Application of Deep Learning to IVC Filter Detection from CT Scans

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Inferior vena cava filters (IVCF) are crucial for preventing blood clots in select patients.
• Delayed removal of temporary IVCF can lead to severe health complications.
• Current tracking methods for IVCF are insufficient, leading to patient and provider unawareness.

Purpose of the Study:

• To develop a deep learning model for segmenting IVCF in CT scan slices.
• To integrate the segmentation model into a prediction algorithm for automated IVCF detection in scans.
• To improve the identification and management of patients with IVCF.

Main Methods:

• A deep learning model was trained to segment IVCF from axial plane CT scan slices.
• The model achieved a Dice score of 0.82 during training on 372 CT slices.
• The segmentation model was integrated with a prediction algorithm for scan-level IVCF detection.

Main Results:

• The deep learning segmentation model achieved a Dice score of 0.82.
• The integrated prediction algorithm demonstrated 92.22% accuracy in detecting IVCF in CT scans.
• The system effectively identifies the presence of IVCF on abdominopelvic CT scans.

Conclusions:

• Deep learning models can accurately segment IVCF from CT scans.
• Automated detection of IVCF using AI can significantly improve patient care and safety.
• This technology offers a promising solution for identifying patients with IVCF who may have forgotten their presence.