Development of artificial intelligence for automated measurement of cervical lordosis on lateral radiographs

Abstract

Cervical sagittal alignment is an essential parameter for the evaluation of spine disorders. Manual measurement is time-consuming and burdensome to measurers. Artificial intelligence (AI) in the form of convolutional neural networks has begun to be used to measure x-rays. This study aimed to develop AI for automated measurement of lordosis on lateral cervical x-rays. We included 4546 cervical x-rays from 1674 patients. For all x-rays, the caudal endplates of C2 and C7 were labeled based on consensus among well-experienced spine surgeons, the data for which were used as ground truth. This ground truth was split into training data and test data, and the AI model learned the training data. The absolute error of the AI measurements relative to the ground truth for 4546 x-rays was determined by fivefold cross-validation. Additionally, the absolute error of AI measurements was compared with the error of other 2 surgeons’ measurements on 415 radiographs of 168 randomly selected patients. In fivefold cross-validation, the absolute error of the AI model was 3.3° in the average and 2.2° in the median. For comparison of other surgeons, the mean absolute error for measurement of 168 patients was 3.1° ± 3.4° for the AI model, 3.9° ± 3.4° for Surgeon 1, and 3.8° ± 4.7° for Surgeon 2. The AI model had a significantly smaller error than Surgeon 1 and Surgeon 2 (P = 0.002 and 0.036). This algorithm is available at (https://ykszk.github.io/c2c7demo/). The AI model measured cervical spine alignment with better accuracy than surgeons. AI can assist in routine medical care and can be helpful in research that measures large numbers of images. However, because of the large errors in rare cases such as highly deformed ones, AI may, in principle, be limited to assisting humans.

AI-Driven Segmentation and Automated Analysis of the Whole Sagittal Spine from X-ray Images for Spinopelvic Parameter Evaluation

“…The ICC values were above 0.90 for C2-7 SVA, C2S, C27L, and L1I, indicating excellent agreement between the AI-based measurements and those obtained by specialists. Our study found an mean error of 3.625 for C2-7L measurements, which is comparable to the mean error of 3.3 for cervical lordosis obtained from an AI model trained on 4546 cervical X-rays [17]. The relatively high error may be attributed to including patients with deformities, those with poorly visible C7 endplates due to obesity, and patients with short necks who may have a different shoulder girdle position in the training process.…”

“…In contrast to previous studies, our research utilized over 900 images for training, resulting in an average error of 8.789 • for Lumbar Lordosis Angle (LLA) measurements. Another study that trained on 629 individuals for LLA measurements showed a similar level of mean error at 8.055 • [17]. Our model also demonstrated lower error compared to a model trained on 493 individuals using EOS, which reported a mean absolute error (MAE) of 11.5 • [18].…”

“…In comparison to previous studies, our AI detection and segmentation models achieved high accuracy in detecting and segmenting spinal structures in whole-spine lateral radiographs. For instance, Fujimori et al reported an AP score of 0.6 for detecting vertebral corners in cervical spine radiographs using Faster R-CNN, while our best model achieved an AP score of 0.75 for detecting spinal structures in whole-spine lateral radiographs [17].…”

Cervical Spondylosis Diagnosis Based on Convolutional Neural Network with X-ray Images

“…While analyses based on single image features have produced satisfactory outcomes as evidenced by studies such as Jebri’s [ 27 ] use of a machine learning-based random forest classifier model for detecting intervertebral space narrowing and osteophyte formation, Tamai’s [ 28 ] segmentation of cervical ligament calcification using the EfficientNet-B2 model, Fujimori’s [ 29 ] application of the EfficientNet-B4 model to identify cervical lordosis, and Chen’s [ 30 ] diagnosis of cervical spine scoliosis using the ResNet model. These contributions notwithstanding, CS cannot be adequately represented by single image features alone but rather as a constellation of multiple imaging features.…”