Objective To study the accuracy,clinical value and limit in assessing the vessels surrounding pancreas involved by pancneutic carcinoma with low field MRI.Methods MRI scanning was performed in 55 patients with pancreatic carcinoma.We prospectively evaluated the MRI findings of vascular involvement by pancreatic carcinoma,and compared with the surgical results.Results Of total 55 cases,50(91%)cases were correctly diagnosed on MRI .MRI carries an accuracy of 86% for prediction of resectility,and accuracy of 93% for unresectility.Conclusion Low field MRI can accurately assess the peripancreative vessels ,but it has limited value in the patients with pancreatic carinomaⅠa orⅠb.

Objective To access the value of three dimensional gradient echo two-point Dixon water-fat separation sequence (3 D-DIXON) in quantitative diagnosis of non-alcoholic fatty liver disease(NAFLD) by comparing with in-phase/opposed-phase (IN/OPP)and proton magnetic resonance spectroscopy (1 H-MRS).Methods A total of 69 patients with NAFLD diagnosed by CT were divided into two groups by the ratio of liver spleen in CT (group A,low-grade fatty liver,n=32;group B,moderate and severe fatty liver,n=37).All patients underwent MR 3D-DIXION,1H MRS and IN/OPP examination twice within 1 week after CT examination.Fat fraction(FF) was evaluated and analyzed respectively.Results Pearson linear correlation analysis showed significant positive correlation between group A and group B about FF3D-DIXON,FFMRs and FFIN/OPP.Area under ROC curve (AUC) was different:AUCsYNTHESIZE＞AUCMRs＞AUCIN/OPP＞AUC3D-DIXON.Kruskal-Wallis H revealed significant difference among the three sequences no matter in group A or group B (P＜0.05).FF values of the three sequences (FF3D-DIXON,FFMRs and FFIN/OPP) were negatively correlated with CT liver/ spleen ratio in each group.The cut-off value of 3D-DIXON,IN/OPP and 1 H-MRS to diagnose presence or absence of fatty liver was 4.9 3 9 ％,5.2 8 4 ％ and 10.4 6 0 ％.Conclusion FF values measured by MRI methods might significantly vary based on different sequences.1 H-MRS is also the main method for quantitative assessment of fatty liver,3D-DIXON showed significant advantage in quantifying moderate and severe fatty liver.

Lung cancer is one of the most serious malignant tumors threatening human health. Early detection and accurate risk assessment are crucial for improving the survival rate and prognosis of lung cancer patients. In this review paper, the research progress in early-stage lung cancer risk assessment methods based on predictive factors and medical imaging was summarized. The results indicated that by utilizing more diverse machine learning algorithms and larger-scale datasets, independent risk prediction factors can be further mined to achieve an accurate assessment of individual lung cancer risk.

Objective To evaluate the value of 1.5T MRI in differential diagnosis of circular enhanced brain lesions.Methods A total of 120 patients,admitted to our hospital from April 2009 to March 2013 and confirmed by clinical pathology,including 40 of brain metastasis,40 of astrocytoma and 40 of brain abscess,were chosen in our study; The characteristic signs of MRI findings of these patients were retrospectively analyzed.Results There were significant differences in the location,number,size,morphology,ring structure,multiple mural nodules,and smooth outside wall of the lesions between patients with brain metastasis and astrocytoma.There were significant differences in the number,size,ring structure,ring wall thickness,and smooth outside wall of the lesions,T2-weighted imaging (T2WI) low signal ring sign,hydraulic fluid level sign,lesion center diffusion weighted imaging (DWI) hyperintensity and apparent diffusion coefficient (ADC) map low signal between patients with brain metastasis and abscess.There were significant differences in the location,shape and multi locular lesions between patients with astrocytoma and abscess.The signs indicating brain metastasis included multi lesions,grade Ⅲ peripheral edema,small lesions,lesions located at the juncture of gray and white matter,with displaying percentages of 79.5％,70.3％,67.5％,48.5％ and 47.9％.The signs indicating astrocytoma included multi locular lesion,lesions located at the white matter,multi mural nodule,rough outside wall and irregular in shape,with displaying percentages of 85.7％,78.0 ％,77.8％,65.4 ％ and 51.9％.The signs indicating abscess included daughter ring sign,DWI hyperintensity and ADC map low signal,circular wall notch sign,hype-intensity rim on T2WI and fluid-fluid level,with displaying percentages of 100％,100％,100％,92.6％ and 82.4％.Conclusion The 1.5T MRI can displaythe characteristic signs of brain metastases,gliomas and brain abscess,which enjoys high value in differential diagnosis.

Objective To investigate the impact of prescribed adaptive statistical iterative reconstruction V (Pre-ASiR-V) on imaging quality and radiation dosage of pure ground glass nodules (pGGN) in chest phantom,in order to obtain the optimal level of Pre-ASiR-V.Methods CT scanning for a chest phantom containing 4 artificial pGGNs was performed with Revolution CT,and the Pre-ASiR-V level was set as 0,20％,40％,60％,80％ and 100％ group,respectively.The mean noise,effective dose (ED) and the subjective scores of pGGN imaging were recorded and compared.Results The mean noise of groups (Pre-ASiR-V 0,20％,40％,60％,80％ and 100％) was (17.93±2.20)HU,(17.30±3.68)HU,(18.20± 3.44)HU,(18.80±0.20)HU,(19.87±2.56)HU and (15.90±4.56)HU,respectively (F=0.568,P=0.723).ED of these groups was 7.40 mSv,5.16 mSv,3.36 mSv,1.97 mSv,0.97 mSv and 0.33 mSv,respectively.Compared with imaging of PreASiR-V 0,the reduction percentage of ED was 30.27％,54.59％,73.38％,86.89％ and 95.54％,respectively.The subjective score of the image quality evaluated by the 2 observers had high agreement (Kappa=0.778,P=0.003),and all the scores were greater than 3.The subjective score of Pre-ASiR-V 80％ and 100％ group was slightly lower than those in other groups.Conclusion Different Pre-ASiR-V level slightly impacts the noise of high-resolution CT images of chest phantom,while reduces radiation dosage significantly.Pre-ASiR-V level of 60％ is the optimal protocol.

Objective:To design a lung function prediction method that combines transfer learning and multimodal feature fusion, aiming to improve the accuracy of lung function prediction in patients with idiopathic pulmonary fibrosis (IPF).Methods:CT images and clinical text data were reprocessed, and an adaptive module was designed to find the most suitable lung function attenuation function for IPF patients. The feature extraction module was utilized to comprehensively extract features. The feature extraction module comprises three sub-modules, including CT feature extraction, clinical text feature extraction, and lung function feature extraction. A multimodal feature prediction network was used to comprehensively evaluate the attenuation of lung function. The pre-trained model was fine-tuned to improve the predictive performance of the model.Results:Based on the OSIC pulmonary fibrosis progression competition dataset, it is found through the adaptive module that the linear attenuation hypothesis is more in line with the trend of pulmonary function decline in patients. Different modal data prediction experiments show that the model incorporating clinical text features has better predictive ability than the model using only CT images. The model combining CT images, clinical text features, and lung function features have optimal predictive results. The lung function prediction method combining transfer learning and multimodal feature fusion has modified version of the Laplace log likelihood (LLLm) of ?6.706 5, root mean squared error (RMSE) of 184.5, and mean absolute error (MAE) of 146.2, which outperforms other methods in terms of performance. The pre-trained model has higher prediction accuracy compared to the zero base training model.Conclusions:The lung function prediction method designed by combining transfer learning and multimodal feature fusion can effectively predict the lung function status of IPF patients at different weeks, providing important support for patient health management and disease diagnosis.

Rectal cancer is one of the most common gastrointestinal malignancies in China. Accurate and reasonable assessment of the preoperative staging of rectal cancer can significantly enhance treatment outcomes and improve patient prognosis. Magnetic resonance imaging is the technique of choice for local staging of rectal cancer and has significant advantages in the diagnosis of rectal primary tumors (T) and peri-intestinal lymph nodes (N). In this review paper, the research ideas and progress of traditional radiomics and deep learning methods for preoperative TN staging prediction of rectal cancer were reviewed around multimodal magnetic resonance images, with the aim of providing new ideas for realizing fully automated TN staging algorithms for rectal cancer.

Objective To explore the image quality improvement of portal vein computed tomography venography(CTV)using CE-Boost technique with low dose and low contrast media usage.Methods A total of 50 patients with suspected portal vein disorders who underwent abdominal non-contrast and biphasic contrast-enhanced CT scans using the Canon 320-row CT machine were retrospectively selected.Images of portal venous phase(PVP)were postprocessed with CE-Boost technique.The CT values of each area,standard deviation(SD)values of the paraspinal muscles,volume CT dose index(CTDIvol),and dose length product(DLP)before and after CE-Boost were measured and recorded.The signal-to-noise ratio(SNR),contrast-to-noise ratio(CNR),effective dose(ED)of each blood vessel before and after CE-Boost were calculated.Subjective image quality was analyzed by two senior radiologists using a five-point scale in a double-blinded method.Statistical analysis was performed using paired t-test and Wilcoxon test.Results The CT values of each area with CE-Boost images were significantly higher than those without CE-Boost images(P＜0.001).SNR and CNR of each blood vessel with CE-Boost images were significantly higher than those without CE-Boost images(P＜0.001).The subjective scores of both images were above 3 points,which met the requirements of clinical diagnosis with good consistency(Kappa=0.772,0.697).The median subjective scores of images with CE-Boost were 5(5,5),which were significantly higher than those without CE-Boost images 5(5,4),(P=0.002).CTDIvol,DLP and ED were(1.85±1.12)mGy,(94.66±44.68)mGy·cm and(1.42±0.67)mSv,respectively.Conclusion CE-Boost technique can significantly improve the image quality of portal vein CTV with low dose and low contrast media usage.

Objective To investigate a novel stroke recurrence risk prediction model,which utilized radiomics machine learning methods and specifically combined carotid computed tomography angiography(CT A)with the Essen stroke risk score(ESRS).Methods A total of 136 patients who underwent carotid CT A were analyzed retrospectively.The features of carotid plaque were extrac-ted by machine learning to construct a radiomics feature model,as well as combined with ESRS.Based on clinical outcomes at one-year follow-up,the stroke recurrence risk prediction model was constructed using the logistic regression(LR)machine learning model.To construct an effective and robust model,the dataset was divided into a training set and a validation set in a ratio of 7∶3.The performance of this model was evaluated using area under the curve(AUC)of receiver operating characteristic(ROC)curve,sensi-tivity and specificity.Results The model had strong predictive value.In the training set,AUC,sensitivity and specificity of this model were 0.903,0.796 and 0.761,respectively.In the validation set,AUC,sensitivity and specificity of this model were 0.869,0.667 and 0.850,respectively.Conclusion The stroke recurrence risk prediction model constructed based on radiomics analysis of carotid plaque characteristics in carotid CTA,in combination with the ESRS,can provide reliable predictions for stroke prognosis.

Objective To evaluate the application value of percutaneous angle positioning system based on augmented reality in improving the accuracy of liver puncture.Methods A canine liver with an embedded marking ring was used as the target for puncture.A skilled physician with over 5 years of experience in liver puncture and a novice physician with limited experience in liver puncture separately performed liver puncture using either the augmented reality-based percutaneous angle localization system(navigation)or CT-guided technique alone(non-navigation).The corresponding puncturing data of non-navigation skilled group(Group A),non-navigation non-skilled group(Group B),navigation skilled group(Group C),and navigation non-skilled group(Group D)were obtained.The differences in the evaluation indicators,including the number of CT scans,number of needle adjustment,time spent for operation,and distance of error,between Group A and Group B,between Group C and Group D,between Group A and Group C,and between Group B and Group D,were analyzed.Results Statistically significant differences in the number of CT scans,number of needle adjustment,time spent for operation,and distance of error existed between Group A and Group B,between Group A and Group C,and between Group B and Group D(all P＜0.0 5),while the differences in the number of CT scans,number of needle adjustment,time spent for operation,and distance of error between Group C and Group D were not statistically significant(all P＞0.05)Conclusion In performing liver puncture,the use of percutaneous angle localization system can reduce the number of CT scans,number of needle adjustment,time spent for operation and distance of error,and improve the puncture accuracy as well,which provides a basis for the clinical utilization of this system and the employment of this system-guided puncture technology in primary hospitals.(J Intervent Radiol,2024,33:507-511)