GLRLM feature by using CT texture analysis is prognostic factor , Evaluation of pancreatic cancer heterogeneity

Date 24 November 2018
Event ESMO Asia 2018 Congress
Session Poster display - Cocktail
Topics Imaging
Pancreatic Cancer
Presenter HyungSun Kim
Citation Annals of Oncology (2018) 29 (suppl_9): ix46-ix66. 10.1093/annonc/mdy432
Authors H. Kim1, J.S. Park2, Y.J. Kim3, K.G. Kim4
  • 1Department Of Surgery, Gangnam Severance Hospital, Yonsei University College of Medicine, 135-720 - Seoul/KR
  • 2Department Of Surgery, Yonsei University, Seoul/KR
  • 3Department Of Biomedical Engineering, Gachon University College of Medicine, 21565 - Incheon/KR
  • 4Department Of Biomedical Engineering, Gachon University College of Medicine, Incheon/KR

Abstract

Background

Pancreatic cancer is the most lethal disease. Chemotherapy resistance is very important to predict the prognosis (recurrence or metastasis) of the patient after surgery in pancreatic cancer patients. Tumour heterogeneity is considered to be an important factor in the progression of chemoresistance. In pancreatic cancer, a few studies reported analysis of tumour heterogeneity by texture analysis. These texture analyses have the advantage of quantifying tumour heterogeneity. So, in our study, we performed analysis of tumour heterogeneity in preoperative CT scan by texture analysis using GLRLM (Gray level run length matrices) and analysed correlation of survival and these values.

Methods

We analysed 116 consecutive patients who underwent surgical resection during 2001-2017, had pre-operative contrast enhanced CT available for analysis (prior to surgery). A ROI was drawn on all the slices with a visible tumour and normal pancreas on the arterial phase CT scan. We analysed correlation of pathological characteristics and GLRLM features. And performed Kaplan meier survival curve among pancreatic cancer patients.

Results

The values of GLN in GLRLM features in tumour are higher than normal pancreas. The high GLN values represent non-uniform texture, heterogeneity. Kaplan meier survival curve using these GLRLM features showed recurrence free survival was shorter in the group with high GLN135 values(p = 0.025). High GLN135 value associated with poor prognosis.

Conclusions

Tumour heterogeneity is a significant as a prognostic factor. According to analyze correlation of pathological outcomes and GLRLM features in pancreatic cancer patients, we found powerful value, GLN in GLRLM features to predict prognosis.

Editorial acknowledgement

Clinical trial identification

Legal entity responsible for the study

Joon Seong Park.

Funding

Has not received any funding.

Disclosure

All authors have declared no conflicts of interest.