All methods were performed in accordance with current guidelines and regulations. This study was approved by the Ethics Committee of the First Affiliated Hospital of Chongqing Medical University, which waived the requirement of informed consent.

Patient selection

One hundred consecutive patients admitted to our hospital and diagnosed with CFVO were retrospectively reviewed. Inclusion criteria were legible profile plain radiographs and one or more acute or chronic osteoporotic vertebral compression fractures at the thoracolumbar junction (T10-L2). Exclusion criteria were spinal tumours, spinal tuberculosis and spinal pyogenic infections.

Measurement methods

The thoracolumbar kyphosis was defined as the included angle between the superior plateau of T10 and the inferior plateau of L2 (or the included angle between the vertical lines normal to the two aforementioned plateaus)^2.5 (Fig. 1).

Measurement of thoracolumbar kyphosis with a PACS was implemented using the PACS’s built-in system to mark the lines manually and read them automatically.

The smartphone used in the study is a HUAWEI Mate 20 Pro. The resolutions of the three Leica rear cameras are 40 million, 20 million and 8 million pixels. The inclinometer built into this smartphone was used to measure thoracolumbar kyphosis on a single lateral x-ray film. The accuracy of the built-in inclinometer sensor is one degree. The X-ray film was placed on the film viewer, which was located on the wall; therefore, the x-ray film was assumed to be plumb (Fig. 2). First, the vertebral body of T10 was confirmed, the certain edge of the smartphone was aligned with the upper end plate of T10, and the angle displayed on the screen of the smartphone was recorded (Fig. 2A) . Second, the vertebral body of L2 was confirmed, the same edge of the smartphone was aligned to the bottom plate of L2, and the angle displayed on the screen of the smartphone was recorded (Fig. 2B). Third, we calculated and recorded the difference between these two angles, which represented the thoracolumbar kyphosis. Since the smartphone is not transparent, errors may occur when aligning the edge of the smartphone with the line of the end plate; observers should try to avoid this potential error.

The inclinometer built into a smartphone was used to measure thoracolumbar kyphosis on standard lateral radiographs. (A) First, the vertebral body of T10 was confirmed, a certain edge of the smartphone was aligned with the upper end plate of T10, and the angle displayed on the screen of the smartphone was recorded. (B) Second, the vertebral body of L2 was confirmed, the same edge of the smartphone was aligned to the lower end plate of L2, and the angle displayed on the screen of the smartphone was recorded.

Without knowledge of patient information, two observers (Observer A is a treating spine surgeon and Observer B is an orthopedic intern undergoing training in general orthopedics) who had had sufficient training and practice with this new method independently evaluated X-ray films. Each angle obtained by the PACS and the smartphone was measured twice over 2 weeks. The order of the four measurements was as follows: first trial with the PACS (first week), first trial with the smartphone (third week), second trial with the PACS (fifth week) and second trial with the smartphone (seventh week). To reduce possible recall, the order of the radiographs was randomly rearranged using the RAND function of the Excel software.

A stopwatch was used to record the time taken for each measurement. To simulate the actual measurement process used in the clinic, the PACS measurement time started when the X-ray number was entered into the PACS and ended when the angle was recorded; the smartphone measurement time started when the built-in inclinometer function was activated and ended when the angle was recorded.

Excel 2016 was used to record a total of 800 (one hundred thoracolumbar junction × two methods × twice × two observers) thoracolumbar kyphotic angles.

statistical analyzes

Statistical analysis was performed blind. All data was analyzed by SPSS 21.0. The Kolmogorov-Smirnov test was used to check if the values ​​were normally distributed. Bland-Altman analysis was performed in MedCalc software to assess the agreement between mean values ​​by these two methods. One-way ANOVA was used to assess differences between all measurements by different observers and different methods. The reliability of the two measurement methods was analyzed by two-way random intraclass correlation coefficients (ICC). ICC values ​​divide reliability into five levels: poor (0.00–0.20), fair (0.21–0.40), moderate (0.41–0.60), substantial or good (0.61 –0.80) and very good (0.81–1.00). The time required for different measurement methods was also compared by an independent-samples t-test. The significance level was set at 0.05.