MRI-Based Deep Learning Segmentation and Radiomics of Sarcoma in Mice.

M D Holbrook,S J Blocker,G A Johnson,A Badea,Y Qi,C T Badea,E S Xu,Y M Mowery,D G Kirsch

doi:10.18383/j.tom.2019.00021

Abstract

Small-animal imaging is an essential tool that provides noninvasive, longitudinal insight into novel cancer therapies. However, considerable variability in image analysis techniques can lead to inconsistent results. We have developed quantitative imaging for application in the preclinical arm of a coclinical trial by using a genetically engineered mouse model of soft tissue sarcoma. Magnetic resonance imaging (MRI) images were acquired 1 day before and 1 week after radiation therapy. After the second MRI, the primary tumor was surgically removed by amputating the tumor-bearing hind limb, and mice were followed for up to 6 months. An automatic analysis pipeline was used for multicontrast MRI data using a convolutional neural network for tumor segmentation followed by radiomics analysis. We then calculated radiomics features for the tumor, the peritumoral area, and the 2 combined. The first radiomics analysis focused on features most indicative of radiation therapy effects; the second radiomics analysis looked for features that might predict primary tumor recurrence. The segmentation results indicated that Dice scores were similar when using multicontrast versus single T2-weighted data (0.863 vs 0.861). One week post RT, larger tumor volumes were measured, and radiomics analysis showed greater heterogeneity. In the tumor and peritumoral area, radiomics features were predictive of primary tumor recurrence (AUC: 0.79). We have created an image processing pipeline for high-throughput, reduced-bias segmentation of multiparametric tumor MRI data and radiomics analysis, to better our understanding of preclinical imaging and the insights it provides when studying new cancer therapies.

Highlights

Because imaging is a standard means for assessing disease state and therapeutic response in clinical oncology, small-animal imaging for coclinical cancer trials enhances the simulation of clinical practice in animals
This study describes and evaluates our small-animal Magnetic resonance imaging (MRI)-based image analysis pipeline on sarcomas treated with radiation therapy (RT), including automated tumor segmentation and radiomics analysis
Our results show that convolutional neural networks (CNNs)-based segmentations with supervised learning using either T2-weighted or multicontrast MRI images are viable methods for automatic tumor volumetric measurements

Summary

Introduction

Because imaging is a standard means for assessing disease state and therapeutic response in clinical oncology, small-animal imaging for coclinical cancer trials enhances the simulation of clinical practice in animals. Assessing treatment response requires tumor measurements that are both accurate and precise. Segmentation algorithms based on convolutional neural networks (CNNs) have shown comparable efficacy in identifying tumors as other automated methods [2]. Several CNN-based methods have been proposed for tumor segmentation from multicontrast MRI, based on both 2D slices [3] or 3D volumes [4]. Many current architectures for tumor segmentation use a patch-based approach, in which a 2D or 3D patch is processed by convolutional and fully connected layers to classify the center pixel of the patch [3, 5]. Other networks operate semantic-wise by classifying each pixel in an input image or a patch using fully convolutional networks or U-nets [2, 6]. Deep learning solutions are attractive for processing multichannel, volumetric image data, where conventional processing methods are often computationally expensive [7]

Objectives

Methods

Results

Conclusion