Image-based Classification of Tumor Type and Growth Rate using Machine Learning: a preclinical study

Tien T Tang,Janice A Zawaski,M Waleed Gaber,Kathleen N Francis,Amina A Qutub

doi:10.1038/s41598-019-48738-5

Tien T Tang, Janice A Zawaski + Show 3 more

Open Access

https://doi.org/10.1038/s41598-019-48738-5

Copy DOI

Abstract

Medical images such as magnetic resonance (MR) imaging provide valuable information for cancer detection, diagnosis, and prognosis. In addition to the anatomical information these images provide, machine learning can identify texture features from these images to further personalize treatment. This study aims to evaluate the use of texture features derived from T1-weighted post contrast scans to classify different types of brain tumors and predict tumor growth rate in a preclinical mouse model. To optimize prediction models this study uses varying gray-level co-occurrence matrix (GLCM) sizes, tumor region selection and different machine learning models. Using a random forest classification model with a GLCM of size 512 resulted in 92%, 91%, and 92% specificity, and 89%, 85%, and 73% sensitivity for GL261 (mouse glioma), U87 (human glioma) and Daoy (human medulloblastoma), respectively. A tenfold cross-validation of the classifier resulted in 84% accuracy when using the entire tumor volume for feature extraction and 74% accuracy for the central tumor region. A two-layer feedforward neural network using the same features is able to predict tumor growth with 16% mean squared error. Broadly applicable, these predictive models can use standard medical images to classify tumor type and predict tumor growth, with model performance, varying as a function of GLCM size, tumor region, and tumor type.

Highlights

The prognoses for gliomas are extremely poor compared to medulloblastomas for the current standard of care, which includes surgical resection, radiation therapy and/or chemotherapy, in which 5 year survival is 15–35% compared to 70%, respectively[9,10]
We focused on texture features, derived from the gray level co-occurrence matrix (GLCM) which are commonly used in many different texture analysis
To construct our classification and prediction models, texture features were first extracted from the tumor region using in-house MATLAB program for three different types of tumors: GL261, U87 and Daoy

Summary

Introduction

The prognoses for gliomas are extremely poor compared to medulloblastomas for the current standard of care, which includes surgical resection, radiation therapy and/or chemotherapy, in which 5 year survival is 15–35% compared to 70%, respectively[9,10]. The GLCM is a square matrix that captures the frequency in which a combination of gray scale intensities occur with the dimensions determined by the number of gray levels[23] Features derived from this matrix are informative of the spatial relationships between grayscale intensities such as amount of variation, disorder, or contrast within an image[23]. These features, can be sensitive to image processing which include acquisition, reconstruction protocols, and inter-scanner variability[24,25,26,27,28] Independent of these systemic variations, the values of these features can be affected by the GLCM size or the number of gray levels which is determined a priori to feature extraction. Using machine learning we assess whether radiomics approaches have the potential to classify tumor type and predict tumor growth rate noninvasively by allowing clinicians to make better informed treatment decisions using standard medical images

Objectives

Methods

Results

Discussion

Conclusion