Diagnosis of Acute Appendicitis with Machine Learning-Based Computer Tomography: Diagnostic Reliability and Role in...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Surgical Emergencies

Background:

• Acute appendicitis (AA) is a prevalent surgical emergency requiring prompt diagnosis to mitigate severe outcomes.
• Current diagnostic methods for AA include physical examination, lab tests, and imaging like ultrasonography and computed tomography (CT).

Purpose of the Study:

• To assess the efficacy of artificial intelligence (AI) in analyzing CT images for the early detection of AA.
• To explore AI's potential in preventing AA-related complications through improved diagnostic accuracy.

Main Methods:

• Analysis of 1200 CT images from patients diagnosed with AA between January 2019 and June 2023.
• Evaluation of four distinct AI models, with performance metrics determined via confusion matrix and receiver operating characteristic analysis.

Main Results:

• MobileNet v2 achieved the highest accuracy (0.7908) and precision (0.8203), while Inception v3 yielded the highest F-score (0.7928).
• MobileNet v2 demonstrated a significant area under the curve (AUC) of 0.8767 in receiver operating characteristic analysis.
• The study reported a maximum sensitivity of 77% and specificity of 86% for AI models in AA diagnosis.

Conclusions:

• AI-powered analysis of CT images can significantly enhance the accuracy of acute appendicitis diagnoses.
• Integrating AI systems with CT imaging offers a valuable tool for early AA detection and management.
• The findings support the expanding role of AI in clinical practice, particularly in diagnosing surgical emergencies.