Research And Application Of Automatic Generation Of Medical Imaging Report Based On Multimodal Machine Learning

With the continuous development of scientific technology and medical standards,a large amount of medical imaging data is generated every day,which increases the burden of diagnosis and imaging report writing for radiologists.In recent years,the research of the automatic generation of image reports is mostly based on the model design of image caption.Although some results have been achieved,there are still certain limitations.There is no sufficient analysis of medical image features and medical semantics.The training of the model forcibly aligns the image features with the text features in the report,resulting in a low-quality medical image report and limited clinical significance.Based on this,this study proposes a medical imaging report automatic generation model based on Topic attention mechanism named TAMRGM.The model can fully consider the medical semantic features of image reports and the advantages of deep learning training models,and can automatically generate more accurate image reports.It first extracts the features of medical images,and fully integrates the medical semantic features of the image report.By adding attention mechanism to multi-modal fusion of image images and text features,the topic generator is used to complete the qualitative analysis of the image,and then generate the detailed information of the image report sentence by sentence to complete the generation of high-quality image report.Finally,the TAMRGM model is trained and evaluated based on the OpenI data set,and the image reports generated by the model are evaluated using BELU Score,METEOR,ROUGE,and CIDEr evaluation indicators.The experimental results show that the effect of the TAMRGM model is better than the CNN-RNN,Hierarchical Generation,and Co_Attention models,proving the effectiveness of the TAMRGM model for generating high-quality medical image reports.In summary,on the basis of the research and investigation of the automatic generation of medical image reports,the author proposes a model for automatic generation of medical image reports based on the Topic attention mechanism,TAMRGM,and the construction and effectiveness of the model on the chest X-ray data set OpenI verification.The research results are hopefully assist the radiologist to quickly complete the writing of the image report and provide better diagnosis and treatment services for the patients.