Utilizing deep learning technology to develop a novel CT image marker for categorizing cervical cancer patients at early stage

Wei Liu; Abolfazl Zargaria; Theresa C. Thai; Tara Castellano; Camille C. Gunderson; Kathleen Moore; Robert S. Mannel; Hong Liu; Bin Zheng; Yuchen Qiu

doi:10.1117/12.2510037

7 March 2019 Utilizing deep learning technology to develop a novel CT image marker for categorizing cervical cancer patients at early stage

Wei Liu, Abolfazl Zargaria, Theresa C. Thai, Tara Castellano, Camille C. Gunderson, Kathleen Moore, Robert S. Mannel, Hong Liu, Bin Zheng, Yuchen Qiu

Author Affiliations +

Proceedings Volume 10879, Biophotonics and Immune Responses XIV; 108790I (2019) https://doi.org/10.1117/12.2510037
Event: SPIE BiOS, 2019, San Francisco, California, United States

Abstract

The purpose of this investigation is to verify the feasibility of using deep learning technology to generate an image marker for accurate stratification of cervical cancer patients. For this purpose, a pre-trained deep residual neural network (i.e. ResNet-50) is used as a fixed feature extractor, which is applied to the previously identified cervical tumors depicted on CT images. The features at average pooling layer of the ResNet-50 are collected as initial feature pool. Then discriminant neighborhood embedding (DNE) algorithm is employed to reduce the feature dimension and create an optimal feature cluster. Next, a k-nearest neighbors (k-NN) regression model uses this cluster as input to generate an evaluation score for predicting patient’s response to the planned treatment. In order to assess this new model, we retrospectively assembled the pre-treatment CT images from a number of 97 locally advanced cervical cancer (LACC) patients. The leave one out cross validation (LOOCV) strategy is adopted to train and optimize this new scheme and the receiver operator characteristic curve (ROC) is applied for performance evaluation. The result shows that this new model achieves an area under the ROC curve (AUC) of 0.749 ± 0.064, indicating that the deep neural networks enables to identify the most effective tumor characteristics for therapy response prediction. This investigation initially demonstrates the potential of developing a deep learning based image marker to assist oncologists on categorizing cervical cancer patients for precision treatment.

Citation Download Citation

Wei Liu, Abolfazl Zargaria, Theresa C. Thai, Tara Castellano, Camille C. Gunderson, Kathleen Moore, Robert S. Mannel, Hong Liu, Bin Zheng, and Yuchen Qiu "Utilizing deep learning technology to develop a novel CT image marker for categorizing cervical cancer patients at early stage", Proc. SPIE 10879, Biophotonics and Immune Responses XIV, 108790I (7 March 2019); https://doi.org/10.1117/12.2510037

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available