Predicting Invasiveness in Lepidic Pattern Adenocarcinoma of Lung: Analysis of Visual Semantic and Radiomic Features

Area of Science:

• Radiology
• Oncology
• Pulmonary Medicine

Background:

• Accurate differentiation of lung adenocarcinoma subtypes is crucial for treatment planning.
• Adenocarcinoma in situ (AIS) and minimally invasive adenocarcinoma (MIA) require different management than invasive lepidic predominant adenocarcinoma (iLPA).
• CT-guided biopsy can sometimes yield results that underestimate tumor invasiveness.

Purpose of the Study:

• To evaluate the utility of visual semantic and computer-aided detection (CAD)-based texture features on CT scans for differentiating iLPA from AIS/MIA.
• To identify predictive features that can help assess invasiveness in lung nodules initially diagnosed as AIS or MIA.

Main Methods:

• Retrospective analysis of 33 patients with CT-guided biopsy results of AIS or MIA who later underwent resection.
• Assessment of visual semantic features and 95 CAD-based quantitative texture variables from pre-biopsy CT scans.
• Feature selection using LASSO or forward selection to identify predictive markers for invasiveness.

Main Results:

• Of 33 patients, 24 (72.7%) had iLPA upon resection, while 9 (27.3%) had AIS/MIA.
• Visual CT features included part-solid (63.6%), pure ground glass (15.2%), and solid (21.2%) nodules.
• While LASSO-selected features were not significant, 'volume' was significant via backward selection. LASSO also identified 'tumor_Perc95', 'nodule surround', and 'small cyst-like spaces' as potentially relevant.

Conclusions:

• Initial biopsy results of noninvasive lepidic predominant adenocarcinoma may underestimate true invasiveness.
• Certain semantic CT features, such as absent septal stretching in noninvasive cases and solid consistency in invasive cases, show promise for predicting invasiveness.
• Further investigation into these CT-based features is warranted to improve diagnostic accuracy.