Breast Cancer Detection (BCD)

The aim of this project is to predict the state of breast cancer through mammograms using Neural Network.

Motivation

I like this topic.

Data Selecting

• The dataset I used is from here: https://github.com/LearnToCode180/Breast-Cancer-Classification/blob/master/DataSet.zip
• It consists of 7632 labeled breast-cancer-related images (benign vs malignant). The dataset was then loaded into Google Drive, which was corporated into Google Colab Framework, due to the convenience of preconfigured GPU/TPU environments without the need to install CUDA Toolkit, cuDNN, or Visual Studio locally.

Data Preprocessing

• Data Augmentation was implemented on the training dataset.

Network Architecture

• I used Transfer Learning for this model. The base model is DenseNet121 with rmsprop optimizer. After that, I added a Fully Connected Layer with 128 nodes using Relu activation Function, a DropOut(0.5), and finally an output layer with 1 node using Sigmoid Function.

Result

• The accuracy of the model is 83%.

Future

• I will continue working on this project in the future, probably add Regularization Methods because the model still experience slight overfitting.