Research Output
Classification of Multiple Visual Field Defects using Deep Learning
Analysis of the effectiveness of Metaheuristic methods on Bayesian optimization in the classification of visual field defects
1 - 2 of 2
Publication
In this work, a custom deep learning method is proposed to develop a detection of visual fields defects which are the markers for serious optic pathway disease. Convolutional Neural Networks (CNN) is a deep learning method that is mostly used in images processing. Therefore, a custom 10 layers of CNN algorithm is built to detect the visual field defect. In this work, 1200 visual field defect images acquired from the Humphrey Field Analyzer 24-2 collected from Google Image have been used to classify 6 types of visual field defect. The defect patterns are including defects at central scotoma, right/left/upper/lower quadratopia, right/left hemianopia, vision tunnel, superior/inferior field defect and normal as baseline. The custom designed CNN is trained to discriminate between defect patterns in visual field images. In the proposed method, a mechanism of pre-processing is included to improve the classification of visual field defects. Then, the 6 visual field defect patterns are detected using a convolutional neural network. The dataset is evaluated using 5-fold cross-validation. The results of this work have shown that the proposed algorithm achieved a high classification rate with 96%. As comparison, traditional machine learning Support Vector Machine (SVM) and Classical Neural Network (NN) is chose and obtained classification rate at 74.54% and 90.72%.
Publication
Bayesian optimization (BO) is commonly used to optimize the hyperparameters of transfer learning models to improve the model’s performance significantly. In BO, the acquisition functions direct the hyperparameter space exploration during the optimization. However, the computational cost of evaluating the acquisition function and updating the surrogate model can become prohibitively expensive due to increasing dimensionality, making it more challenging to achieve the global optimum, particularly in image classification tasks. Therefore, this study investigates and analyses the effect of incorporating metaheuristic methods into BO to improve the performance of acquisition functions in transfer learning. By incorporating four different metaheuristic methods, namely Particle Swarm Optimization (PSO), Artificial Bee Colony (ABC) Optimization, Harris Hawks Optimization, and Sailfish Optimization (SFO), the performance of acquisition function, Expected Improvement (EI), was observed in the VGGNet models for visual field defect multi-class classification. Other than EI, comparative observations were also conducted using different acquisition functions, such as Probability Improvement (PI), Upper Confidence Bound (UCB), and Lower Confidence Bound (LCB). The analysis demonstrates that SFO significantly enhanced BO optimization by increasing mean accuracy by 9.6% for VGG-16 and 27.54% for VGG-19. As a result, the best validation accuracy obtained for VGG-16 and VGG-19 is 98.6% and 98.34%, respectively.