Establishing predictive models for malignant and inflammatory pulmonary nodules using clinical data and CT imaging...

Area of Science:

• Pulmonology
• Radiology
• Oncology

Background:

• Pulmonary nodule detection is common but diagnosis is challenging.
• Distinguishing malignant from inflammatory nodules requires accurate methods.

Purpose of the Study:

• To develop and validate models for differentiating malignant from inflammatory solid lung nodules.
• To utilize clinical data and computed tomography (CT) imaging features for improved diagnostic accuracy.

Main Methods:

• A study included 948 patients with pulmonary nodules across four centers.
• Four models were developed based on nodule diameter (≤10mm, >10-≤20mm, >20-≤30mm, all nodules).
• Independent risk factors were identified and models were evaluated using receiver operating characteristic curve analysis (AUC).

Main Results:

• 17 features (2 clinical, 15 imaging) were identified, with lobulation, age, and lesion characteristics being significant.
• The models achieved high performance, with AUCs ranging from 0.861 to 0.943.
• Accuracy, sensitivity, and specificity varied across models but remained high, indicating effective differentiation.

Conclusions:

• Novel subgrouping models effectively distinguish between inflammatory and malignant lung nodules.
• The models utilize a reduced feature set, simplifying the diagnostic process.
• These models can facilitate accurate diagnosis for patients with potentially malignant lung lesions.