CT image segmentation of bone for medical additive manufacturing using a convolutional neural network

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Additive Manufacturing

Background:

• Medical additive manufacturing (AM) relies heavily on image segmentation, a process that is typically time-consuming and labor-intensive.
• Accurate bone segmentation in computed tomography (CT) scans is crucial for creating patient-specific implants.

Purpose of the Study:

• To develop and train a convolutional neural network (CNN) for automated bone segmentation in CT scans.
• To evaluate the accuracy and efficiency of the CNN in segmenting skull bone for AM applications.

Main Methods:

• A CNN was trained using CT scans from multiple scanners and validated against "gold standard" Standard Tessellation Language (STL) models derived from patient data.
• Segmentation accuracy was quantified using the Dice similarity coefficient and surface deviation analysis.

Main Results:

• The CNN achieved a high Dice similarity coefficient of 0.92 ± 0.04, indicating excellent overlap with gold standard segmentations.
• Mean surface deviations of CNN-derived models were within clinically acceptable ranges (-0.19 ± 0.86 mm to 1.22 ± 1.75 mm).
• Performance was consistent across CT scans from different scanners.

Conclusions:

• The automated CNN effectively segments skull bone in CT scans with high accuracy.
• This AI-driven approach significantly reduces the time and effort required for image segmentation, thereby increasing the accessibility of patient-specific AM constructs.