METHODS TO IMPROVE SURGICAL EFFICIENCY BY USING ARTIFICIAL INTELLIGENCE IN THE MANAGEMENT OF PATIENTS WITH MAXILLARY SINUS CYSTS

Abstract

Maxillary sinus cystic lesions include mucous retention cysts, pseudocysts, postoperative maxillary cysts, mucoceles, and odontogenic cystic lesions extending into the sinus. Mucous retention cysts and pseudocysts are usually benign incidental findings and may be seen in up to 13% of adults, but differential diagnosis remains important because CT or CBCT is often needed to distinguish them from lesions that require surgery. In recent years, artificial intelligence has emerged as a promising adjunct for detecting maxillary sinus pathology, segmenting lesion volume, and supporting surgical planning. Systematic reviews of CT and CBCT studies report diagnostic accuracy roughly in the 85% to 97% range, sensitivity from 87% to 100%, and specificity from 87.2% to 99.7%, while automated maxillary sinus segmentation models based on nnU-Net v2 have achieved F1-scores around 0.96, sensitivity 0.96, Dice coefficients about 0.96, and IoU near 0.93. The aim of this modeled study was to evaluate whether an AI-assisted workflow could improve surgical efficiency and perioperative decision-making in 140 patients with maxillary sinus cysts managed at the TDTU ENT Department. The modeled protocol combined standard clinical assessment, nasal endoscopy, CT or CBCT imaging, AI-based lesion triage, automatic 3D segmentation, and a decision-support layer for choosing observation, standard endoscopic surgery, or extended endoscopic access. Synthetic outcomes suggested that AI support improved diagnostic concordance, shortened time to final operative planning, reduced additional imaging, improved selection of the endoscopic corridor, and modestly reduced recurrence and operative time in anatomically complex lesions.