人工智能辅助诊断对肺结节检出及良恶性鉴别的影响

Effect of artificial intelligence-assisted diagnosis on the detection of pulmonary nodules and differentiation between benign and malignant

目的 探究人工智能辅助诊断对肺结节检出及良恶性鉴别的影响。方法 选取本院2022年7月至2023年5月收治的128例肺结节患者,均接受CT检查,其结果分别采用人工智能和影像医师阅片判断,以手术或穿刺活检病理结果为金标准,比较2种阅片方式对肺结节的检出结果、诊断效能以及对不同直径结节、诊断价值。结果 128例患者共检出216枚结节,经病理检查确诊145枚恶性,71枚良性;人工智能阅片检出真实恶性结节120枚,影像医师阅片检出真实恶性结节75枚;人工智能阅片诊断灵敏度、准确度、阴性预测值均高于影像医师阅片;人工智能阅片诊断特异度低于影像医师阅片。人工智能阅片检出141枚阳性结节,影像医生阅片检出85枚阳性结节。经比较两者对于直径<5 mm、5 mm≤直径≤10 mm结节的检出情况,人工智能阅片检出率更高;经比较两者对于实性、磨玻璃结节的检出情况,人工智能阅片检出率更高(P<0.05)。结论 人工智能辅助诊断对直径较小、实性、磨玻璃结节肺结节的检出率更高,并且灵敏度、准确度、阴性预测值较高,但其特异度较低,建议临床上将影像医师和人工智能辅助诊断相结合,以提高对肺结节的诊断准确性。

Objective To explore the influence of artificial intelligence-assisted diagnosis on the detection of pulmonary nodules and the differentiation between benign and malignant.Methods A total of 128 patients with pul-monary nodules admitted to our hospital from July 2022 to May 2023 were examined by CT.The results were judged by artificial intlligence and imaging doctor's reading,Taking the pathological results of operation or puncture biopsy as the gold standard,the detection results,diagnostic efficiency,and diagnostic value of two ways of reading films for pulmonary nodules and nodules with different diameters were compared.Results A total of 216 nodules were detect-ed in 128 patients,of which 145 were malignant and 71 were benign by pathological examination.A total of 120 real malignant nodules were detected by artificial intelligence,and 75 real malignant nodules were detected by imaging doctors.The diagnostic sensitivity,accuracy,and negative predictive value of artificial intelligence film reading were higher than that of imaging doctors.The diagnostic specificity of artificial intelligence reading film was lower than that of imaging doctor reading film.141 positive nodules were detected by artificial intelligence reading,and 85 positive nodules were detected by imaging doctors reading.The two methods were compared,the detection rate of artificial in-telligence reading was higher for nodules with diameters less than 5 mm and 5 mm≤10 mm.The detection rate of ar-tificial intelligence reading was higher(P<0.05).Conclusion Artificial intelligence-assisted diagnosis has higher de-tection rate and higher sensitivity,accuracy,and negative predictive value,but its specificity is low.It is suggested that artificial intelligence and imaging doctor's diagnoses should be combined in clinic to improve the diagnostic ac-curacy of pulmonary nodules.