Automated detection of gallbladder stones using a deep learning algorithm on computed tomography scans.

Area of Science:

• Radiology
• Artificial Intelligence
• Medical Imaging

Background:

• Gallstones (cholelithiasis) are common, often diagnosed via ultrasound.
• CT scans can detect gallstones, but automated detection methods are needed.
• Deep learning offers potential for improving diagnostic accuracy in medical imaging.

Purpose of the Study:

• To develop and evaluate a deep learning algorithm for automated gallstone detection on CT scans.
• To assess the diagnostic accuracy, sensitivity, and specificity of the algorithm.
• To compare CT-based gallstone detection with ultrasound and MRI where available.

Main Methods:

• Retrospective single-center study using CT scans from January 2018 to June 2019.
• Deep learning models trained on segmented gallbladders and gallstone bounding-box labels.
• 5-fold cross-validation and evaluation on a test set of 90 scans.
• Analysis included Dice coefficient, sensitivity, specificity, ROC analysis, and comparison with ultrasound/MRI reports.

Main Results:

• The segmentation algorithm achieved a median Dice score of 94.4%.
• The detection pipeline showed 96.7% sensitivity and 83.3% specificity for gallstone detection.
• High sensitivity was observed for hyperdense (97.3%) and large stones (>10 mm, 97.1%).
• CT showed low false-positive (0.6%) and false-negative (4.5%) rates compared to ultrasound/MRI.

Conclusions:

• The developed deep learning algorithm demonstrates high sensitivity for detecting gallstones in CT scans.
• The algorithm is particularly effective for large and conspicuous gallstones.
• This AI tool shows promise for automated gallstone detection in clinical practice.