Clinical Scorecard: AI Shows Promise for Intraoperative Lung Adenocarcinoma Assessment

At a Glance

Key Highlights

• SuRImage demonstrated moderate-to-good discrimination in diagnostic and grading tasks.
• Outperformed frozen section analysis in predefined comparisons.
• Achieved 97% sensitivity and 92% specificity for binary identification of invasive lung adenocarcinoma.
• Average accuracy of 89% for multiclass grading compared to 65% for frozen section analysis.
• Maintained robust diagnostic performance among patients with ambiguous frozen section findings.

Guideline-Based Recommendations

Diagnosis

• Use SuRImage for intraoperative identification of invasive lung adenocarcinoma vs preinvasive lesions.

Management

• Intraoperative assessment can guide the extent of resection and lymph node evaluation.

Monitoring & Follow-up

• Monitor diagnostic accuracy and grading performance in clinical settings.

Risks

• Model not designed for all lung nodules or other lung cancer histologies.

Patient & Prescribing Data

1,727 patients with indeterminate lung nodules.

Real-world feasibility remains uncertain due to exclusion of patients with poor image quality.

Clinical Best Practices

Related Resources & Content

• The Lancet Digital Health