Automated Classification of Urinary Cells: Using Convolutional Neural Network Pre-trained on Lung Cells

Urine cytology, which is based on the examination of cellular images obtained from urine, widely used for diagnosis bladder cancer. However, sometimes difficult in highly heterogeneous carcinomas exhibiting weak atypia. In this study, we propose a new deep learning method that utilizes image information another organ automated classification urinary cells. We first extracted 3137 291 lung cytology specimens biopsies and trained process benign malignant cells using VGG-16, convolutional neural network (CNN). Subsequently, 1380 were 123 urine to fine-tune CNN was pre-trained with To confirm effectiveness proposed method, introduced three different training methods compared their performances. The evaluation results showed accuracy fine-tuned 98.8% regarding sensitivity 98.2% specificity cells, higher than those only or could be automatically classified high rate. These suggest possibility building versatile deep-learning model organs.