Virtual Library
Start Your Search
Ravindra Patil
Author of
-
+
P2.01 - Advanced NSCLC (ID 618)
- Event: WCLC 2017
- Type: Poster Session with Presenters Present
- Track: Advanced NSCLC
- Presentations: 1
- Moderators:
- Coordinates: 10/17/2017, 09:00 - 16:00, Exhibit Hall (Hall B + C)
-
+
P2.01-052 - Does Radiomics Improves the Survival Prediction in Non Small Cell Lung Cancer? (ID 10460)
09:00 - 09:00 | Presenting Author(s): Ravindra Patil
- Abstract
Background:
Non small cell lung cancer (NSCLC) accounts for 85% of all the lung cancer worldwide. An accurate survival time prediction is important so that subject can plan his activities and also it aids physician in arriving at the best treatment plan. There have been multiple studies that are conducted to build the survival analysis model for lung cancer. The factors that are considered in most of the models includes age, gender, tumor size, weightloss, smoking history and TNM staging to arrive at the survival prediction. However, the current focus is to make the prediction of the survival analysis more personalized and accurate. With the advent of radiomics, which deals with extraction of quantifiable features from the CT images promises to aid in personalized medicine. The objective of this work is to validate the use of radiomic features and arrive at a radiomic signature, which has better prediction power. Also, to analyze the role of radiomic features in improving the accuracy of survival prediction in NSCLC.
Method:
The dataset consist of 237 subjects CT images with NSCLC with the follow up data of 5 years with different histologies (Adenocarcinoma-39;Large cell-102 and Squamous cell carcinoma-96). The data also contained, the clinical information such as age, gender, TNM staging and the survival status. Furthermore, 432 radiomic features were extracted from the gross tumor volume of the CT images for all the corresponding subjects. Assessment was performed using cox regression model between different groups of features (clinical information, radiomic features and combination of clinical information and radiomics features) to arrive at the survival prediction model. Also, unique radiomic signature was identified with 16 features that has maximum influence on the survival model.
Result:
The results showed that the concordance Harrell’s concordance index (c-index) for only clinical information was observed to be 0.56, with only radiomics features being 0.64 and with combination of radiomics features along with clinical information was observed to be 0.69.
Conclusion:
In this study we observed that the radiomic features along with clinical information aids in providing better survival prediction model for NSCLC. Also, a unique radiomic signature was obtained which is used as an input to the survival prediction model for improving the accuracy. The study also highlights the role of imaging features driving towards personalized treatment in NSCLC.
