AI predicts Fishman skeletal maturity from hand-wrist X-rays

Title

Prediction of Fishman’s skeletal maturity indicators using artificial intelligence

Authors

Harim Kim; CheolSoon Kim; JiMin Lee; Jae Joon Lee; Jiyeon Lee; JungSuk Kim; SungHwan Choi

Summary

This study evaluates an artificial intelligence–based automated system for predicting Fishman’s Skeletal Maturity Indicators (SMI) using handwrist radiographs. The proposed hybrid system integrates GreulichPyle, TannerWhitehouse 3, and Fishman’s SMI methods. The workflow includes automated ROI detection, regionwise classification, and SMI stage mapping. The model achieved clinically reliable performance, with an overall accuracy of 0.772 and MAE of 0.27 SMI stages.

Key Words

Artificial intelligence; Skeletal maturity; Fishman SMI; Handwrist radiographs; Orthodontics

Extracted Data

Year: 2023

Modality: Handwrist radiographs

Dataset: 3,304 hand-wrist radiographs

Dataset Split: Training / internal validation: 2,593 ; External testing: 711

Network Architecture: Detection Transformer; RetinaNet; Swin Transformer

Metrics: Accuracy; Mean Absolute Error (MAE); Root Mean Square Error (RMSE); Sensitivity; Specificity

AP – Professional Qty: 3 (+1 in case of disagreement)

AP – Supervisor Presence: No information

AP – Experience Level: No information

AP – Expertise Area: Orthodontist

AP – Tool or System: No information

ML Task: Classification

Project Objective: To evaluate the performance of an AI-based automated assessment system for Fishman's SMI prediction to improve clinical efficiency and reproducibility

Clinical Relevance

The proposed AI system enhances efficiency and reproducibility in skeletal maturity assessment, offering clinically reliable SMI predictions that support orthodontic treatment planning and growthrelated decisionmaking.