its potential to transform educational systems toward greater equity and social justice.

III. LITERATURE SURVEY

1. In a study by James. E. Galvin, et al. (2020), the heightened risk of Alzheimer's disease among rural, ethnically diverse residents compared to their urban counterparts was explored. The research highlighted the impact of chronic diseases like diabetes and hypertension on increasing dementia susceptibility, especially in regions with limited access to specialized healthcare providers. To address this issue, the study tested a novel approach aimed at enhancing dementia detection and treatment rates within government-assisted independent living facilities among rural residents. The participants (N = 139) were predominantly non-White (78%) and exhibited low health literacy (70% below 5th grade). Cognitive screening identified 28 residents at risk, of whom 25 consented to in-depth assessments by adult gerontological nurse practitioners (AGNP). Remarkably, 60% (15 out of 25) of those completing subsequent primary provider referrals were diagnosed with dementia, marking a statistically significant outcome ([χ2(1) = 76.67, p < .001, Phi = 0.743]). The study's findings underscore the potential of a community engagement approach to home-based dementia management, aligning with the objectives of Healthy People 2030 by facilitating early intervention, and treatment, and reducing the need for costly institutionalization.
2. In the study by L. Jin et al. (2023), a pioneering approach for the detection of Alzheimer's disease and the prediction of Mini-Mental State Examination (MMSE) scores is introduced. Their innovative methodology, termed Complementary and Simultaneous Ensemble (CONSEN) algorithm, leverages multilingual spontaneous speech data. By defining speech pauses and interventions to extract disfluency features and incorporating various acoustic features, the researchers developed generalized models. Through the implementation of the CONSEN algorithm, their model achieved remarkable results, boasting an outstanding 86.69% accuracy in Alzheimer's disease detection and a low 3.727 root mean square error (RMSE) for MMSE prediction. Impressively, these outcomes secured the first-place ranking in both tasks within the ICASSP Signal Processing Grand Challenge: ADReSS-M Challenge 2023. This work demonstrates the potential of utilizing speech-based features and advanced algorithms to advance Alzheimer's disease diagnosis and prediction methodologies.
3. In a study by P. Kishore, et al.(2021), the authors introduced an innovative approach for Alzheimer's disease detection and prediction using artificial intelligence and machine learning. Addressing the complexity of Alzheimer's impact on memory and cognition, the study emphasized the absence of a single diagnostic test. Instead, relying solely on brain scans is insufficient, and clinicians currently consider familial reports and medical history. The paper proposed the integration of AI and machine learning to enhance this diagnostic process. Engaging big data processing due to the multifaceted nature of information sources, the study presented a significant processing model within data mining. By employing diverse classifiers, the research trained on Alzheimer's rates and characteristics, treating them as a complex system through various machine learning algorithms. While prior research found limitations in Alzheimer's detection accuracy using Support Vector Machine classifiers, this study explored alternative algorithms. Impressively, the Support Vector Machine with a linear kernel model outperformed others in accuracy. This work holds the promise of advancing Alzheimer's detection precision through innovative machine-learning techniques.
4. In this study by Madgi, M, et al.( 2022), the authors conduct a comprehensive review of Alzheimer's disease (AD) and explore two machine learning (ML) methods for its early-stage identification. Alzheimer's disease, a neurocognitive disorder, primarily afflicting individuals in their early years, leads to memory loss, behavioral changes, and language difficulties. Early detection is crucial for advancing treatments. Machine learning, a subset of Artificial Intelligence, employs probabilistic and optimization techniques to enable computers to learn from intricate data. Researchers commonly employ machine learning to diagnose early-stage AD. The survey provides an extensive overview of ongoing research in this domain, scrutinizing classification methods employed in ADNI datasets. It examines key aspects such as dataset utilization, evaluation metrics, and ML methodologies. The study proposes a model that synthesizes current research, highlighting challenges and opportunities for innovation. It particularly focuses on two methods: the 18-layer convolutional network and the 3D convolutional network, demonstrating that multi-layered CNNs offer superior accuracy compared to 3D CNNs, achieving a remarkable 98% accuracy for the 18-layer CNN in early Alzheimer's prediction using ADNI data. The study significantly contributes to the realm of ADNI dataset utilization and its optimal division for classification purposes. Overall, the research centers on early Alzheimer's prediction via machine learning, underlining the promising accuracy potential achievable with advanced ML methodologies.
5. In this study by S. S. Hammed. et al.( 2022), the authors conduct a study devised four distinct machine learning models for the purpose of detecting Alzheimer's disease, which is categorized as a classification challenge. The evaluation encompassed several classification algorithms, including logistic regression, support vector classifier, decision tree, and random forest classifier. To optimize the models, parameter values influencing accuracy were meticulously fine-tuned. This was achieved through a K-fold cross-validation score, allowing the generation of refined models.

The dataset employed originated from longitudinal cross-sectional data extracted from the OASIS database. Notably, the results indicate that the random forest classifier outperforms other models, showcasing its superior performance in Alzheimer's disease identification.

IV. ARCHITECTURE

A. Pre-processing of Brain MRI and CT Scanned Images

The initial stage encompasses the preparation of brain images for subsequent analysis. Direct utilization of MRI, CT scans, and molecular images as inputs for machine learning classifiers is unfeasible. Therefore, a set of pre-processing techniques is employed, encompassing actions like resizing to consistent dimensions, rectifying brightness variations, filtering, addressing illumination inconsistencies, refining focus, eliminating noise, applying thresholding, introducing geometric alterations, and converting to grayscale. These pre-processing procedures serve the purpose of rendering the images suitable and primed for subsequent analytical phases.

B. Feature Extraction

For every MRI or scanned brain image, we conduct feature extraction to gather meaningful and important attributes from the refined data. Feature extraction is a method that simplifies raw information into smaller, more understandable groups, all the while accurately representing the initial dataset. We employ machine learning methods to extract specific attributes from the enhanced brain images, which are then used as inputs for the classifiers.

C. Classification using CNN

Utilizing their capacity to learn hierarchical patterns, CNNs are adept at analyzing brain images, making them invaluable for Alzheimer's disease classification. By fine-tuning pre-trained architectures like VGG16 or ResNet on your dataset, their effectiveness for the task is enhanced. The CNN's output materializes as a succinct feature vector, embodying crucial traits from brain images. This vector holds pivotal information, pivotal for distinguishing between healthy and Alzheimer 's-affected brains. As a result, the CNN operates as a potent feature extractor, significantly aiding the subsequent hybrid classification process for precise Alzheimer's disease identification.

D. Classification using SVM(Hybrid Approach)

In the hybrid approach for Alzheimer's disease classification, Support Vector Machines (SVM) play a key role. SVMs are renowned for their ability to find optimal decision boundaries in high-dimensional spaces, making them well-suited for complex data like brain images. After extracting features from the images using a Convolutional Neural Network (CNN), these features are then fed into the SVM for classification.

The SVM learns to differentiate between healthy and Alzheimer 's-affected brains by finding the most effective hyperplane that separates the feature vectors. Through this process, the SVM captures intricate relationships among the features, aiding accurate classification. Regularization parameters and kernel functions, such as linear or radial basis function kernels, can be fine-tuned for optimal performance.

By incorporating the extracted features from the CNN with SVM's classification capabilities, the hybrid approach ensures a robust and accurate Alzheimer's disease identification model. This combination leverages the strengths of both techniques to enhance the overall classification performance, facilitating improved diagnostic accuracy and potentially aiding medical professionals in early detection and intervention.

E. Assessment of Accuracy

The evaluation of machine learning classifiers will be conducted through diverse metrics, including Confusion Matrix, Classification Accuracy, Sensitivity Analysis, Precision, Area under the Curve, and Sensitivity Matrices. These metrics yield significant understanding regarding the precision and correctness of the classifiers in identifying brain disorders. Key terms in the Confusion Matrix—True Positives, True Negatives, False Positives, and False Negatives—play a crucial role in assessing classifier effectiveness. Additionally, Precision, Sensitivity, and Specificity matrices contribute to a comprehensive assessment of the precision and efficiency of the classification algorithms.

Conclusion

This groundbreaking study holds immense significance by using advanced machine-learning techniques to predict progressive brain diseases. Conditions like Alzheimer\'s disease and multiple sclerosis pose complex challenges with broad societal impacts. What makes this research extraordinary is its potential to revolutionize early detection and achieve an impressive 90 percent accuracy in predicting these serious diseases. By creating advanced predictive models, the study aims to provide healthcare professionals with powerful tools for identifying high-risk individuals. This could lead to timely interventions, better treatment results, and improved quality of life for those affected. Furthermore, this study represents a noteworthy merging of healthcare and machine learning, seamlessly combining data analysis and technology with medical expertise. The models\' ability to analyze extensive clinical records, genetic data, and brain scans opens new avenues for personalized medicine. By harnessing machine learning in healthcare, we imagine a future where diagnosis and treatment become more accurate, efficient, and patient-centered.

References

[1] J. Gaugler, B. James, T. Johnson, I. Marin, and J. Weuve, “Alzheimer’s Disease Facts and Figures. Alzheimer’Association. Washinton DC [internet]. 2019 [acceso: 23/11/2020]; 15 (3):87-321. [2] L. Jin et al., \"CONSEN: Complementary and Simultaneous Ensemble for Alzheimer’s Disease Detection and MMSE Score Prediction,\" ICASSP 2023 - 2023 IEEE Int.Conf. on Acoustics, Speech and Signal Processing (ICASSP), Rhodes Island, Greece, 2023, pp. 1-2. [3] P. Kishore, C. U. Kumari, M. Kumar, and T. Pavani, “Detection and analysis of Alzheimer’s disease using various machine learning algorithms,” Mater. today Proc., vol. 45, pp. 1502–1508, 2021. [4] Patil, V., Madgi, M. & Kiran, A. Early prediction of Alzheimer\'s disease using convolutional neural network: a review. Egypt J Neurol Psychiatry Neurosurg 58, 130 (2022). https://doi.org/10.1186/s41983-022-00571-w [5] 58. S. Pavalarajan, B. A. Kumar, S. S. Hammed. et al., \"Detection of Alzheimer\'s disease at Early Stage using Machine Learning,\" 2022 International Conference on Advanced Computing Technologies and Applications (ICACTA), Coimbatore, India, 2022, pp. 1-5, doi: 10.1109/ICACTA54488.2022.9752827.

Copyright

Copyright © 2023 Laukik Khade, Hardik Kotangale, Girish Navale, Dr. Archana Bhise. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.