MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel
  1. Home
  2. Meeting Program
  3. All
  4. Histology Classification in Early-Stage Non-Small-Cell Lung Cancers Using Radiomic Features

Histology Classification in Early-Stage Non-Small-Cell Lung Cancers Using Radiomic Features

A Haga1,2*, W Takahashi2 , S Aoki2 , K Nawa2 , H Yamashita2 , J Kotoku3 , O Abe2 , K Nakagawa2 , (1) University of Tokushima, Tokushima, (2) The University of Tokyo Hospital, Tokyo, (3) Teikyo University

Presentations

(Sunday, 7/29/2018) 4:00 PM - 4:55 PM

Room: Room 205

Purpose: To evaluated the potential application of radiomics for predicting the histology of early-stage non-small-cell lung cancer (NSCLC) using treatment planning computed tomography (CT).

Methods: Forty patients for whom tumor biopsies was performed before stereotactic body radiation therapy (SBRT) were included in this study. Adenocarcinoma and squamous cell carcinoma were diagnosed in 21 and 19 patients, respectively. All 40 patients underwent four-dimensional CT using a 16-detector row Aquilion LB model scanner (Toshiba) with the same protocol, for use in SBRT planning. Maximum exhale CT and the corresponding gross tumor volume (GTV) delineated with FDG-PET information were used in radiomic feature extraction. A three-dimensional wavelet transform was applied to CT images, and then, various texture features as well as size-shape and histogram based features were extracted from each component. In total, 476 features were extracted from each dataset. With this training data, a histology prediction model was constructed using a naïve Bayesian model. A feature selection was performed with the correlation analysis and interobserver delineation analysis. For the validation of the model, the 29 early-stage GTV-completed datasets from the Cancer Imaging Archive (TCIA dataset) was used. Through the analysis, both training and validation data were normalized each other.

Results: The feature selection with the correlation analysis and interobserver delineation analysis showed that the number of features with a p-value < 0.1 in univariate analysis was only 8 for NSCLC histology classification. Area-under-the-curve (AUC) value was 0.895 for training data and 0.793 for validation data (TCIA dataset).

Conclusion: A radiomic analysis has shown that robust features have a high prognostic power in predicting early-stage NSCLC histology subtypes. Further study is desirable, including collection of patient data at other institutions.

Keywords

Lung, Image Analysis

Taxonomy

IM/TH- Image Analysis (Single modality or Multi-modality): Classification methods

Contact Email

Additional Links