Digital Health Innovation and Informatics

PV QA 2 - Poster Viewing Q&A 2

MO_17_2627 - Quantitative Analysis of Weight of Prognostic Factors Related to Radiation Pneumonitis using Statistical Analysis and Artificial Neural Network

Monday, October 22
10:45 AM - 12:15 PM
Location: Innovation Hub, Exhibit Hall 3

Quantitative Analysis of Weight of Prognostic Factors Related to Radiation Pneumonitis using Statistical Analysis and Artificial Neural Network
E. Ju1, S. Lee1, K. H. Kim1,2, S. W. Choi3, K. H. Chang4, Y. J. Cao5, J. B. Shim1, N. K. Lee1, D. S. Yang1, W. S. Yoon1, Y. J. Park1, and C. Y. Kim1; 1Department of Radiation Oncology, College of Medicine, Korea University, Seoul, Korea, Republic of (South), 2Proton Therapy Center, National Cancer Center, Gyeonggi, Korea, Republic of (South), 3MEDICALSTANDARD Co.,Ltd., Gyeonggi-do, Korea, Republic of (South), 4Department of Radiation Oncology, College of Medicine, Yonsei University, Seoul, Korea, Republic of (South), 5Department of Radiation Oncology, Tianjin Medical University Cancer Institute and Hospital, Tianjin, 300060, China

Purpose/Objective(s): Radiation pneumonitis is one of the major radiation toxicity of radiation therapy for thoracic tumors. Numerous studies have been conducted to investigate the factors influencing the development of radiation pneumonitis. However, there is a limit to quantify the impact of each prognostic factor on the incidence of radiation pneumonitis. The aim of this study is to investigate the correlation between each prognostic factors and radiation pneumonitis using correlation analysis, and to quantify the weighting factor of each prognostic factor on pneumonitis using logistic regression and artificial neural network.

Materials/Methods: We retrospectively analyzed 110 lung cancer patients treated with radiotherapy in the Republic of Korea. The incidence of radiation pneumonitis was evaluated using the Common Toxicity Criteria (CTC) 3.0. Correlation analysis was performed on 25 prognostic factors to determine Pearson's linear correlation coefficient and Spearman's rho. P-values were calculated from Student's t distribution to confirm whether the correlation coefficients were statistically significant using using a multi-paradigm numerical computing environment and proprietary programming language. The significance values of prognostic factors were analyzed using the logistic regression and the artificial neural network using statistical software. The results of this study were validated by comparing the odds ratio of the meta-analysis literature.

Results: There were 24 patients (21.8%) who had ≥grade 2 pulmonary toxicity. In the entire population, the correlation analysis revealed that MLD (Mean lung dose), CCRT (Concurrent chemotherapy), Smoking, Histology, and Tumor location were significantly associated with radiation pneumonitis (p<0.05). Logistic regression analysis showed that the weights were high in the order of CCRT, MLD, and V20, and were 1.12, 0.47, and 0.25, respectively. The odds ratios of the meta-analysis literature were higher in order of MLD, V20, and CCRT, respectively, and 3.20, 2.80, and 1.41, respectively. Among the significant prognostic factors(p<0.05), MLD has the highest correlation (r: 0.39, rs: 0.31) and the highest weighting factor related to radiation pneumonitis from logistic regression (Weight: 0.47) and from meta-analysis literature (OR: 3.20), also MLD has the highest weighting factor among the 7 prognostic factors from the artificial neural network (MLD: 41%, V20: 26%, Age: 15%, CCRT: 6%, V30: 6%, Lung disease: 3%, Tumor location: 3%).

Conclusion: We propose a method to quantify the importance of radiation toxicity incidence using artificial neural network and evaluate the method using statistical analysis. The quantitative analysis method of weighting factors of prognostic factors related to the radiation toxicity in this study would be clinically useful for minimizing the risk of incidence of radiation toxicity in patients treated with radiotherapy.

Author Disclosure: E. Ju: None. S. Lee: None. S. Choi: None. K. Chang: None.

Send Email for Eunbin Ju


Assets

MO_17_2627 - Quantitative Analysis of Weight of Prognostic Factors Related to Radiation Pneumonitis using Statistical Analysis and Artificial Neural Network



Attendees who have favorited this

Please enter your access key

The asset you are trying to access is locked. Please enter your access key to unlock.

Send Email for Quantitative Analysis of Weight of Prognostic Factors Related to Radiation Pneumonitis using Statistical Analysis and Artificial Neural Network