Cervical smear image diagnosis system on basis of convolutional neural networks and transfer learning

Abstract

The invention discloses a cervical smear image diagnosis system on the basis of convolutional neural networks and transfer learning. The cervical smear image diagnosis system comprises a sample database, a pre-training model, a reconstruction model and a to-be-tested sample acquisition module. The sample database is used for storing training samples; the pre-training model is a convolutional neural network model and comprises a convolutional layer, a pooling layer and a full-connection layer, and training can be carried out by the aid of the training samples in the sample databases, so that multi-classification requirements can be met by the pre-training model; network structures and parameters of the convolutional layer and the pooling layer in the pre-training model can be transferred bythe reconstruction model, the architecture of the full-connection layer can be modified, and accordingly requirements of cervical smear image classification tasks can be met by the reconstruction model; the to-be-tested sample acquisition module is used for acquiring to-be-tested cervical smear image samples, and the to-be-tested cervical smear image samples can be inputted into the reconstruction model, so that corresponding classification results can be obtained. The cervical smear image diagnosis system has the advantages that images with different sizes are allowed to be inputted into thecervical smear image diagnosis system, accordingly, high-performance classification effects can be realized by the cervical smear image diagnosis system, and excellent auxiliary effects can be realized by the cervical smear image diagnosis system for cervical cancer early screening and diagnosis.

Description

technical field [0001] The invention relates to intelligent diagnosis technology in biomedical information processing, in particular to a cervical smear image diagnosis system based on convolutional neural network and transfer learning. Background technique [0002] Cervical cancer is the fourth leading cause of cancer death in women and can be treated if detected early. Therefore, cervical cancer screening is very important for the treatment of cervical cancer. A Pap smear is a physical examination technique that is widely used to prevent cervical cancer and to find cells that are potentially cancerous. However, the diagnosis process relies heavily on the experience of doctors, which is very time-consuming on the one hand and prone to human misjudgment on the other hand. Due to the recent development of computer technology, computer-aided diagnosis system is becoming an important tool for pathologists to detect and diagnose cervical cancer, which can overcome subjective i...

AI Technical Summary

Problems solved by technology

Whereas transfer learning is able to exploit the lack of labeled data to train convolutional neural network models

Images