Long-Term Survival and CANARY-Based Artificial Intelligence for Multifocal Lung Adenocarcinoma

Area of Science:

• Oncology
• Artificial Intelligence in Medicine
• Pulmonary Medicine

Background:

• Multifocal lung adenocarcinoma (MFLA) presents unique diagnostic and treatment challenges.
• Accurate assessment of tumor invasiveness is crucial for guiding therapeutic strategies in MFLA.

Purpose of the Study:

• To evaluate the efficacy of a computer-aided nodule assessment and risk yield (CANARY)-based artificial intelligence (AI) model in predicting tumor invasiveness in MFLA patients.
• To determine if AI can assist in differentiating between indolent and invasive nodules within MFLA.

Main Methods:

• A prospective registry trial (NCT01946100) enrolled patients with MFLA undergoing surgical resection.
• Retrospective analysis using CANARY-based AI was performed on identified nodules to quantify invasiveness.
• Clinical data, including survival, and pathologic findings were collected and analyzed.

Main Results:

• The majority of nodules at diagnosis (75.8%) were classified as non-invasive by AI-CANARY.
• AI-CANARY scores were significantly higher for surgically removed nodules compared to non-resected nodules (P=.001).
• Five-year and 10-year survival rates for the MFLA cohort were 91% and 73.6%, respectively.

Conclusions:

• Long-term survival for surgically treated MFLA patients (N0, M0) may be comparable to stage I non-small cell lung cancer.
• CANARY-based AI shows potential for stratifying nodules to guide surgical intervention versus observation in MFLA.