Automated design prediction for definitive obturator prostheses: A case-based reasoning study

This study aims to evaluate the effectiveness of a case-based reasoning (CBR) system in predicting the design of definitive obturator prostheses for maxillectomy patients.

Data from 209 maxillectomy cases, including extraoral images of obturator prostheses and occlusal images of maxillectomy defects, were collected from Institute of Science Tokyo Hospital. These cases were organized into a structured database using Python's pandas library. The CBR system was designed to match new cases with similar historical cases based on specific attributes such as aramany class, abutment details, defect extension, and oronasal connection size. The system's performance was evaluated by clinicians who assessed the accuracy of prosthesis designs generated for 33 test cases.

A correlation analysis demonstrated a significant positive relationship (ρ = 0.84, p < 0.0001) between the CBR system's confidence scores and the number of correct prosthesis designs identified by clinicians. The median precision at five cases was 0.8, indicating that the system effectively retrieved relevant designs for new cases.

The study shows that the developed CBR system effectively predicts the design of obturator prostheses for maxillectomy patients. Clinically, the system is expected to reduce clinician workload, simplify the design process, and enhance patient engagement by providing prompt insights into their final prosthetic design.