Text mining approach for the prediction of disease status from discharge summaries using CCBE and NEROA-CNN

A necessary document that encompasses clinical and administrative information indispensable for the continuity of care subsequent to the patients being discharged as of hospitals is called the hospital Discharge Summary (DS). Yet, it is complex to extract information automatically from DSs with natural language. Thus, a Text Mining (TM) approach has been proposed to predict the disease status utilizing DSs via Clustering centered Clinical Bert Embedding (CCBE) and Nudged Elastic Rider Optimization Algorithm with Convolutional Neural Network (NEROA-CNN). Initially, the input data is amassed from the publically available patient information dataset in the proposed model. Next, pre-processing of the patient information module occurs. After that, to extort the header details from the dataset, the header extraction is done. Subsequently, by utilizing Harris Hawks Gravitational Search Strategy Optimization Algorithm (H2GS2), the header selection is performed. Next, the collection of DS modules occurs, and it undergoes pre-processing. The subsequent step after pre-processing is the medical term identification. Then, utilizing the Patient ID attribute, the patient information module and DS module will be combined. After that, utilizing the CCBE method, word embedding is carried out. Lastly, by employing the NEROA-CNN, the classification is executed. Ultimately, the proposed technique's performance is examined regarding the performance metrics by analogizing it with the prevailing methodologies. The findings exhibit that the proposed model surpasses the other top-notch models.