Assisting Scalable Diagnosis Automatically via CT Images in the Combat against COVID-19

Introductory paragraphThe pandemic of coronavirus Disease 2019 (COVID-19) caused enormous loss of life globally. 1-3 Case identification is critical. The reference method is using real-time reverse transcription PCR (rRT-PCR) assays, with limitations that may curb its prompt large-scale application. COVID-19 manifests with chest computed tomography (CT) abnormalities, some even before the onset of symptoms. We tested the hypothesis that application of deep learning (DL) to the 3D CT images could help identify COVID-19 infections. Using the data from 920 COVID-19 and 1,073 non-COVID-19 pneumonia patients, we developed a modified DenseNet-264 model, COVIDNet, to classify CT images to either class. When tested on an independent set of 233 COVID-19 and 289 non-COVID-19 patients. COVIDNet achieved an accuracy rate of 94.3% and an area under the curve (AUC) of 0.98. Application of DL to CT images may improve both the efficiency and capacity of case detection and long-term surveillance.

CT and MR features of hepatic ischemia/necrosis after hepatosplenic surgery and upper gastrointestinal hemorrhage / 中华肝胆外科杂志

Objective@#To investigate the CT and MR imaging features of hepatic ischemia/necrosis after hepatosplenic surgery and upper gastrointestinal hemorrhage.@*Methods@#A total of 36 patients diagnosed with hepatic ischemia/necrosis by both medical imaging and clinical diagnosis shortly after hepatosplenic surgery and upper gastrointestinal hemorrhage were collected, including 9 patients with liver cancer resection, 5 patients with liver cancer ablation (microwave ablation/radiofrequency ablation, argon-helium knife, alcohol injection), 11 patients with spleen resection, and 11 patients with upper gastrointestinal bleeding. Conventional liver CT and / or MR plain and dynamic enhancement scan were performed to comprehensively analyze the morphology and density/signal performance of the lesions.@*Results@#(1) Number of lesions: All cases had multiple lesions. (2) Distribution of lesions: scattered in the liver lobes, clustered or regional distribution, mainly in the periphery of the liver. (3) Size of lesions: the boundary of the nodular lesion was clear, and the single maximum diameter was 1.0-1.5 cm. It can be fused into a wedge-shaped patch or a segmental/sub-segmental large patch with a slight mass effect. (4) CT density or MR signal characteristics: CT plain scan showed slightly low density; MR plain scan showed slightly low signal on T1WI, high signal on T2WI, slightly high signal on DWI and no lipid/fat on dual phase imaging; 24 out of 36 cases (66.7%) showed no enhancement, while some lesions showed thin ring enhancement on the edge; emboli were found in the main and/or branches of portal vein (21/36 cases, 58.3%). (5) In the short-term review (minimum 5 days), the lesions became smaller or disappeared, and the local liver volume became smaller or the surface was depressed.@*Conclusions@#Hepatic ischemia/necrosis occurs after hepatosplenic surgery and upper gastrointestinal hemorrhage. The imaging manifestations are multiple nodular or flaky hypovascular foci, and the short-term review shows a markedly improvement. It needs to be differentiated from infection and metastasis of malignant tumors after operation.

CT and MR features of hepatic ischemia/necrosis after hepatosplenic surgery and upper gastrointestinal hemorrhage / 中华肝胆外科杂志

Objective To investigate the CT and MR imaging features of hepatic ischemia/necrosis after hepatosplenic surgery and upper gastrointestinal hemorrhage.Methods A total of 36 patients diag-nosed with hepatic ischemia/necrosis by both medical imaging and clinical diagnosis shortly after hepato-splenic surgery and upper gastrointestinal hemorrhage were collected,including 9 patients with liver cancer resection,5 patients with liver cancer ablation(microwave ablation/radiofrequency ablation,argon-helium knife,alcohol injection),1 1 patients with spleen resection,and 11 patients with upper gastrointestinal bleeding.Conventional liver CT and/or MR plain and dynamic enhancement scan were performed to com-prehensively analyze the morphology and density/signal performance of the lesions.Results(1)Number of lesions:AU cases had multiple lesions.(2)Distribution of lesions:scattered in the liver lobes,clustered or regional distribution,mainly in the periphery of the liver.(3)Size of lesions:the boundary of the nodu-lar lesion was clear.and the single maximum diameter Was 1.0-1.5 ca.It can be fused into a wedge-shaped patch or a segmental/sub-segmental large patch with a slight mass effect.(4)CT density or MR sig-nal characteristics:CT plain scan showed slightly low density;MR plain scan showed slightly low signal on T1 WI,high signal on T2WI,slightly high signal on DWI and no lipid/fat on dual phase imaging;24 out of 36 cases(66.7％)showed no enhancement,while some lesions showed thin ring enhancement on the edge;emboli were found in the main and/or branches of portal vein(21/36 cases,58.3％).(5)In the short-term review(minimum 5 days),the lesions became smaller or disappeared,and the local liver volume be-came smaller or the surface was depressed.Conclusions Hepatic ischemia/necrosis OCCURS after hepato-splenic surgery and upper gastrointestinal hemorrhage.The imaging manifestations are multiple nodular or flaky hypovascular foci,and the short-term review shows a markedly improvement.It needs to be differentia-ted from infection and metastasis of malignant tumors after operation.

Value of diffusion?weighted imaging in the evaluation of therapeutic effect of cyber knife on hepatocellular carcinoma / 中华放射学杂志

Objective To investigate the value of DWI in the evaluation of the therapeutic effect of cyber knife in the treatment of hepatocellular carcinoma.Methods Ninety patients who were clinically diagnosed with hepatocellular carcinoma (95 lesions) in the 5th Medical Center of General Hospital of PLA from February 2011 to December 2013 were retrospectively analyzed. All 90 patients underwent pre‐treatment dynamic contrast‐enhanced MRI scans of liver. Fourty six of them underwent liver MRI scan 3 months after treatment with cyber knife, and 49 patients underwent liver MRI scans 6 months after treatment. According to the evaluation criteria of solid tumor effect, complete necrosis tumor lesions and postoperative residual active lesions were determined. The ADC values of residual active tissue lesion, necrosis lesion in tumor and normal liver were measured. Paired sample t test was used to compare the difference of ADC values of tumor necrosis lesions and normal liver tissues between 3 and 6 months after treatment and before treatment, and ROC was used to evaluate the efficacy of ADC values in predicting complete tissue necrosis after treatment. Results In the 95 lesions, 91 lesions were completely necrotic, and active tissues were found in 4 lesions after treatment. There were statistically significant differences in the ADC values of the lesions at 3 and 6 months after treatment and before treatment (P<0.05),and the ADC values of the tissues after treatment were higher than those before treatment. There was no statistically significant difference in ADC values between 3 and 6 months after treatment and before treatment in normal liver tissue (P>0.05). The area under the ROC of using ADC value to predict the complete necrosis after treatment with cyber knife was 0.767, and the ADC value was 1.23×10‐3 mm2/s as the diagnostic cutoff value. The sensitivity and specificity of the diagnosis were 76.5% and 70.6%, respectively. Conclusion ADC values can distinguish the necrotic component from the active component of hepatocellular carcinoma after treatment, and can be used to evaluate the therapeutic effect of cyber knife in hepatocellular carcinoma.

Application of intravoxel incoherent motion diffusion weighted imaging in inflammation activity of chronic hepatitis B / 中华放射学杂志

Objective To discuss the diagnostic value of intravoxel incoherent motion(IVIM)DWI in the inflammation activity of chronic hepatitis B. Methods One hundred and seventy-five patients with chronic hepatitis B were prospectively collected as inflammation group, and 15 healthy volunteers with no history of hepatitis and normal liver function as control group.All subjects were performed MR plan scans for liver and IVIM tests.The liver apparent diffusion coefficient(ADC),the diffusion coefficient of the true water molecule (D), microcirculation perfusion coefficient (D*) and perfusion fraction (f) were measured. After MRI examination, the inflammation group received liver biopsy within two days, and obtained the pathological hepatitis activity grades.According to the degree of inflammation activity(G),the inflammation group was divided into G1,G2 and G3+G4 groups.The independent sample t test was used for the parameter values between the inflammation group and the control group.A single factor variance analysis was used to compare the parameters among different levels of inflammation group. Spearman correlation analysis was used to evaluate the correlation between inflammation activity grades and IVIM parameters.ROC curve was used to evaluate each parameter in the diagnosis of hepatitis activity pathology classification efficiency.Results The values of ADC, D, D*, f between control group and inflammation group were statistically significant(P<0.05).The results of pathological puncture showed that there were 61 cases of G1,62 cases of G2 level and 52 cases of G3+G4.The ADC,D*and f values in the G1,G2,G3+G4 groups decreased with the increase of the level of inflammation activity, and the difference was statistically significant (P<0.05). There was no statistically significant difference between D values (P>0.05). ADC, D* and f values were correlated negatively with low,moderate and low levels of inflammation activity(r=-0.377,-0.434,-0.226, P<0.05).There was no correlation between D value and inflammation activity(r=-0.076,P>0.05).Areas of ADC,D*and ADC combined with D*values under ROC curves to diagnose≥G2 grade inflammation activity were 0.732, 0.748, 0.837 respectively. Areas of f values under ROC curve to diagnose ≥G3+G4 grade inflammation activity was 0.600.Conclusion ADC and D*values help to identify≥G2 grade inflammation activity,and ADC combined with D*value has the highest diagnostic performance,and has certain reference value for clinical treatment.

Differential Diagnosis of Common Liver Lipid-containing Tumors Using Diffusion Weighted Imaging / 中国医学影像学杂志

Purpose To investigate the application value of diffusion weighted imaging (DWI) in differential diagnosis of common liver lipid-containing tumors.Materials and Methods Sixty patients with pathologically confirmed liver lipid-containing tumors from January 2011 to August 2015 were retrospectively analyzed,including 25 cases ofhepatocellular carcinoma (HCC),20 cases of hepatic adenoma (HA) and 15 cases of hepatic angiomyolipoma (HAML).Apparent diffusion coefficient (ADC) value and ratio of ADCtumor/ADCliver (rADC) of tumors were respectively measured.Routine MRI findings were analyzed.Results The MRI findings of these three common liver lipid-containing tumors were with their own characteristics.For HCC,HA and HAML,the average ADC value was (1.225±0.221)×10-3 mm2/s,(1.318±0.212)×10-3 mm2/s and (1.317±0.297)×103 mm2/s,respectively,(b=800 s/mm2);the average rADC value was 1.004± 0.151,0.984 ± 0.146 and 1.027±0.223,respectively.There was no significant difference in ADC and rADC among these three tumors (P＞0.05).Conclusion ADC and rADC had no differential diagnosis value in common liver lipid-containing tumors.Routine MRI signals and enhanced imaging features are the main diagnostic basis.

Comparative study of multi-detector CT and breathhold 3D MR coronary angiography in patients with coronary artery disease / 中华放射学杂志

50%) in patients with suspected coronary artery disease.Methods Both coronary MDCT angiography (CTA) and MR angiography (MRA) was performed within 3 days in 40 patients with suspected coronary artery disease, and conventional coronary angiography (CAG) was performed within 2 weeks after MDCT and MR scan in 31 patients. CTA was performed with a 16-MDCT scanner. MRA was performed on a 1.5 T MR scanner with breathhold 3D fast imaging employing steady state acquisition sequence. CTA and MRA image quality was evaluated in 9 coronarysegments by two experienced radiologists in concensus using a four-point grading scale. Sensitivity, specificity, positive predictive value, and negative predictive value were calculated for detection of significant stenosis on a segmental basis using CAG as reference and gold standard. Results MDCT showed higher image quality in most coronary segments except middle RCA (P