Detection rate of pulmonary nodules by CT scan at different doses and application value of artificial intelligence assistance / 中国医师进修杂志

ABSTRACT

Objective:To investigate the detection rate of pulmonary nodules (PN) by CT scan at different doses and the application value of artificial intelligence(AI) system assistance.Methods:From October 2019 to October 2021, 210 patients with PN in Beijing Hospital of Traditional Chinese Medicine, Capital Medical University, were retrospectively selected, and they were divided into the study group (106 cases) and the control group (104 cases) by CT scan at different doses. The control group used the conventional average dose (169 mAs) CT scan, the study group used an average low-dose (54 mAs) CT scan. The PN detection rate of different gender, age, body mass index (BMI) between the two groups were compared. The morphological characteristics, radiation dose, CT image quality between the two groups were compared. The diagnostic efficiency of radiologists and AI system was compared.Results:The detection rate of PN in the study group and the control group had no significant difference 73.58% (78/106) vs. 80.77%(84/104), χ2 = 1.54, P>0.05. The detection rate of PN with different gender, age group and BMI in the two groups had no significant differences ( P>0.05). The diameter of nodules and the rates of calcification, cavitation, bronchial sign, lobar sign, burr sign and pleural adhesion sign in the two groups had no significant differences ( P>0.05). The mean effective tubular bulb dose, length product of radiation dose, total tubular bulb dose, radiation volume dose index in the study group were higher than those in the control group (46.15 ± 7.38) mAs vs. (104.39 ± 10.53) mAs, (169.24 ± 19.77) mGy·cm vs. (427.17 ± 43.58) mGy·cm, (972.65 ± 58.34) mAs vs. (2 861.26 ± 181.37) mAs, (3.55 ± 1.16) mGy vs. (8.95 ± 2.07) mGy, there were statistical differences ( P<0.05). The excellent, good, acceptable, poor of 1.0 mm image quality in the study group were 26, 60, 18, 2, and in the control group were 32, 64, 8, 0, there was statistical difference ( u =1.71, P = 0.087). The excellent, good, acceptable, poor of maximum intensity projection (MIP) image quality in the study group were 58, 42, 6, 0 and in the control group were 70, 34, 0, 0, there was statistical difference ( u = 1.81, P = 0.070). The detection rate of PN by AI low-dose CT scan was higher than that of radiologists 88.68%(94/106) vs. 73.58%(78/106), there was statistical difference ( χ2 = 7.89, P = 0.005). Conclusions:The low-dose CT chest scans for PN, the results of detection rate, morphological characteristics, CT image quality are basically the same as those of conventional-dose CT chest scans, and can greatly reduce the radiation dose, which is more suitable for PN screening, and combined with AI system can significantly improve the detection rate of PN.