Objective: the purpose of this study was to develop an MRI image-based nomogram for predicting patient survival in the diagnosis of osteosarcoma. Methods: 150 patients with 102 pathologically confirmed osteosarcoma diagnosed from January 2008 to March 2011 were collected, all of whom had received preoperative MRI examination. According to the order of diagnosis, patients were divided into two groups: training cohort and independent verification cohort. 102 cases from the training group were used to construct the prediction model. The independent validation group of 48 cases was used to test the predictive power of the model. Patients with survival ≥5 years after treatment were classified as survival group, while patients who died within 5 years after surgery were classified as non-survival group. Clinical data were collected including age, sex, tumor anatomy, tumor surgical staging (Enneking), and pathologic fractures. Each sequence image archived in the PACS system is used and the original image is exported in DICOM format. The minimum follow-up was 5 years. Itk-snap software was used for manual ROI segmentation. The best survival related imaging features were then extracted, including tumor strength, shape and size, texture, and petit's sign. Then, LASSO logistic regression was used to construct the rad-score for survival prediction of osteosarcoma patients, and combined with the multivariate analysis of clinical data, a normogram personalized prediction model was constructed. Results: 1. The scores of patients in the survival group were generally higher than those in the non-survival group. Significant differences in radiology scores (p <.0001) were observed between the survival and non-survival groups in the training and validation groups. The AUC of the training group was 0.79. The AUC of the validation group was 0.76. 2. The imaging omics nomogram showed better calibration and classification ability than the clinical model, with an AUC of 0.86, compared with 0.79 in the training group and 0.84 versus 0.73 in the validation group. Decision curve analysis proves the clinical usefulness of radiomics nomogram. 3. Significant differences were observed between the survival curves of the survival group and the non-survival group predicted by the normogram (p value b<.05; Log-rank test). Conclusion: imaging omics nomogram can help clinicians customize the appropriate treatment plan.

Acknowledgements

First and foremost, I would like to show my deepest gratitude to mysupervisor, Dr. Wang Meiyun, a respectable, responsible and resourceful scholar,who has provided me with valuable guidance in every stage of the writing of thisthesis. Without her enlightening instruction, impressive kindness and patience, Icould not have completed my thesis. Her keen and vigorous academic observationenlightens me not only in this thesis but also in my future study. My sincere appreciation also goes to seniors, who participated this study with great cooperation.Last but not least, I' d like to thank all my friends, especially my three lovely roommates, for their encouragement and support.