Development of one novel phantom-less quantitative computed tomography (QCT) technique and its application in diagnosis of osteoporosis

Abstract

Osteoporosis is one of the most common degenerative diseases worldwide. Osteoporosis leads to the progressive weakening of bone and is one major cause of fractures. With aging populations in many countries, the osteoporosis diagnosis issue in the older population needs to be urgently addressed. Bone mineral density (BMD) is widely used to predict osteoporotic fractures. However, due to limitations of clinical imaging technologies, few studies have focused on BMD measurements via routine computed tomography (CT) scans. Moreover, CT data with different scanning parameters greatly influence the accuracy of BMD measurements. Therefore, retrospective large-scale CT scans for automatic BMD testing using internal calibration methods has not yet been investigated in previous studies.

This study aimed to develop and validate the use of one automatic BMD measurement system based on routine CT scans, which can be applied in retrospective and respective data to improve the osteoporosis screening rate. Specifically, the objectives are 1) to develop one novel automatic phantom-less quantitative computed tomography (PL-QCT) method for accurate volumetric BMD measurements from routine CT scans, 2) to validate the accuracy and precision of this novel method compared with phantom-based QCT via clinical validation projects, and 3) to investigate the application of this system for osteoporosis screening in low-dose chest CT (LDCT) scans originally obtained for COVID-19 or cancer screening and compare it with Dual Energy X-ray Absorptiometry (DXA), 4) to explore the prediction capability of the system for osteoporotic vertebral compression fractures (OVCF).

The results revealed the reproducible performance of the automatic PL-QCT system in terms of the precision test (CV<1%). LDCT data obtained for lung cancer or COVID-19 screening of 649 patients were utilized for BMD calibration by the PL-QCT system. In the ROC results, the AUC index remained above 0.75, indicating the novel system could predict osteopenia and osteoporosis with high accuracy according to the clinical golden standard. The multi-center clinical validation showed this novel system was highly correlated with the phantom-based QCT method in spine BMD measurements (bias < 1 mg/cc, r = 0.98). Additionally, it was found that many patients with osteoporosis were not properly diagnosed by DXA due to osteophytes and vascular calcification. QCT system with three dimensional BMD results could partly solve the problem by locating suitable region of interest (ROI). Furthermore, the OVCF prediction capability of the automatic PL-QCT system has been validated with high accuracy in this study.

In conclusion, this study developed one automatic PL-QCT system by the combination of internal calibration methods and image processing algorithms. Based on a series of clinical validation projects, this novel system was demonstrated to assess BMD with high accuracy and precision compared with traditional internal calibration approaches. Moreover, it also has great potential for large population osteoporosis screening and OVCF prediction through its application in all types of routine CT scans collected retrospectively and prospectively in future.