WIDINet: A diagnostic model for staging pneumoconiosis based on data expansion and KL entropy judgement

Abstract

Occupational pneumoconiosis brings great difficulties to the diagnosis of professional doctors due to its complex lung characteristics. Computer-aided diagnosis has opened up new research ideas for researchers, but due to the inherent small data characteristics of medical data sets, it performs poorly in specific practice. In this paper, we propose a weighted information flow data incremental network (WIDINet) for pneumoconiosis staging based on existing research results. WIDINet first uses random attribute masking technology to obtain incremental data with higher data quality, and then uses prior knowledge and multi-granularity attention modules to solve the problem of low reliability of feature screening networks based on incremental data. At the end of the WIDINet diagnostic model, this article uses the KL weighted judgment algorithm to measure the difference between the test chest radiograph and China's current national gold standard to achieve staged diagnosis of the target task. In the experimental verification part, we used the pneumoconiosis data set collected in our laboratory to prove that the WIDINet staging diagnosis model has an accuracy of 93.4%, a precision of 87.8%, a sensitivity of 78.4%, a specificity of 95.6%, and an F1score of 84.9%. The area under the curve (AUC) reaches 96%.Code: https://github.com/nonoXwb/SDom (github.com)