Predictive value of clinical-multiparametric MRI radiomics-based nomogram in prognosis of breast cancer patients treated with chemotherapy(PDF)
《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]
- Issue:
- 2024年第9期
- Page:
- 1115-1121
- Research Field:
- 医学影像物理
- Publishing date:
Info
- Title:
- Predictive value of clinical-multiparametric MRI radiomics-based nomogram in prognosis of breast cancer patients treated with chemotherapy
- Author(s):
- HOU Hong; ZHANG Yalin; SUN Huajing; MU Qiang; FANG Kun
- Department of Breast Surgery, Qingdao Central Hospital, Qingdao 266042, China
- Keywords:
- Keywords: breast cancer magnetic resonance imaging radiomics nomogram chemotherapy prognosis
- PACS:
- R737.9；R816.4
- DOI:
- DOI:10.3969/j.issn.1005-202X.2024.09.008
- Abstract:
- Abstract: Objective To investigate the predictive value of a nomogram model constructed with clinical variables and multiparametric magnetic resonance imaging (MRI) radiomics features in prognosis of patients undergoing chemotherapy for breast cancer. Methods A total of 150 breast cancer patients who received chemotherapy in Qingdao Central Hospital from January 2020 to December 2021 were enrolled for the study. All patients underwent breast MRI examination before chemotherapy, and the optimal radiomics features were extracted from the tumor regions of interest on multiparametric MRI image which were manually outlined using segmentation software. After 2-year follow-up, the patients were divided into good prognosis group and poor prognosis group according to whether there were disease recurrence, metastasis and death, and they were divided into training set (n=105) and test set (n=45) in a ratio of 7:3. The clinical data and multiparametric MRI radiomics features were collected, and the factors affecting the prognosis of breast cancer patients receiving chemotherapy were analyzed by univariate and multivariate Logistic regression analyses. Three models, namely clinical model, radiomics model, and clinical-multiparametric MRI radiomics-based nomogram model were constructed, and their predictive value in the prognosis of breast cancer patients treated with chemotherapy was analyzed with receiver operating characteristic curves. Results Out of 150 cases of breast cancer treated with chemotherapy, 52 cases were included in poor prognosis group, with 33 cases of recurrence and/or metastases and 19 cases of deaths. Logistic regression analysis identified TNM stage, estrogen receptor and progesterone receptor as clinical predictors for prognosis of breast cancer patients treated with chemotherapy (P<0.05). There were 69 and 36 patients with good and poor prognoses in training set, and 29 and 16 patients with good and poor prognoses in test set. Through least absolute shrinkage and selection operator algorithm, 16 radiomics features related to the prognosis of breast cancer patients treated with chemotherapy were screened out from TIRM, DWI, and DCE-MRI sequences, and their radiomics scores were calculated. Compared with patients with poor prognosis, those with good prognosis in training set and test set had higher radiomics scores (P<0.05). ROC curve analysis showed that the area under the curve of radiomics model, clinical model, and nomogram model were 0.853, 0.741, 0.923 in training set (P<0.05), and 0.797, 0.749, 0.896 in test set (P<0.05), respectively. Conclusion Based on pre-chemotherapy radiomics scores, TNM stage, estrogen receptor and progesterone receptor are identified as independent predictors for prognosis of breast cancer patients treated with chemotherapy, and the clinical-multiparametric MRI radiomics-based nomogram can effectively improve the predictive efficacy of prognosis in breast cancer patients treated with chemotherapy.
Last Update: 2024-09-26