The following is a summary of “AI-driven Characterization of Solid Pulmonary Nodules on CT Imaging for Enhanced Malignancy Prediction in Small-sized Lung Adenocarcinoma,” published in the April 2024 issue of Pulmonology by Kudo et al.
The accurate differentiation of solid nodules from those with ground-glass opacities in lung cancer, particularly in tumors ≤2 cm, poses a significant diagnostic challenge. Human interpretation of these nodules exhibits considerable variability among observers, underscoring the need for an objective and reliable diagnostic tool. This study aims to harness artificial intelligence (AI) to automatically analyze such tumors and develop prospective AI systems capable of independently distinguishing highly malignant nodules.
A retrospective analysis was conducted on 246 patients diagnosed with negative clinical lymph node metastases (cN0) using positron emission tomography-computed tomography (PET/CT) imaging, who subsequently underwent surgical resection for lung adenocarcinoma. AI was employed to detect tumors ≤2 cm in size. By classifying these nodules as either solid (AI_solid) or non-solid (non-AI_solid) based on confidence scores, the study sought to establish correlations between AI determinations and pathological findings, thereby enhancing the precision of preoperative assessments.
Solid nodules identified by AI with a confidence score ≥0.87 exhibited significantly higher solid component volumes and proportions in patients categorized as AI_solid compared to those classified as non-AI_solid, with no discernible differences in overall diameter or total volume of the tumors. Among patients classified as AI_solid, 16% demonstrated lymph node metastasis, and a noteworthy 94% harbored invasive adenocarcinoma. Additionally, 44% experienced upstaging postoperatively. These AI_solid nodules were indicative of high-grade malignancies.
In the context of small-sized lung cancer diagnosed as cN0, AI demonstrates the capability to automatically identify tumors as solid nodules ≤2 cm and assess their malignancy preoperatively. The AI classification holds the potential to inform the necessity for lymph node assessment in sub-lobar resections, reflecting the metastatic potential of the nodules.
Source: sciencedirect.com/science/article/abs/pii/S152573042400069X
