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

Cervical sagitta I alignment is an essential parameter for the evaluation of spine disorders. Manual measurement is time-consuming and burdensome to measurers. Artificial intelligence (Al) in the form of convolutional neural networks has begun to be used to measure x-rays. This study aimed to develop Al 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 Al model learned the training data. The absolute error of the Al measurements relative to the ground truth for 4546 x-rays was determined by fivefold cross-validation. Additionally, the absolute error of Al 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 Al model was 3.3 degrees in the average and 2.2 degrees in the median. For comparison of other surgeons, the mean absolute error for measurement of 168 patients was 3.1 degrees +/- 3.4 degrees for the Al model, 3.9 degrees +/- 3.4 degrees for Surgeon 1, and 3.8 degrees +/- 4.7 degrees for Surgeon 2. The Al model had a significantly smaller error than Surgeon I and Surgeon 2 (P = 0.002 and 0.036). This algorithm is available at (https://ykszk.github.io/c2c7demo/). The Al model measured cervical spine alignment with better accuracy than surgeons. Al 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, Al may, in principle, be limited to assisting humans.