Differentiating non-lactating mastitis and malignant breast tumors by deep-learning based AI automatic classification...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Oncology

Background:

• Accurate differentiation between non-lactating mastitis (NLM) and malignant breast tumors is crucial for effective patient management.
• Traditional ultrasound interpretation relies heavily on sonographer experience, leading to variable diagnostic accuracy.

Purpose of the Study:

• To evaluate the diagnostic performance of a deep learning-based artificial intelligence (AI) automatic classification system.
• To compare the AI system's accuracy against traditional ultrasound interpretations by sonographers with varying experience levels.
• To assess the AI system's utility in differentiating NLM from malignant breast tumors.

Main Methods:

• A retrospective study involving 707 patients with breast lesions (475 malignant tumors, 232 NLM) across three medical centers.
• Ultrasound images were interpreted independently by intermediate and senior sonographers, and by a deep learning-based AI system.
• Kappa test was used to evaluate consistency between interpretations and postoperative pathological diagnoses.

Main Results:

• The deep learning AI system achieved an accuracy of 83.00%, sensitivity of 87.20%, and specificity of 85.33% (AUC 92.6).
• Senior sonographers demonstrated high accuracy (87.35%), sensitivity (86.27%), and specificity (87.89%), with high consistency (kappa=0.72) with pathological results.
• Intermediate sonographers had lower accuracy (76.92%), sensitivity (84.71%), and specificity (73.95%), with moderate consistency (kappa=0.53).

Conclusions:

• The deep learning-based AI automatic classification system demonstrates high accuracy, sensitivity, and specificity.
• The AI system shows high consistency with pathological diagnoses, comparable to senior sonographers.
• AI is a promising auxiliary tool for the differential diagnosis of NLM and malignant breast tumors.