Sorafenib plus Memory Like Natural Killer Cell Immunochemotherapy in Hepatocellular Carcinoma and Assessment of Local Response using Machine Learning

Purpose: Despite recent advancements in treatment strategies, the complex tumor microenvironment of hepatocellular carcinoma (HCC) continues to hinder therapeutic efficacy and limit overall survival. In this study, we explored the potential of combining sorafenib with memory-like natural killer (ML-NK) cell therapy in a preclinical HCC model and developed machine learning models for noninvasive evaluation of local tumor response.

Materials and Methods: HCC tumors were established in 24 Buffalo rats by implanting 2×10⁶ McA-RH7777 cells. Following diagnosis, animals were randomly allocated to control, sorafenib, ML-NK, and combination therapy groups and received corresponding treatments. Therapeutic responses were monitored over a four-week period using multi-parametric MRI, and tissue samples were collected for histological analysis. The local treatment response was characterized using texture features and machine learning models were developed for assessment of local response.

Results: Combination therapy significantly outperformed ML-NK monotherapy from the second experimental week onwards. Sorafenib demonstrated inferior efficacy in comparison to ML-NK cell immunotherapy, while the combination regimen exhibited the most pronounced inhibition of tumor growth. Additionally, combination therapy resulted in the highest levels of tumor necrosis and apoptosis (p< 0.05). RF models constructed using T1w MRI data achieved a strong correlation coefficient when incorporating four specific features. Conversely, RF models based on T2w MRI data exhibited variable performance that was not directly related to model complexity, likely due to the limitations of T2w MRI in capturing unique characteristics associated with tumor necrosis. The integration of both T1w and T2w MRI data within a single RF model yielded the most robust predictive performance, with a correlation coefficient of 0.85.

Conclusion: Preliminary findings suggest that a therapeutic regimen combining sorafenib with ML-NK cell immunotherapy holds potential as a promising treatment modality for HCC. Furthermore, the statistical modeling demonstrated significant promise for non-invasive assessment of the tumor microenvironment.