Individual Radiotherapy Plan Based on Multi-Omics Model Study for Cervical Cancer Patients with Ovarian Preservation

Purpose/Objective(s)

For cervical cancer patients who underwent ovarian preservation radical surgery before radiotherapy, bone marrow suppression occurred with a high probability due to the limitation of the fields angle with the ovarian preservation treatment plan. Therefore, physicists need to evaluate the patient’s comprehensive condition in advance to weigh the advantages and disadvantages to make a clinical treatment plan for these patients.

Materials/Methods

Firstly, we tracked the radiotherapy process of 75 cervical cancer patients who underwent ovarian preservation radical surgery in our hospital from 2022 to 2024. Through bioinformatics analysis, we identified genes related to bone marrow suppression in these patients. Then, 3D-slicer software was used to process CT image data of patients, and physical dose information was calculated through DVHmetrics package from R software. Combined with patient clinical characteristics, multi omics model features were screened related to DVH dose, CT imaging, and genes by machine learning methods (T-test, LASSO, random forest, XGboost, etc.). Finally, these parameters were integrated and scored, and a nomograph was created, which may play an important role in selecting optimal plan for physicist.

Results

Genomics revealed a negative correlation between HnRNPA1 and bone marrow suppression in ovarian conserving patients, patients with low expression of hnRNPA1 were more prone to bone marrow suppression than those with high expression. Five dosomics features were positively correlated with bone marrow suppression, including ovarian volume, distance from ovary to skin surface, distance from ovary to PTV, abdominal circumference of patients at the navel slice, pelvic V70. And 7 imaging omics features were positively correlated with bone marrow suppression, including diagnostics_Mask-interpolated_Minimum, original_shape_MajorAxisLength, etc. A correlation score based on genomics, dosomics, and radiomics was established through machine learning, which achieved an AUC of over 0.67 in predicting bone marrow suppression after radiotherapy.

Conclusion

Our research focused on cervical cancer patients who underwent ovarian preservation radical surgery. Through multi omics mode analysis (genomics+dosomics+radiomics), relevant features were identified, and their relationships were clarified with clinical treatment scoring strategy for rapid evaluating bone marrow suppression possibility. Both physical and biological significance were considered from this score, which can serve as a reference for physicists to provide better individual clinical ovarian conserving radiotherapy plans.