A Novel Tumor Progression Prediction Method for Multimode Ablation Treatment.

Area of Science:

• Oncology
• Medical Imaging
• Data Science

Background:

• Multimode ablation (radio-frequency heating post-pre-freezing) shows promise for liver cancer treatment, enhancing therapeutic effects and anti-tumor immunity.
• Post-treatment, ablated lesions remain, necessitating accurate methods for immediate outcome assessment and long-term disease monitoring.

Purpose of the Study:

• To develop a novel method for predicting tumor progression after multimode ablation for liver cancer.
• To enable simultaneous postoperative evaluation and prognosis analysis.

Main Methods:

• Leveraging intraoperative therapeutic information from thermal dose distribution.
• Developing a survival analysis framework using features from clinical, preoperative, intraoperative, and postoperative data.
• Employing random survival forest for feature selection and deep neural networks for survival prediction.

Main Results:

• The proposed method achieved a C-index of 0.855±0.090, outperforming existing state-of-the-art survival analysis techniques.
• Thermal dose information was a significant predictor, accounting for 21.7% of the overall feature importance.

Conclusions:

• The developed data-driven methods are effective for tumor progression prediction in multimode ablation therapy.
• This approach can benefit personalized medicine and streamline patient follow-up processes.