Computer-Aided Diagnosis and Staging of Pancreatic Cancer Based on CT Images

Min Li,Xiaoyi Lv,Wei Han,Shuailei Zhang,Ziwei Yan,Yushuai Yuan,Yilidan Reheman,Chen Chen,Xiaohan Nie,Cheng Chen,Pan Huang

doi:10.1109/access.2020.3012967

Abstract

Pancreatic cancer (PC) is a malignant tumor that seriously threatens the survival of patients. Artificial classification has practical difficulties, such as unstable classification accuracy, a heavy workload, and the classification results depend on the subjective judgment of the clinician during the diagnosis and staging of PC. In addition, accurate PC staging could better help clinicians deliver the optimal therapeutic schedule for PC patients of different stages. Therefore, this study proposes a comprehensive medical computer-aided method for preoperative diagnosis and staging of PC based on an ensemble learning-support vector machine (EL-SVM) and computed tomography (CT) images. The least absolute shrinkage and selection operator (LASSO) algorithm was chosen for feature selection. In contrast to no feature selection, the model optimization time decreased by 19.94 seconds while maintaining precision. The EL-SVM learner was used to classify 168 CT images of normal pancreas and different stages of PC. The experimental results demonstrated that the normal pancreas (normal)-pancreatic cancer early stage (early stage) classification accuracy was 86.61%, the normal-pancreatic cancer stage III (stage III) classification accuracy was 87.04%, the normal-pancreatic cancer stage IV (stage IV) classification accuracy was 91.63%, the normal-PC classification accuracy was 87.89%, the early stage-stage III classification accuracy was 75.03%, and the early stage-stage IV classification accuracy was 81.22%, and the stage III-stage IV classification accuracy was 82.48%. Our experimental results prove that our proposed method is feasible and promising for clinical applications for the preoperative diagnosis and staging of PC via CT images.

Highlights

Pancreatic cancer (PC) is a highly malignant tumor of the digestive tract that presents considerable challenges in both the early screening stage and later treatment [1]–[3]
Aiming at the problems in the current research on the diagnosis of PC based on Computed tomography (CT) images, this study proposed a method of computer-aided diagnosis and staging of PC based on the ensemble learning-support vector machine (EL-supported vector machine (SVM))
This study proposed a PC nonenhanced CT image classification model based on the EL-SVM

Summary

Introduction

Pancreatic cancer (PC) is a highly malignant tumor of the digestive tract that presents considerable challenges in both the early screening stage and later treatment [1]–[3]. Comprehensive preoperative diagnosis and staging of PC are important, especially in the detection of PC staging, which could better help the clinicians to deliver the optimal therapeutic schedule for different stages of PC and allow the patients to receive early medical interventions before advanced PC are formed [9], [10]. The imaging modalities commonly used in the diagnosis of PC are as follows: (1) Computed tomography (CT) is the common imaging for many patients due to low cost and high penetration. Several studies have shown that CT plays a significant role in the preoperative diagnosis of PC and has become the preferred approach for many

Methods

Results

Conclusion