Differentiating Small-Cell Lung Cancer From Non-Small-Cell Lung Cancer Brain Metastases Based on MRI Using...

Area of Science:

• Radiology
• Oncology
• Artificial Intelligence

Background:

• Distinguishing small-cell lung cancer (SCLC) from non-small-cell lung cancer (NSCLC) brain metastases is critical due to differing clinical behaviors.
• Accurate non-invasive classification is needed to guide treatment decisions for lung cancer patients with brain metastases.

Purpose of the Study:

• To develop and evaluate a deep learning and transfer learning model for non-invasively classifying SCLC versus NSCLC brain metastases using conventional MRI.
• To assess the model's performance using multiparametric MRI data.

Main Methods:

• A deep learning model (EfficientNet) was trained on conventional MRI data (T1-weighted post-contrast, T2-weighted, FLAIR) from 69 patients with lung cancer brain metastases (44 NSCLC, 25 SCLC).
• Classification was performed on lesion area crop images.
• Model evaluation utilized 5-fold cross-validation, assessing accuracy, precision, recall, F1 score, and AUC.

Main Results:

• The best classification performance was achieved using multiparametric MRI (T1WI+c+FLAIR+T2WI).
• Mean accuracy reached 0.90 ± 0.04 on validation data and 0.87 ± 0.05 on test data.
• Mean F1 scores were 0.92 ± 0.05 (NSCLC) and 0.87 ± 0.08 (SCLC) for validation, and 0.88 ± 0.05 (NSCLC) and 0.85 ± 0.05 (SCLC) for testing.

Conclusions:

• The proposed deep learning approach offers a highly sensitive and specific non-invasive method for differentiating SCLC from NSCLC brain metastases.
• This technique can serve as a valuable diagnostic tool to enhance treatment decision-making for patients with brain metastases.
• Further validation with larger patient cohorts is warranted to confirm these findings.