Application of deep neural network to prostate cancer diagnosis

Zhang, Feifan

Permalink

https://hdl.handle.net/2142/104053 Copy

Description

Title

Application of deep neural network to prostate cancer diagnosis

Author(s)

Zhang, Feifan

Contributor(s)

Popescu, Gabriel

Issue Date

2019-05

Keyword(s)

• Prostate Cancer
• Semantic Segmentation
• U-Net
• Gleason Score

Abstract

Deep neural network (DNN) has been widely used in biomedical fields to help in understanding diseases. A well-trained DNN can learn inconspicuous features and give a better overall diagnosis than a human. Prostate cancer, one of the most dangerous cancers, has a 5-year survival rate of around 70% when diagnosed in Stage II but this rate decrements to 40% when diagnosed in Stage III. Clearly, we need to find a way help diagnose prostate cancer as early as possible. In order to build such a diagnosis network, we make use of the distinctive gland distribution in Prostate cells, training a U-Net (one kind of deep convolutional neural network) to do the cell semantic segmentation. This segmentation gives us a 3-label output image which labels the areas of gland, stroma and background in the input image. Based on the segmentation result, the U-net is trained to learn and output the prediction stage of prostate cancer data. This project uses Tensorflow library in Python to build and train the U-net.

Type of Resource

text

Language

en

Permalink

http://hdl.handle.net/2142/104053

Owning Collections