Abstract:

Objective Using computer learning software（SPSS 21.0）,to establish a peripheral blood blast intelligent determination model,and to lay a foundation for the intelligent review of blood routine reports and the intelligent diagnosis of laboratory determination results. Methods A total of 6 173 outpatients and inpatients were enrolled,and their data were collected through laboratory information system（LIS） and hospital information system（HIS） of West China Hospital of Sichuan University. A total of 64 blood cell determination indexes were enrolled,and the P value of blast was obtained by computer statistical software. The screening performance was evaluated. Results The indexes,including hemoglobin（Hb）,hematocrit（HCT）,lymphocyte absolute value（LYMP#）,nucleated red blood cell percentage（NRBC%）,lymphocyte corrected absolute value（LYMPH#）,forward scatter intensity distribution width of lymphocyte（LY-WZ）,immature granulocyte absolute value（IG#）,forward scatter intensity distribution width of neutrophil（NE-WZ）,white blood ecell count in WDF channel（WBC-D）,white blood cell（WBC） count,white blood cell in WNR channel（WBC-N）,nucleated cell count in WDF channel（TNC-D）,total nucleated cell（TNC） count,nucleated cell count in WNR channel（TNC-N）,side scatter intensity distribution width of neutrophil（NE-WX）,side fluorescence intensity distribution width of monocyte（MO-WY）,side fluorescence intensity distribution width of neutrophil（NE-WY）,side scatter intensity distribution width of monocyte（MO-WX）,nucleated red blood cell absolute value（NRBC#） count and monocyte absolute value（MONO#）,were included in the model. The P value of blast was 96.4%. In this model,989 cases of outpatients were verified by blind method,the predictive sensitivity was 100.0%,and the specificity was 53.2%. Furthermore,493 cases of inpatients were verified by blind method,the predictive sensitivity was 68.1%,and the specificity was 52.4%. Conclusions The model of peripheral blood blast intelligent determination based on machine learning plays a role in blood cell analysis,which provides experience for the application of artificial intelligence in intelligent diagnosis in the future. The model's ability for determining blood diseases under treatment needs to be improved,and the machine learning algorithm should be optimized for a better model for determining peripheral blood blast.

Key words: Machine learning, Blood cell count, Blast