Abstract:
This project trained and validated the Alexnet convolutional neural network algorithm
using breast histopathology image data. The Alexnet neural network is one of many deep
learning algorithms that has been utilized to preform supervised learning classi cation tasks.
The goal of the project was to accurately classify breast histopathology images into the cat-
egories invasive ductal carcinoma positive and negative. The algorithm and supported data
manipulation scripts were implemented using MATLAB software utilizing the transfer learn-
ing capabilities of the Deep Learning Toolbox. The results of the trained network validation
indicated a correct classi cation rate of approximately 85%. Alternative performance metrics
indicated a high false classi cation positive rate. The performance metrics in their totality
demonstrate a practical ability of the trained network to serve as an automation tool to
assist medical experts in their detection of invasive ductal carcinoma.
