Prediction model of radiotherapy outcome for Ocular Adnexal Lymphoma using informative features selected by chemometric algorithms
Min Zhou,Jiaqi Wang,Jiahao Shi,Guangtao Zhai,Xiaowen Zhou,Lulu Ye,Lunhao Li,Menghan Hu,Yixiong Zhou
DOI: https://doi.org/10.1016/j.compbiomed.2024.108067
IF: 7.7
2024-02-02
Computers in Biology and Medicine
Abstract:Background: Ocular Adnexal Lymphoma (OAL) is a non-Hodgkin's lymphoma that most often appears in the tissues near the eye, and radiotherapy is the currently preferred treatment . There has been a controversy regarding the prognostic factors for systemic failure of OAL radiotherapy, the thorough evaluation prior to receiving radiotherapy is highly recommended to better the patient's prognosis and minimize the likelihood of any adverse effects. Purpose: To investigate the risk factors that contribute to incomplete remission in OAL radiotherapy and to establish a hybrid model for predicting the radiotherapy outcomes in OAL patients. Methods: A retrospective chart review was performed for 87 consecutive patients with OAL who received radiotherapy between Feb 2011 and August 2022 in our center. Seven image features, derived from MRI sequences, were integrated with 122 clinical features to form comprehensive patient feature sets. Chemometric algorithms were then employed to distill highly informative features from these sets. Based on these refined features, SVM and XGBoost classifiers were performed to classify the effect of radiotherapy. Results: The clinical records of from 87 OAL patients (median age: 60 months, IQR: 52–68 months; 62.1% male) treated with radiotherapy were reviewed. Analysis of Lasso (AUC = 0.75, 95% CI: 0.72–0.77) and Random Forest (AUC = 0.67, 95% CI: 0.62–0.70) algorithms revealed four potential features, resulting in an intersection AUC of 0.80 (95% CI: 0.75–0.82). Logistic Regression (AUC = 0.75, 95% CI: 0.72–0.77) identified two features. Furthermore, the integration of chemometric methods such as CARS (AUC = 0.66, 95% CI: 0.62–0.72), UVE (AUC = 0.71, 95% CI: 0.66–0.75), and GA (AUC = 0.65, 95% CI: 0.60–0.69) highlighted six features in total, with an intersection AUC of 0.82 (95% CI: 0.78–0.83). These features included enophthalmos , diplopia , tenderness, elevated ALT count, HBsAg positivity, and CD43 positivity in immunohistochemical tests. Conclusion: The findings suggest the effectiveness of chemometric algorithms in pinpointing OAL risk factors, and the prediction model we proposed shows promise in helping clinicians identify OAL patients likely to achieve complete remission via radiotherapy. Notably, patients with a history of exophthalmos, diplopia, tenderness, elevated ALT levels, HBsAg positivity, and CD43 positivity are less likely to attain complete remission after radiotherapy. These insights offer more targeted management strategies for OAL patients. The developed model is accessible online at: https://lzz.testop.top/ .
engineering, biomedical,computer science, interdisciplinary applications,mathematical & computational biology,biology