/Medical-Imaging-Malaria-Detection

CNNs for Malaria

Primary LanguageJupyter Notebook

Malaria Image Analysis using CNNs

Designing and developing a CNN for malaria image analysis

This Deep learning + medical imaging system can help reduce the 400,000+ deaths per year caused by malaria.

Data Characteristics

Size – 407.5 MB

Source – National Library of Medicine

Shape –  The dataset contains a total of 27,558 cell images with equal instances of parasitized and uninfected cells

Labelled Target – (Parasitized/Uninfected)

Steps

  1. Downloaded data and cleaned and organized the images.

  2. Feature extraction using pretrained neural networks - VGG 16 Alt text

  3. Defining a model - Basic Vs. Multilayer Perceptron - A classifier that sits on top of the feature extractor.

Best Model Metrics

Alt text

  1. Visuvalize the predictions Alt text

  2. Future Work:

     Try Conv2d and Max Pooling with strides
 Augment the data to prevent overfitting 
 Also considering using an All-CNN model as it generally performs significantly better than VGG-16.