# Radiographic signature in apical periodontitis improves prediction of apical lesion healing through survival prediction model

## TL;DR

A new method using radiographic features improves the prediction of healing time for apical periodontitis lesions.

## Contribution

A radiomics-based survival model using lesion boundary features significantly improves healing prediction in apical periodontitis.

## Key findings

- The FA method improved survival model performance with an AUC increase to 0.884–0.905 at 12–18 months.
- The Delong test confirmed significant improvement in model performance (p < 0.05).
- Clinical benefit was higher with the FA-derived radiomics score in predicting healing speed.

## Abstract

This retrospective study aimed to evaluate the effectiveness of radiographic signatures of apical periodontitis (AP), particularly lesion boundary features, in predicting lesion healing periods using survival analysis. A total of 254 AP cases with apical lesions were included. Canny edge detection and fragment analysis (FA) were used to define the regions of interest (ROI) S1-S4 on radiographs. Radiographic signatures were extracted, and a radiomics score (rad-score) was developed using the least absolute shrinkage and selection operator (LASSO) Cox regression. Preliminary validation was performed using Kaplan-Meier survival analysis. Survival models were fitted, and model performance was evaluated. Clinical benefit was assessed through decision curve analysis. The results showed that radiographic signatures of the lesion boundary identified via the FA method significantly improved the performance of the survival model (Delong test; p < 0.05), with optimization of the calibration curve and an increase in the area under the curve (AUC) from 0.566–0.619 (reference model) to 0.884–0.905 at 12, 15, and 18 months. These findings were maintained in a small external validation cohort. The clinical benefit was also greater when using the rad-score derived via the FA method. In summary, the FA method proved to be an effective tool for quantifying the apical lesion boundary and predicting the healing speed using a survival model.