Artificial Intelligence Assisted Morphological Classification of Nucleated Red Blood Cells in Peripheral Blood
Objective To evaluate the performance of artificial intelligence (AI )-based blood cell morphology analyzer (AI reader)in detecting peripheral blood nucleated red blood cells (NRBC). Methods We collected 1191 venous blood samples (automatic blood cell analyzer alarm indicating nucleated red blood cells)and then divided these samples into three groups:low-,medium-,and high-absolute NRBC values.AI reader was used to detect NRBC％,and statistical analysis was performed on the pre-classification results and the results were reviewed by morphological experts;We collected 494 venous blood samples (NRBC absolute value ≥1×109/L),and classified peripheral blood cell morphology using an AI reader.We then corrected the results by morphological experts, and created a confusion matrix for blood cell classification, and evaluated the consistency and detection rate of the AI reader for peripheral blood cell classification.Results There was no statistically significant difference (P<0.05)between the NRBC％ detection results of the AI film reader and the expert review results,with R2>0.99.The AI film reader detected a total of 99137 NRBCs,in which 99031 sample were identified correctly,while the expert correction result was 99976.The compliance rate of the AI film reader for detecting NRBC was 99.89％,and the detection rate was 99.05％.Conclusion The AI film reader shows a high consistency and detection rate between the detection results of NRBC in peripheral blood and the results reviewed by experts,providing an efficient auxiliary detection method for screening NRBC related diseases.
Artificial intelligence(AI)Nucleated red blood cellsNRBC
万方数据
维普
