Malaria Cell Image Classification using CNN

Overview

This repository implements a Convolutional Neural Network (CNN)–based image classification system to detect malaria-infected red blood cells from microscopic cell images.

The goal of this project is to explore how deep learning can assist in automated malaria screening by classifying cell images as Parasitized or Uninfected.

This is a research and learning project, not a clinical diagnostic tool.

---

Problem Statement

Malaria diagnosis traditionally requires manual microscopic examination of blood smears, which is:

• time-consuming
• dependent on expert availability
• prone to human error at scale

This project investigates whether a CNN can learn discriminative visual patterns from cell images to classify malaria infection status automatically.

---

Dataset

• Source: Publicly available malaria cell image dataset
• Image Type: Microscopic images of red blood cells
• Classes:
  • Parasitized
  • Uninfected
• Task: Binary image classification

Note: Dataset files are not included in this repository.

---

Approach

Model Architecture

• Custom CNN built from scratch
• Multiple convolutional blocks with:
  • Convolution layers
  • Activation functions (ReLU)
  • Pooling layers
• Fully connected layers for classification
• Sigmoid / Softmax output for binary classification

Preprocessing

• Image resizing
• Normalization
• Train / validation split
• Batch-based data loading

---

Training Setup

• Framework: TensorFlow / Keras
• Loss Function: Binary Cross-Entropy
• Optimizer: Adam
• Metrics:
  • Accuracy
• Batch training with validation monitoring

---

Results

The trained CNN is able to learn meaningful visual features from microscopic images and achieves strong classification performance on the validation/test set.

Exact accuracy may vary depending on:

• training epochs
• random initialization
• dataset split