Deep Learning Advancements in Malaria Diagnosis: A PyTorch-Based Ensemble Approach for Image Classification

Deep Learning Advancements in Malaria Diagnosis: A PyTorch-Based Ensemble Approach for Image Classification

Saravana Kumar, Saraswathi Meena R., Hirthick S., Surya Devi B.
Copyright: © 2024 |Pages: 25
DOI: 10.4018/979-8-3693-7462-7.ch010
OnDemand:
(Individual Chapters)
Available
$37.50
Current Special Offers
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

The authors introduce a robust convolutional neural network (CNN) model for malaria-infected cell identification, achieving over 96.5% test accuracy using PyTorch and GPU acceleration. Data augmentation ensures dataset suitability, while this MosquitoNet CNN architecture effectively extracts hierarchical features through three convolutional and fully linked layers. Training over 20 epochs with cross-entropy loss and Adam optimizer yields high accuracy on independent testing subsets, supported by detailed class-wise metrics and a confusion matrix visualization. This approach integrates deep learning, data augmentation, and advanced visualization for comprehensive malaria detection, promising significant advancements in medical diagnostics. Future work may explore hyperparameter tuning and transfer learning for further enhancement. This research contributes to the field with its robust methodology and high accuracy, offering a promising tool for malaria diagnosis and beyond.
Chapter Preview
Top

Background

The discipline of malaria detection has made significant strides in recent years thanks to the integration of cutting-edge technologies and innovative methodologies. A comprehensive review of the literature on the topic reveals that a number of investigations have been carried out with the aim of enhancing the accuracy and efficacy of techniques for identifying cells infected with malaria.

Key Terms in this Chapter

Deep Residual Networks: An advanced type of neural network that introduces skip connections or shortcuts to jump over some layers. This approach helps mitigate the vanishing gradient problem and allows for the training of much deeper networks.

MosquitoNet: A deep learning-based CADx (Computer-Aided Diagnosis) system designed specifically for malaria diagnosis. It employs a combination of CNNs for feature extraction and traditional machine learning models for classification.

Data Augmentation: A technique used to increase the diversity of data available for training models without collecting new data. In the context of image processing, this can include transformations such as rotations, translations, and scaling, which help improve the robustness of the model.

Grad-CAM (Gradient-weighted Class Activation Mapping): A visualization technique for deep learning networks that highlights the regions in an image that are important for the prediction. This helps interpret and understand the model's decision-making process.

Complete Chapter List

Search this Book:
Reset