Deep Learning-enhanced Opportunistic Osteoporosis Screening in Ultralow-Voltage (80 kV) Chest CT: A Preliminary Study

Area of Science:

• Radiology
• Artificial Intelligence
• Medical Imaging

Background:

• Bone mineral density (BMD) assessment is crucial for osteoporosis diagnosis.
• Current methods often require dedicated scans, limiting opportunistic screening.
• Low-voltage CT scans, commonly used for lung cancer screening, offer potential for incidental BMD measurement.

Purpose of the Study:

• To evaluate the feasibility of deep learning (DL) for automated BMD measurement.
• To utilize ultralow-voltage 80 kV chest CT scans for this purpose.
• To assess the accuracy of DL-based BMD measurements compared to standard quantitative CT (QCT).

Main Methods:

• A cohort of 987 patients undergoing 80 kV chest and 120 kV lumbar CT was analyzed.
• Four convolutional neural networks (CNNs) were developed for segmentation, ROI extraction, and BMD calculation.
• BMD values from 80 kV DL analysis were compared against 120 kV QCT as the reference standard.

Main Results:

• DL-based BMD measurements from 80 kV CT showed high correlation (R²=0.991-0.998) and agreement with 120 kV QCT.
• The DL method demonstrated superior agreement compared to 80 kV QCT.
• Diagnostic performance for osteoporosis and low BMD was excellent, with AUCs ranging from 0.997 to 1.000.

Conclusions:

• Deep learning enables fully automated and accurate BMD calculation from 80 kV chest CT scans.
• This approach facilitates opportunistic osteoporosis screening during lung cancer screening.
• The DL method offers a highly accurate and efficient tool for widespread BMD assessment.