Performance of Deep-learning-based Artificial Intelligence on Detection of Pulmonary Nodules in Chest CT / 中国肺癌杂志

BACKGROUND@#The detection of pulmonary nodules is a key step to achieving the early diagnosis and therapy of lung cancer. Deep learning based Artificial intelligence (AI) presents as the state of the art in the area of nodule detection, however, a validation with clinical data is necessary for further application. Therefore, the aim of this study is to evaluate the performance of AI in the detection of malignant and non-calcified nodules in chest CT.@*METHODS@#Two hundred chest computed tomography (CT) data were randomly selected from a self-built nodule database from Tianjin Medical University General Hospital. Both the pathology confirmed lung cancers and the nodules in the process of follow-up were included. All CTs were processed by AI and the results were compared with that of radiologists retrieved from the original medical reports. The ground truths were further determined by two experienced radiologists. The size and characteristics of the nodules were evaluated as well. The sensitivity and false positive rate were used to evaluate the effectiveness of AI and radiologists in detecting nodules. The McNemar test was used to determine whether there was a significant difference.@*RESULTS@#A total of 889 non-calcified nodules were determined by experts on chest CT, including 133 lung cancers. Of them, 442 nodules were less than 5 mm. The cancer detection rates of AI and radiologists are 100%. The sensitivity of AI on nodule detection was significantly higher than that of radiologists (99.1% vs 43%, P<0.001). The false-positive rate of AI was 4.9 per CT and decreased to 1.5 when nodules less than 5 mm were excluded.@*CONCLUSIONS@#AI achieves the detection of all malignancies and improve the sensitivity of pulmonary nodules detection beyond radiologists, with a low false positive rate after excluding small nodules.

Low-dose CT pulmonary angiography: a phantom study / 中华放射学杂志

Objective To investigate the feasibility of reduced radiation dose for CT pulmonary angiography (CTPA) and the possible lowest radiation threshold by a phantom study.Methods The CT value difference between air within the trachea and the extracorporeal background region was measured in132 consecutive patients.A noise-measurement phantom and a pulmonary embolism (PE) phantom were made of phenol-formaldehyde, and both phantoms and a water phantom were scanned with standard and lower radiation doses as follow: 280, 200, 160, 100, 90, 80, 70, 60, 50, 40, 30, 20, 15, and 10 mA respectively, at a fixed voltage of 120 kVp.Standard and soft tissue algorithms were used to reconstruct the images.Three experienced doctors independendy evaluate the image quality and the efficiency of detecting PE of the images with various doses.The Pearson correlation analysis, two-tailed paired t test, ANOVA, and Kappa test were employed for the statistical analysis.Results The CT value difference between air within the trachea and the extracorpereal background region in 132 consecutive patients ranged from 20.00 to 55.00 HU, which had a positive correlation with weight[(64.99±11.86) kg], weight-height ratio [(38.71±6.13) kg/m], and BMI[(23.11±3.38) kg/m2](r=0.228,0.374,0.449 respectively; P <0.01).The image noise level with soft-tissue reconstruction algorithm[(16.55±9.08), (16.42±9.40) HU]was significantly lower than that of the image with standard reconstruction algorithm[(22.43±11.25),(21.99±11.67) HU](F=4.316, P < 0.05).The image noise level with soft-tissue reconstruction algorithm at 100 mA was similar to that of the images with standard reconstruction algorithm at 280 mA, and the signal-w-noise ratio (SNR) and contrast-to-noise ratio (CNR) of the image of PE phantom was 23.05 and 20.52 respectively, without any impairment in detectability of embolus.The image noise level with soft-tissue reconstruction algorithm at 60 mA was similar to that of the image with standard reconstruction algorithm at 160 mA, while the SNR and CNR was 18.01 and 15.97 respectively, also with acceptable detectability of embolus.When the tube current was reduced below 30 mA, the image quality decreased significantly, with the SNR and CNR was lower than 12.36 and 10.95 respectively, and the detectability of embolus was degraded.The consistency of the image quality grading by 3 observers was excellent(K=0.807,0.712,0.904 ,respectively; P < 0.01).Conclusions The 100 mA may potentially be the ideal low dose tube current setting, with radiation dose only equal to 36% of 280 mA (standard dose).The 30 mA may possibly be a minimum radiation dose for detecting PE.The soft-tiasue reconstruction algorithm was favorable in preserving the SNR when the radiation dose was reduced.