Evaluation of deep learning-based quantitative computed tomography for opportunistic osteoporosis screening

Abstract

To evaluate diagnostic efficacy of deep learning (DL)-based automated bone mineral density (BMD) measurement for opportunistic screening of osteoporosis with routine computed tomography (CT) scans. A DL-based automated quantitative computed tomography (DL-QCT) solution was evaluated with 112 routine clinical CT scans from 84 patients who underwent either chest (N:39), lumbar spine (N:34), or abdominal CT (N:39) scan. The automated BMD measurements (DL-BMD) on L1 and L2 vertebral bodies from DL-QCT were validated with manual BMD (m-BMD) measurement from conventional asynchronous QCT using Pearson's correlation and intraclass correlation. Receiver operating characteristic curve (ROC) analysis identified the diagnostic ability of DL-BMD for low BMD and osteoporosis, determined by dual-energy X-ray absorptiometry (DXA) and m-BMD. Excellent concordance were seen between m-BMD and DL-BMD in total CT scans (r = 0.961/0.979). The ROC-derived AUC of DL-BMD compared to that of central DXA for the low-BMD and osteoporosis patients was 0.847 and 0.770 respectively. The sensitivity, specificity, and accuracy of DL-BMD compared to central DXA for low BMD were 75.0%, 75.0%, and 75.0%, respectively, and those for osteoporosis were 68.0%, 80.5%, and 77.7%. The AUC of DL-BMD compared to the m-BMD for low BMD and osteoporosis diagnosis were 0.990 and 0.943, respectively. The sensitivity, specificity, and accuracy of DL-BMD compared to m-BMD for low BMD were 95.5%, 93.5%, and 94.6%, and those for osteoporosis were 88.2%, 94.5%, and 92.9%, respectively. DL-BMD exhibited excellent agreement with m-BMD on L1 and L2 vertebrae in the various routine clinical CT scans and had comparable diagnostic performance for detecting the low-BMD and osteoporosis on conventional QCT.

Figure 1

ROC curves of m-BMD and DL-BMD for diagnosis of osteoporosis and low BMD. (A, B). The AUC values were calculated for diagnosing osteoporosis (A) and low BMD (B) based on the central DXA. m-BMD = manual BMD, DL-BMD = deep learning-based automated bone mineral density, AUC = area under the receiver operating characteristic curve, Low BMD = osteoporosis or osteopenia, DXA = dual-energy X-ray.

Figure 2

ROC curves of DL-BMD for diagnosing osteoporosis and low BMD. (A–C) The AUC values were calculated for diagnosing osteoporosis (A) and low BMD (B) based on m-BMD values. DL-BMD = deep learning-based automated bone mineral density, m-BMD = manual BMD, AUC = area under the receiver operating characteristic curve (AUC), Low BMD = osteoporosis or osteopenia a.

Figure 3

Images obtained from a 68-year-old man who was diagnosed via DXA to have normal CT scans (A–C). (A) T-scores for lumbar DXA, femoral neck, and total hip were 1.4, − 0.5, and − 0.5, respectively. The trabecular BMD of L1-L2 was 77.125 mg/cm³, showing a diagnosis of osteoporosis according to the American College of Radiology (ACR) guidelines. A lateral lumbar spine radiograph (B) and an axial CT image (C) depicted severe osteophytes and end plate sclerosis of the lumbar vertebrae. Abdominal aortic calcifications also were noted.

Figure 4

Flow diagram of the study patients. QCT = quantitative computed tomography, DXA = dual-energy X-ray absorptiometry.

Figure 5

Manually measuring v-BMD on abdominal CT using asynchronous QCT. v-BMD = volumetric bone mineral density, QCT = quantitative computed tomography.

Figure 6

Overview of the 2D U-Net architecture with an input image matrix size of 512 × 512 and multiple layers. Arrows of different colors indicate different operations. Conv = Convolutional layer, ReLU = Rectified Linear Unit.

Figure 7

Process of DL-BMD measurement on chest CT. (A–C) The selected slice shows the raw data (A), area of basivertebral vein (B), and the ROI drawn (C) excluding the basivertebral vein (B) area. DL-BMD = deep learning-based automated bone mineral density, ROI = region of interest.