CT-Based Deep Learning Model for Invasiveness Classification and Micropapillary Pattern Prediction Within Lung...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Preoperative identification of tumor invasiveness in pulmonary adenocarcinomas is crucial for surgical planning.
• Accurate diagnosis of lung adenocarcinoma with micropapillary patterns is vital for clinical decision-making.

Purpose of the Study:

• To evaluate the accuracy of deep learning models in classifying the invasiveness degree of lung adenocarcinoma.
• To assess the capability of deep learning models in predicting the micropapillary pattern in lung adenocarcinoma.

Main Methods:

• Retrospective analysis of 291 histopathologically confirmed lung adenocarcinoma patient records.
• Development of two deep learning models, Lung-DL (LeNet) and Dense model (DenseNet architecture).
• Evaluation of model performance using Area Under the Curve (AUC) for invasiveness and accuracy for micropapillary pattern prediction.

Main Results:

• The Lung-DL model achieved an AUC of 0.88 for invasiveness classification, while the Dense model achieved 0.86.
• For micropapillary pattern prediction, the Lung-DL model reached 92% accuracy, and the Dense model reached 72.91% accuracy.
• Deep learning models demonstrated high accuracy in both invasiveness classification and micropapillary pattern prediction for lung adenocarcinoma.

Conclusions:

• Deep learning is effective for classifying the invasiveness of pulmonary adenocarcinomas.
• This study represents the first application of deep learning for predicting micropapillary patterns in lung adenocarcinoma.
• These deep learning applications can enhance efficiency and support the development of precise, individualized treatment strategies for lung cancer.