Deep Learning-Based Malaria Detection: A Robust CNN Framework for Accurate and Automated Diagnosis

This research suggests a strong framework for automated malaria detection using a Convolutional Neural Network (CNN) model. The dataset, sourced from Kaggle, consists of 27,558 high-resolution microscopic cell images, equally distributed across two classes: parasitized and uninfected. To guarantee better model results, several methods including resizing, normalization, and augmentation were performed. The proposed model, which was trained using a balanced dataset, produced a high overall accuracy of 95.73%, and high precision, recall, and F1 scores, indicating that the model can work efficiently with a wide range of samples. The study also employed class weighting to handle imbalances and employed sophisticated callbacks including EarlyStopping to avoid overfitting. The above metrics gave a detailed and precise assessment of the model, for instance, using the confusion matrix and classification report, which showed that there was very little misclassification. The framework has a lot of potential in actual implementations especially in the areas where resources are scarce. The future work can include increasing the size of the dataset to capture more regions and increasing the model's interpretability through using XAI. The integration of this system to portable diagnostic tools can greatly enhance the detection of malaria, which is affordable, accurate and easily accessible to support the fight against malaria.