A Treatment Decision Model for Cutaneous Squamous Cell Carcinoma Based on Bayesian Networks

Area of Science:

• Oncology
• Artificial Intelligence in Medicine
• Clinical Decision Support Systems

Background:

• Cutaneous squamous cell carcinoma (cSCC) is a common non-melanoma skin cancer often treated surgically.
• Systemic therapies, including immune checkpoint inhibitors like Cemiplimab, are crucial for advanced or inoperable cSCC.
• Cemiplimab offers new treatment avenues for patients with advanced cSCC unsuitable for conventional therapies.

Purpose of the Study:

• To develop a clinical decision support tool utilizing Bayesian networks (BNs) for optimal cSCC treatment selection.
• To incorporate diverse patient data, including clinical, histological, and genetic factors, into the decision-making model.
• To address challenges of missing or uncertain data in complex treatment planning.

Main Methods:

• Development of a Bayesian network (BN) model for treatment strategy recommendation in cSCC.
• Inclusion of patient-specific data: tumor type, stage, and PD-L1 expression.
• Retrospective validation using data from 66 patients with cSCC or basal cell carcinoma (BCC).

Main Results:

• The BN model demonstrated a high overall accuracy of 95.5% in treatment recommendation.
• Model performance was statistically significant (p < 0.001) when compared to multidisciplinary tumor board decisions.
• The tool effectively managed complex patient data for treatment planning.

Conclusions:

• Bayesian networks (BNs) are effective tools for modeling intricate clinical decision-making processes in oncology.
• The developed BN tool shows promise in enhancing treatment selection for cSCC patients.
• This approach supports personalized medicine by integrating comprehensive patient data.