요추 X-ray의 딥러닝 분석을 통한 추나 요추 변위 진단 프로토콜

Diagnostic Protocol for Lumbar Malposition in Chuna Manual Therapy via Deep Learning Analysis of Lumbar X-rays

Objectives This study aims to establish a deep learning-based diagnostic protocol for identifying lumbar malposition on lumbar spine X-ray images in Chuna manual therapy. The goal is to replace subjective palpation-dependent diagnosis with ob-jective, reproducible radiographic assessment. Methods Radiographic biomarkers—including vertebral corner points, pedicle mar-gins, and lumbosacral alignment indicators—will be manually annotated by experts and used to train convolutional neural network models. The dataset will be divided by patient into training, validation, and test sets. Landmark detection accuracy and mal-position classification will be evaluated using mean absolute error, intraclass correla-tion coefficients, accuracy, F1-score, Area Under the Curve(AUC), and Cohen’s kappa. Results The proposed model is expected to detect anatomical landmarks with high agreement to expert annotations and classify malposition types with clinically ac-ceptable performance. This protocol may provide a standardized, quantitative frame-work for lumbar alignment assessment in Chuna manual therapy. Conclusions This study established a deep learning-based diagnostic protocol for the automated detection of radiographic biomarkers and the quantification of lumbar displacements in Chuna manual therapy. By providing an objective and reproducible evaluation system, this protocol is expected to enhance diagnostic reliability and sup-port evidence-based clinical practice in Korean medicine.