Artificial intelligence model based on CT imaging for predicting infected upper urinary tract calculi

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Urology

Background:

• Infected upper urinary tract calculi pose diagnostic challenges.
• Accurate preoperative identification is crucial for effective treatment planning.

Purpose of the Study:

• To develop and evaluate an AI model using CT imaging for predicting infected upper urinary tract calculi.
• To assess the model's efficacy in preoperative diagnosis.

Main Methods:

• A RetinaNet model was trained on CT images from 465 patients with urinary tract calculi.
• Infectious stones were defined by >25% struvite or carbonate apatite content.
• Model performance was evaluated using metrics like Mean Average Precision (mAP) and accuracy on a separate test set.

Main Results:

• The AI model achieved high specificity (93.09%) in predicting infectious stones.
• Moderate sensitivity (72.78%) was observed, indicating limitations for ruling out infections.
• The model demonstrated significant consistency with infrared spectroscopy (kappa=0.679).

Conclusions:

• The AI model demonstrates high specificity for predicting infectious upper urinary tract calculi, valuable for preoperative identification.
• Its moderate sensitivity suggests it should be used as a complementary tool alongside laboratory tests.
• The model offers auxiliary guidance for treatment and surgical decisions in patients with urinary tract calculi.