تشخيص بيماري با استفاده از تركيب داده هاي ساختاريافته و بدون ساختار در پرونده سلامت الكترونيك

Disease diagnosis using a combination of structured an‎d unstructured data in electronic health records.

مهندسي نرم افزار

Today, with the digitization of medical records around the world an‎d the advancement of technology, electronic health records have become an important source of real-world data. In electronic health records, patient health information is stored. This information includes structured data (such as vital signs, age, blood pressure, gender) an‎d unstructured information (such as notes, clinical reports, an‎d medical prescriptions). There is very important information in the unstructured data category. Combining this data can lead to more accurate diagnosis of the disease. Since a larger volume of patient information is made up of unstructured data, the use an‎d analysis of unstructured data is of great importance. In this study, unstructured data an‎d structured data from electronic health records are considered for disease diagnosis. In this study, the MIMIC-III dataset, which includes a variety of structured an‎d unstructured temporal data, was used. To do this, in the first step, appropriate columns of structured data were extracted from this dataset, in the second step, unstructured data an‎d medical notes of this dataset were analyzed an‎d unstructured data were transformed into embedded vectors using the Clinical BERT model. In the third step, a table of the new dataset consisting of structured data an‎d unstructured data results was created. In the fourth step, disease diagnosis was selec‎ted an‎d analyzed using this dataset an‎d modeling was performed with the LSTM model an‎d BERT model with this data combination. The input of the model is the data combination an‎d its output is the disease diagnosis code. The results show that the disease code diagnosis, considering all data types, has a precision of 0.94, recall of 0.93 an‎d precision of 0.94. The disease code is the exact identifier of the disease name according to the MIMIC-III dataset. Having a model with these results in the health field will greatly contribute to the treatment process, early detection of diseases, an‎d prevention of mortality.

6

156959