5 Key Takeaways

• 1

  SuRImage, an AI model, was evaluated in a multicenter study using smartphone photos of lung specimens for intraoperative assessment.

• 2

  The model demonstrated moderate-to-good discrimination in diagnostic tasks and outperformed frozen section analysis in predefined comparisons.

• 3

  SuRImage achieved high sensitivity and specificity, with 97% sensitivity and 92% specificity for binary identification of invasive lung adenocarcinoma.

• 4

  The study included 1,727 patients and 2,910 surgical resection images, focusing on clinical stage IA lung adenocarcinoma.

• 5

  SuRImage's performance varied across subgroups, particularly showing lower discrimination for grade 1 invasive adenocarcinoma.