With rapid development of medical information, more and more attention has been paid to the intelligent diagnosis based on electronic medical records. The results of intelligent diagnosis can provide doctors with advice in the course of diagnosis and treatment,and avoid unnecessary mistakes. In the field of intelligent text aided diagnosis, how to extract and express a large amount of text information in the medical record plays a key role. The final result of classification and prediction is based on the characterization of text information using some mathematical models.In the past,we used single feature set to get the final results, which could not achieve our expect, and the results affected the reliability of intelligent diagnosis.This paper puts forward a kind of combined text representation method based on weighted LDA and word vector model, we get the combined feature set by this way and use it to classify liver disease.By combining the weighted LDA and the word vector model, we can make full use of the potential text information of the electronic medical records, improve the accuracy of classification prediction, and increase the reliability of intelligent diagnosis.

Latent Dirichlet allocation (LDA) and other modified topic models have become the prevalent tools for semantic analysis and text data mining. With the rapid development of the medical information, large amount of data has been accumulated in the form of text, while most of which recording in a confused structure. Based on LDA, this paper proposes a revised approach, which enlightens an idea of weighting. This paper also demonstrates the desirable performance of the modified method, thus yields the proof of its effectiveness.

According to the theory of CT reconstruction algorithm, mainly filter back projection algorithm, we make an effort to improve the quality of reconstructed image. In the aspect of interpolation algorithm, we propose a new weighted interpolation algorithm, which take full advantage of both nearest neighbor interpolation and bilinear interpolation. We apply it to FDK reconstruction algorithm and the experimental results show that the new interpolation algorithm could preserve the image edge details and also can effectively suppress noise while takes little computational resource cost.