SPINEX | Validation Study of an AI-Based Automatic Lenke Classification Model for Adolescent Idiopathic Scoliosis: A Literature-Based Analysis

Recently, the Department of Orthopedics of Shanghai Ninth People's Hospital Affiliated to Shanghai Jiao Tong University School of MedicineProfessor Zhao Changqing'steam published a research paper titled "Validation Study of an AI-Based Automatic Lenke Classification Model for Adolescent Idiopathic Scoliosis: A Literature-Based Analysis" in a spinal surgery journal. This study usedthevalidation study of an AI-based automatic Lenke classification model for adolescent idiopathic scoliosis: a literature-based analysis. This study used theSPINEX Medical Technology (Shanghai) Co., Ltd.'sSpineSight image processing software to establish an AI model for measuring the Cobb angle of AIS patients and performing Lenke classification, evaluating its feasibility and practicality in clinical practice.

 

This study developed an artificial intelligence model to automatically measure the Cobb angle of adolescent idiopathic scoliosis (AIS) and perform Lenke classification, aiming to improve measurement efficiency and consistency. Although AI methods (such as CNN) show acceptable consistency with manual measurement results, clinical promotion still faces challenges such as "black-box diagnosis" and medical responsibility. The study emphasizes that enhancing algorithm interpretability and allowing manual verification and adjustment can help increase clinicians' trust in automated measurement and promote its application in practice.

1. Literature Search

Inclusion criteria: (1) clear AIS diagnosis with definite Lenke classification; (2) patient age 10-20 years; (3) clear and complete imaging data.

Exclusion criteria: (1) poor image acquisition or stitching quality; (2) severe pelvic tilt; (3) previous spinal surgery history.

According to the above criteria, 7 papers were finally screened, including 14 cases (1 male and 13 female), aged (14.88 ± 2.55) years. The imaging data of all cases were used as the validation set to evaluate the accuracy of Cobb angle measurement and Lenke classification.

2. Model Construction

2.1. Keypoint Detection

Based on the "SpineSight" image processing software, this study developed an artificial intelligence model for automatically measuring spinal parameters. The core of the model adopts an asymmetric encoder-decoder network structure. The encoder is based on the RepVGG-A1 architecture, which introduces ResNet-like shortcut connections during the training phase to ensure smooth gradient propagation, while being equivalent to an efficient VGG-style convolutional network during the inference phase. The total number of model parameters is about 13M, balancing strong feature extraction capability and real-time inference speed.

The model performs two key tasks: first, generating pixel-wise segmentation masks of vertebral bodies; second, predicting the offsets between each vertebral pixel and its four corner points. During training, the segmentation task uses a binary cross-entropy loss function for supervised learning, while the corner regression task is optimized using a smooth L1 loss function. During the prediction phase, the algorithm combines the segmentation masks and predicted offsets to finally compute precise vertebral corner coordinates.

Performance validation results show that the average radial error (MRE) of the model in locating vertebral corners on standing full-length spinal X-rays is only 1.24 mm and 1.28 mm on frontal and lateral films respectively, and the success detection rate (SDR) within a 2 mm error range is as high as90.46% and 88.93%.In addition, in the pedicle segmentation task, its Dice Similarity Coefficient (DSC) reached 0.83, further confirming that the model has high precision and good clinical practicality in spinal imaging analysis.

2.2. Cobb AngleMeasurement

This study, based onLenke Classification, determined the Cobb angle measurement regions, and used the AI model to identify vertebral corners on full-length frontal and side-bending spinal X-rays, automatically calculating the Cobb angle of each region. Bland-Altman analysis was used to evaluate consistency. The results show excellent consistency between automated and manual measurement (ICC > 0.86), with small mean absolute error and root mean square error for the main curve Cobb angle, and the measurement differences were within the clinically acceptable range of 5°, proving the reliability and practicality of this automated method in classification and measurement.

2.3.Lenke Classification

This study used the AI model to automatically measure the Cobb angle and sagittal parameters, performing Lenke main classification as well as lumbar coronal (A/B/C) and thoracic sagittal (-/N/+) modifier classification for patients. Through validation on external datasets, the model achieved accuracies of 89%, 89% and 96% on the three classifications respectively, confirming that its automatic classification system has high accuracy and provides a reliable tool for clinical practice.

3. Statistical Analysis

This study used SPSS, GraphPad Prism and PyCharm software for statistical analysis. After the Shapiro-Wilk normality test, continuous data were expressed as mean ± standard deviation, and between-group comparisons used the t-test. Reliability was assessed by the intraclass correlation coefficient, absolute agreement was measured by the mean absolute difference and its standard deviation, and the Bland-Altman plot was used to analyze agreement and systematic bias. The classification performance of the AI model was evaluated by confusion matrix indicators (accuracy, precision, recall, etc.). A P value of < 0.05 was considered statistically significant.