Automated pancreas segmentation and volumetry using deep neural network on computed tomography

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Pancreas segmentation is crucial for diagnosing lesions, understanding anatomy, and predicting patient outcomes.
• Convolutional neural networks (CNNs) are widely used for pancreas segmentation, but data limitations and scarcity of large-scale computed tomography (CT) studies hinder deep learning applications.
• Accurate pancreas segmentation is essential for quantitative analysis and clinical decision support in abdominal CT scans.

Purpose of the Study:

• To perform deep learning-based semantic segmentation of the pancreas.
• To evaluate the performance of four distinct 3D segmentation networks on a large dataset.
• To assess the generalizability of these networks through internal and external validation.

Main Methods:

• Deep learning-based semantic segmentation was applied to a dataset of 1006 participants.
• Four individual 3D convolutional neural network architectures were utilized for pancreas segmentation.
• Internal validation was conducted on the 1006-patient cohort, and external validation was performed using the Cancer Imaging Archive (TCIA) pancreas dataset.

Main Results:

• Internal validation yielded mean precision, recall, and Dice similarity coefficients of 0.869, 0.842, and 0.842, respectively.
• External validation with the TCIA dataset resulted in mean precision, recall, and Dice similarity coefficients of 0.779, 0.749, and 0.735.
• The study demonstrated the effectiveness of deep learning models in achieving accurate pancreas segmentation across different datasets.

Conclusions:

• Deep learning models show significant potential for automated pancreas segmentation in abdominal CT.
• The developed systems can provide accurate segmentation and quantitative information, aiding clinical decisions.
• Further research into generalized deep learning systems is warranted to enhance their clinical utility.