ObjectiveTo realize automatic risk bias assessment for the randomized controlled trial (RCT) literature using BERT (Bidirectional Encoder Representations from Transformers) as an approach for feature representation and text classification.MethodsWe first searched The Cochrane Library to obtain risk bias assessment data and detailed information on RCTs, and constructed data sets for text classification. We assigned 80% of the data set as the training set, 10% as the test set, and 10% as the validation set. Then, we used BERT to extract features, construct text classification model, and evaluate the seven types of risk bias values (high and low). The results were compared with those from traditional machine learning methods using a combination of n-gram and TF-IDF as well as the Linear SVM classifier. The accuracy rate (P value), recall rate (R value) and F1 value were used to evaluate the performance of the models.ResultsOur BERT-based model achieved F1 values of 78.5% to 95.2% for the seven types of risk bias assessment tasks, which was 14.7% higher than the traditional machine learning method. F1 values of 85.7% to 92.8% were obtained in the extraction task of the other six types of biased descriptors except "other sources of bias", which was 18.2% higher than the traditional machine learning method.ConclusionsThe BERT-based automatic risk bias assessment model can realize higher accuracy in risk of bias assessment for RCT literature, and improve the efficiency of assessment.
Objective To construct large medical model named by "Huaxi HongYi"and explore its application effectiveness in assisting medical record generation. Methods By the way of a full-chain medical large model construction paradigm of "data annotation - model training - scenario incubation", through strategies such as multimodal data fusion, domain adaptation training, and localization of hardware adaptation, "Huaxi HongYi" with 72 billion parameters was constructed. Combined with technologies such as speech recognition, knowledge graphs, and reinforcement learning, an application system for assisting in the generation of medical records was developed. ResultsTaking the assisted generation of discharge records as an example, in the pilot department, after using the application system, the average completion times of writing a medical records shortened (21 min vs. 5 min) with efficiency increased by 3.2 time, the accuracy rate of the model output reached 92.4%. Conclusion It is feasible for medical institutions to build independently controllable medical large models and incubate various applications based on these models, providing a reference pathway for artificial intelligence development in similar institutions.