BiLSTM-Attention-Transformer Model Achieves 99.8% Accuracy for Lung Cancer Detection from Clinical Text

A deep learning model (BiLSTM + attention + transformer) achieved 99.8% accuracy for lung cancer detection from clinical text (including radiology reports) in an experimental study. Training time was 13.99 s. Small dataset and overfitting concerns limit generalizability.

• The hybrid model combined Bidirectional LSTM, attention mechanism, and transformer to overcome overfitting, data imbalance, and long-sequence dependency issues.
• Validation loss was reduced to 0.3112, with a fast training time of 13.99 seconds.
• Authors note the small dataset may cause overfitting, and call for larger studies with advanced regularization methods.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.