<p>Electronic Medical Record (EMR) has been widely utilized in disease prevention, clinical diagnosis and scientific research. EMR mining based on Named Entity Recognition (NER) has become a research hotspot in the healthcare field. However, NER faces challenges of poor context modeling ability and severe entity nesting. In order to deal with the above challenges, this paper proposes a Machine Reading Comprehension based Named Entity Recognition for medical text (MRCNER). First, the context sequence and the question are inputs to an embedding and encoding module composed of BERT, Dropout, and BiLSTM to obtain text vector representations and problem vector representations that contain global semantics, as well as better understanding of the relationship between language structure and context. Then, the bidirectional attention is adopted for the aim of fully learning the semantic association between text and entity categories, and focusing more on key information. Next, we propose an answer prediction module. It is consisted of a start index classifier, an end index classifier, and a match classifier. The first two classifiers determine the answer by predicting its starting and ending positions in the original context, avoiding the generation of discontinuous named entities. The match classifier is designed to predict whether the starting position and the ending position match. Finally, the predicted results are obtained. We evaluated MRCNER on the CCKS2020 (China Conference on Knowledge Graph and Semantic Computing) dataset and demonstrated that the accuracy, recall and F1 value of the model have all been improved to 84.48%, 82.96% and 83.71%, respectively.</p>