Transfer learning VGG for histopathological lung cancer image classification.

Lung cancer diagnosis can be done using histopathology techniques. A biopsy is performed on body tissue suspected of having cancer, and the biopsy results are observed through a microscope by a pathologist. This study utilized medical imaging results from lung cancer histopathology. The research aims to classify lung cancer histopathology images. Experimental data comes from the public dataset LC25000. The data has three classes: adenocarcinoma, normal, and squamous cell carcinoma. Each class has 1,000 images. The research steps consist of preprocessing and classification. The preprocessing includes image resizing 224, converting BGR to RGB image format, and splitting training, validation, and testing data. Training vs. testing data has a percentage of 80:20. Meanwhile, for training and validation, 5-fold cross-validation is used. We use Transfer Learning Convolutional Neural Network (CNN) with Visual Geometry Group architecture (VGG16 and VGG19) for classification. Network optimization was done by initializing the learning rate 64, batch-size 32, learning rate 0.0001, and the Adaptive Moment Estimation (Adam) optimizer. Also, using early-stop in the training process when validation accuracy does not improve along with increasing epochs. The experimental results show the best accuracy, precision, recall, and f1-score of 0.97. [ABSTRACT FROM AUTHOR]