Identification of cognitive deterioration in elderly patients using Efficient Net B3 and compare the accuracy with convolutional neural network

• Main Authors: Mahimaa, K., Pramila, P. V., Selvaperumal, S. K., Venu, D.
• Format: Conference Proceeding
• Language: English
• Subjects: Accuracy; Artificial neural networks; Datasets; Machine learning; Medical imaging; Neural networks; Older people
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0229453

The goal of this research is to assess the cognitive deterioration in elderly subjects using an Novel Efficient net B3 model compared over Convolutional neural networks in terms of accuracy. In this study, two primary approaches were employed: the Convolutional Neural Network (CNN) and Efficient Net B3. The research utilized the "Alzheimer’s Dataset," an open-source dataset accessible through Kaggle, comprising a total of 6400 MRI images. These images were divided into training and testing sets to facilitate the analysis. 5121 pictures make up the train dataset, whereas 1279 images make up the test dataset. The value of G power=0.8 was used to calculate the accuracy of Efficient Net B3 for datasets and confidence interval of 95%. The experiment was iterated using the aforementioned models 10 times. The study revealed that Convolutional Neural Networks applied to the Alzheimer’s Dataset achieved an accuracy of 74.69%, whereas the novel technique utilizing Efficient Net B3 demonstrated a significantly higher accuracy of 92.28%. According to the t-test, the Efficient Net B3 technique appears to be more significant (p