Therefore, we established a respiratory system big data platform in Kashi, using machine learning, natural language recognition and extraction methods to discuss the information in patients' electronic medical records, and established algorithm models to achieve high accuracy through model autonomous learning. The dry climate, air pollution and rural biofuels have led to a high incidence of chronic airway diseases, and the limited experience and medical equipment of doctors in primary hospitals in Kashi area of China, make it difficult to identify similar diseases ( 6). Misdiagnosis is common in primary hospitals, and can lead to inappropriate treatment, prolonged recovery time, and potential deterioration, and limited experience of doctors at primary hospitals also worsens the situation ( 5). The similar symptoms among these diseases make timely diagnosis difficult. These diseases have common symptoms such as cough, sputum expectoration, wheezing, and chest pain, but the treatment and follow-up of each disease are completely different ( 1вЂ“ 4).

Respiratory diseases, including pulmonary tuberculosis (PTB), chronic obstructive pulmonary disease (COPD), pulmonary thromboembolism (PTE), and bronchiectasis, are among the most common diseases clinically. Results: The precisions of traditional machine-learning, deep-learning methods and our proposed method were 0.6, 0.81, 0.89, and F1 scores were 0.6, 0.81, 0.88, respectively.Д¬onclusion: Compared with traditional machine learning and deep-learning methods that our proposed method had a significantly higher performance, and provided precise identification of respiratory disease. The partial and total information of the fused feature set was encoded by convolutional layers, while LSTM layers decoded the encoded information.

Pre-processing of clinical records with вЂњjiebaвЂќ word segmentation module, and the Bidirectional Encoder Representation from Transformers (BERT) model was used to perform word vectorization on the text. We collected the medical records of inpatients in the respiratory department, including: chief complaint, history of present illness, and chest computed tomography. Methods: The method combined convolutional neural network (CNN) and long-short-term memory network (LSTM) for prediction and diagnosis of respiratory diseases. Objective: Based on the respiratory disease big data platform in southern Xinjiang, we established a model that predicted and diagnosed chronic obstructive pulmonary disease, bronchiectasis, pulmonary embolism and pulmonary tuberculosis, and provided assistance for primary physicians.

5Department of Software, Sun Yat-sen University, Guangzhou, China.

4Department of Preventive Medicine, School of Public Health, Sun Yat-sen University, Guangzhou, China.

3State Key Laboratory of Pathogenesis, Prevention and Treatment of High Incidence Diseases in Central Asia, Xinjiang Medical University, ГњrГјmqi, China.

2Department of Clinical Research Center of Infectious Diseases (Pulmonary Tuberculosis), First People's Hospital of Kashi, Kashi, China.

1Department of Respiratory and Critical Care Medicine, First People's Hospital of Kashi, Kashi, China.

Li Li 1,2,3 вЂ , Alimu Ayiguli 2 вЂ , Qiyun Luan 2 вЂ , Boyi Yang 4, Yilamujiang Subinuer 2, Hui Gong 2, Abudureherman Zulipikaer 2, Jingran Xu 2, Xuemei Zhong 1, Jiangtao Ren 5 * and Xiaoguang Zou 2 *

Text to table word