Volume 25, Issue 12 (3-2019)
RJMS 2019, 25(12): 84-95 |
Back to browse issues page
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Kazemi A, Fazeli M, Haghighat S. Application of decision tree and logistic regression algorithms to predict lymphedema in breast cancer patients. RJMS 2019; 25 (12) :84-95
URL: http://rjms.iums.ac.ir/article-1-5305-en.html
URL: http://rjms.iums.ac.ir/article-1-5305-en.html
University of Tehran, Tehran, Iran , alyiehkazemi@gmail.com
Abstract: (3530 Views)
Background: Lymphadenitis is a common and debilitating complication of breast cancer patients. This study has predicted and classified lymphedema complications. Moreover, identifying effective factors and discovering patterns for faster diagnosis and prevention of this complication is another goal of this study.
Methods: Data from 1113 patients with breast cancer who were referred to Seyed Khandan Lymphedema clinic during 2009 to 2017 were evaluated. Data analysis was performed by using IBM SPSS Modeler software version 18 and CRISP-DM methodology and in the modeling section, logistic regression and decision tree algorithms were used.
Results: Data from 933 patients including 25 variables were entered into the model after pre-processing. Probability of catching of Lymphedema for each patient predicted by logistic regression algorithm and different decision tree algorithms consist of C5.0, Chaid, C&RT, and Quest with the sensitivity of 79.33%, 74.41%, 71.92%, 72.64% and 77.83%, respectively and finally the factors related to Lymphedema were identified. Ratio of the involved lymph node numbers to the removed lymph node numbers, heaviness, type of surgery, stage of disease, age, body mass index, metastasis, number of chemotherapy courses, comorbidity, number of removed lymph nodes respectively.
Conclusion: The results show that C5.0 decision tree algorithm with the highest sensitivity is the best model for predicting Lymphedema. By applying the rules created for a new sample with specific characteristics, it can be predicted that the patient will probably suffer from Lymphedema or not. Considering that BMI is a changeable factor, weight control regimens are recommended for these patients. In addition, it is necessary to pay attention to the patient's heavy feeling in the early stages.
Methods: Data from 1113 patients with breast cancer who were referred to Seyed Khandan Lymphedema clinic during 2009 to 2017 were evaluated. Data analysis was performed by using IBM SPSS Modeler software version 18 and CRISP-DM methodology and in the modeling section, logistic regression and decision tree algorithms were used.
Results: Data from 933 patients including 25 variables were entered into the model after pre-processing. Probability of catching of Lymphedema for each patient predicted by logistic regression algorithm and different decision tree algorithms consist of C5.0, Chaid, C&RT, and Quest with the sensitivity of 79.33%, 74.41%, 71.92%, 72.64% and 77.83%, respectively and finally the factors related to Lymphedema were identified. Ratio of the involved lymph node numbers to the removed lymph node numbers, heaviness, type of surgery, stage of disease, age, body mass index, metastasis, number of chemotherapy courses, comorbidity, number of removed lymph nodes respectively.
Conclusion: The results show that C5.0 decision tree algorithm with the highest sensitivity is the best model for predicting Lymphedema. By applying the rules created for a new sample with specific characteristics, it can be predicted that the patient will probably suffer from Lymphedema or not. Considering that BMI is a changeable factor, weight control regimens are recommended for these patients. In addition, it is necessary to pay attention to the patient's heavy feeling in the early stages.
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
