A Deep Learning Approach for Gland Segmentation in Cancerous Tissue Images
CSEF · 2017 Mathematical Sciences
Overview
Objectives/Goals The goal of this project was to use a deep learning based fully convolutional neural network (FCN) for pixel level gland segmentation of cancer images, and test the ability of the model learnt on colon cancer images to segment glands in different types of cancer images. Methods/Materials The gland FCN model is initialized by transferring the VGG-16 learnt model, pre-trained on an image database of 1.5 million images. Matlab training and testing scripts in the FCN-Matconvnet toolbox were modified for gland segmentation, and the training was done on a computer with a GPU for 1200 epochs (iterations) over 40 hours of training. Results The algorithm was trained and validated on a colon cancer image database, for which the training and testing accuracy were approximately 96% and 90% respectively. The same model was also tested on a prostate cancer image dataset, for which the segmentation results qualitatively look accurate. Conclusions/Discussion Overall, the results from this model are promising, especially since this accuracy was achieved with a training set of less than 100 images, proving that it is possible to construct a highly accurate deep neural network model to automatically segment a multitude of glands in various cancer type tissue images.
Summary statement
This project aims to develop a deep learning algorithm based on fully convolutional neural networks and transfer learning to segment glands in cancer images, then test its ability to be able to expand to different types of cancers.
Help received
We contacted various pathologists for cancerous tissues images. After downloading the FCN-Matconvnet scripts, we contacted professionals to get more information about the specifics regarding the code. After writing the code, we requested a professional to simply run the code.
Competition history
- CSEF 2017
Resources
Related projects
CSEF · 2018
A Deep Learning Approach for Analyzing Tumor Histopathological Image Slides to Predict Breast Cancer
ISEF · 2024
Image Segmentation of Gastrointestinal Polyps in the Human Gastrointestinal Tract Using Machine Learning
ISEF · 2022
Applying Deep Learning in Recognizing Endometrial Carcinoma
ISEF · 2018
Neural Networks and Colon Polyp Detection
ISEF · 2015
Improved Cancer Detection and Diagnosis through a Novel Combination of Cell Segmentation and Artificial Intelligence Techniques
CSEF · 2019
A Novel Fuzzy Clustering-Based Convolutional Neural Network Ensemble for Histopathologic Cancer Metastasis Detection
CSEF · 2018
Improving Mitosis Detection and Localization in Breast Cancer Histology Using Deep Convolutional Neural Networks
ISEF · 2025
The Development of a U-Net Model for Pancreatic Tumor Segmentation in Computed Tomography Scans
Closest projects by meaning, across every fair and year in the corpus.
Browse more like this
Source: California Science & Engineering Fair public projects