Abstract:

Objective To analyze the predictive value of artificial intelligence （AI） quantitative parameters combined with 256?slice spiral CT scans for assessing the invasiveness of lung ground?glass nodules （GGNs）. Methods This study included 98 GGN patients diagnosed by postoperative pathology at the hospital from May 2021 to July 2024. Preoperative assessments involved AI quantitative parameters and 256?slice spiral CT scans. Patients were categorized into non?invasive （AAH， AIS， MIA） and invasive （IAC） groups based on pathology. AI parameters and CT scan results were compared to analyze factors influencing invasiveness and their predictive value. Results Among the 98 GGN patients， there were 29 AAH cases， 22 AIS， 19 MIA， and 28 IAC. The invasive group had higher average CT values， nodule long?axis diameter， maximum area， presence of air bronchogram， vascular clustering signs， and irregular shapes compared to the non?invasive group （P < 0.05）. Binary logistic regression identified these six features （air bronchogram， vascular clustering， irregular shape， average CT value， nodule long diameter， and maximum area） as significant factors affecting GGN invasiveness （P < 0.05）. ROC curve analysis showed that the combined detection of these parameters had higher sensitivity and specificity than single tests， with an AUC of 0.907， indicating a high predictive value for assessing GGN invasiveness. Conclusion The combination of AI quantitative parameters and 256?slice spiral CT scanning effectively predicts the invasiveness of GGN， providing significant clinical guidance for preoperative evaluation.

Key words: artificial intelligence quantitative parameters, 256-slice spiral CT scans, lung ground-glass nodules, invasiveness