Objective To establish a machine learning radiomics model that can accurately predict MRI enhancement patterns of glioma based on T2 fluid attenuated inversion recovery(T2-FLAIR)images for optimizing the workflow of magnetic resonance imaging(MRI)examinations of glioma patients.Methods We retrospectively collected preoperative MR T2-FLAIR images from 385 patients with pathologically confirmed glioma,who were divided into enhancing and non-enhancing groups according to the enhancement pattern.Predictive radiomics models were established using Gaussian Process,Linear Regression,Linear Regression-Least absolute shrinkage and selection operator,Support Vector Machine,Linear Discriminant Analysis or Naive Bayes as the classifiers in the training cohort(n=201)and tested both in the internal(n=85)and external validation cohorts(n=99).The receiver-operating characteristic curve was used to assess the predictive performance of the models.Results The predictive model constructed based on 15 radiomics features using Gaussian Process as the classifier had the best predictive performance in both the training cohort and the internal validation cohort,with areas under the curve(AUC)of 0.88(95%CI:0.81-0.94)and 0.80(95%CI:0.71-0.88),respectively.In the external validation cohort,the model showed an AUC of 0.81(95%CI:0.71-0.90)with sensitivity,specificity,positive predictive value and negative predictive value of 0.98,0.61,0.76 and 0.96,respectively.Conclusion The T2-FLAIR-based machine learning radiomics model can accurately predict the enhancement pattern of gliomas on MRI.

Objective To establish a machine learning radiomics model that can accurately predict MRI enhancement patterns of glioma based on T2 fluid attenuated inversion recovery(T2-FLAIR)images for optimizing the workflow of magnetic resonance imaging(MRI)examinations of glioma patients.Methods We retrospectively collected preoperative MR T2-FLAIR images from 385 patients with pathologically confirmed glioma,who were divided into enhancing and non-enhancing groups according to the enhancement pattern.Predictive radiomics models were established using Gaussian Process,Linear Regression,Linear Regression-Least absolute shrinkage and selection operator,Support Vector Machine,Linear Discriminant Analysis or Naive Bayes as the classifiers in the training cohort(n=201)and tested both in the internal(n=85)and external validation cohorts(n=99).The receiver-operating characteristic curve was used to assess the predictive performance of the models.Results The predictive model constructed based on 15 radiomics features using Gaussian Process as the classifier had the best predictive performance in both the training cohort and the internal validation cohort,with areas under the curve(AUC)of 0.88(95%CI:0.81-0.94)and 0.80(95%CI:0.71-0.88),respectively.In the external validation cohort,the model showed an AUC of 0.81(95%CI:0.71-0.90)with sensitivity,specificity,positive predictive value and negative predictive value of 0.98,0.61,0.76 and 0.96,respectively.Conclusion The T2-FLAIR-based machine learning radiomics model can accurately predict the enhancement pattern of gliomas on MRI.

Objective To evaluate the performance of magnetic resonance imaging(MRI)multi-sequence feature imputation and fusion mutual model based on sequence deletion in differentiating high-grade glioma(HGG)from low-grade glioma(LGG).Methods We retrospectively collected multi-sequence MR images from 305 glioma patients,including 189 HGG patients and 116 LGG patients.The region of interest(ROI)of T1-weighted images(T1WI),T2-weighted images(T2WI),T2 fluid attenuated inversion recovery(T2_FLAIR)and post-contrast enhancement T1WI(CE_T1WI)were delineated to extract the radiomics features.A mutual-aid model of MRI multi-sequence feature imputation and fusion based on sequence deletion was used for imputation and fusion of the feature matrix with missing data.The discriminative ability of the model was evaluated using 5-fold cross-validation method and by assessing the accuracy,balanced accuracy,area under the ROC curve(AUC),specificity,and sensitivity.The proposed model was quantitatively compared with other non-holonomic multimodal classification models for discriminating HGG and LGG.Class separability experiments were performed on the latent features learned by the proposed feature imputation and fusion methods to observe the classification effect of the samples in two-dimensional plane.Convergence experiments were used to verify the feasibility of the model.Results For differentiation of HGG from LGG with a missing rate of 10%,the proposed model achieved accuracy,balanced accuracy,AUC,specificity,and sensitivity of 0.777,0.768,0.826,0.754 and 0.780,respectively.The fused latent features showed excellent performance in the class separability experiment,and the algorithm could be iterated to convergence with superior classification performance over other methods at the missing rates of 30%and 50%.Conclusion The proposed model has excellent performance in classification task of HGG and LGG and outperforms other non-holonomic multimodal classification models,demonstrating its potential for efficient processing of non-holonomic multimodal data.

Objective To evaluate the performance of magnetic resonance imaging(MRI)multi-sequence feature imputation and fusion mutual model based on sequence deletion in differentiating high-grade glioma(HGG)from low-grade glioma(LGG).Methods We retrospectively collected multi-sequence MR images from 305 glioma patients,including 189 HGG patients and 116 LGG patients.The region of interest(ROI)of T1-weighted images(T1WI),T2-weighted images(T2WI),T2 fluid attenuated inversion recovery(T2_FLAIR)and post-contrast enhancement T1WI(CE_T1WI)were delineated to extract the radiomics features.A mutual-aid model of MRI multi-sequence feature imputation and fusion based on sequence deletion was used for imputation and fusion of the feature matrix with missing data.The discriminative ability of the model was evaluated using 5-fold cross-validation method and by assessing the accuracy,balanced accuracy,area under the ROC curve(AUC),specificity,and sensitivity.The proposed model was quantitatively compared with other non-holonomic multimodal classification models for discriminating HGG and LGG.Class separability experiments were performed on the latent features learned by the proposed feature imputation and fusion methods to observe the classification effect of the samples in two-dimensional plane.Convergence experiments were used to verify the feasibility of the model.Results For differentiation of HGG from LGG with a missing rate of 10%,the proposed model achieved accuracy,balanced accuracy,AUC,specificity,and sensitivity of 0.777,0.768,0.826,0.754 and 0.780,respectively.The fused latent features showed excellent performance in the class separability experiment,and the algorithm could be iterated to convergence with superior classification performance over other methods at the missing rates of 30%and 50%.Conclusion The proposed model has excellent performance in classification task of HGG and LGG and outperforms other non-holonomic multimodal classification models,demonstrating its potential for efficient processing of non-holonomic multimodal data.

Objective:To investigate the reproducibility of inflow-based vascular-space-occupancy (iVASO)-MRI in quantifying skeletal muscle perfusion and its potential clinical value in patients with dermatomyositis (DM).Methods:Totally 15 consecutive patients with DM and 15 healthy volunteers were enrolled in this prospective study from December 2018 to April 2019 at Nanfang Hospital, Southern Medical University. All subjects underwent T 1WI, short-tau inversion recovery (STIR) T 2WI, and iVASO-MRI of thigh on a 3.0 T MR scanner. According to findings on T 1WI and STIR T 2WI, the muscles were divided into normal, unaffected, edematous and atrophic or fat-infiltrated groups. Maximum arteriolar muscular blood volume (MBVa_max) and mean MBVa (MBVa_mean) of these 4 groups of muscles were obtained by 2 radiologists independently. In order to evaluate the reproducibility of iVASO, the repeated scan was performed 3 days later in 17 subjects (12 healthy volunteers and 5 DM patients), and the MBVa values were measured to calculate the intraclass correlation coefficients (ICC). The MBVa_max and MBVa_mean among the 4 groups were compared by using Kruskal-Wallis H test and the differences of each 2 groups was compare by using Mann-Whitney U test. Results:The ICC between the 2 observers was 0.95 and 0.96 for MBVa_max and MBVa_mean, respectively. The ICC between repeated tests was 0.87 and 0.89 for MBVa_max and MBVa_mean, respectively.There was significant difference among normal muscles, unaffected muscles, edematous muscles and atrophic or fat-infiltrated muscles ( P<0.001). Post hoc comparisons of MBVa_max and MBVa_mean showed that compared to normal muscles, unaffected muscles, edematous muscles and atrophic or fat-infiltrated muscles had a significant decrease of MBVa ( P<0.05). Unaffected muscles and edematous muscles showed no significant difference in terms of MBVa_max and MBVa_mean (both P=0.99), which were significantly higher than those of atrophic or fat-infiltrated muscles ( P<0.05). Conclusions:iVASO-MRIcan reliably quantify the MBVa of thigh muscular arteriolar, and it is potentially valuable in the diagnosis of DM.

Objective To investigate the CT and MRI features of supratentorial primitive neuroectodermal tumors (sPNET)in adults,and to analyze the reasons of misdiagnosis.Methods The CT and MRI features of 1 5 patients with sPNET confirmed by pathology were analyzed retrospective.13 of 15 patients were underwent plain and contrast enhanced MRI,4 patients were also performed CT scan and the other 2 patients were underwent CT plain scan only.Results Nine of 1 5 lesions were located in the temporal and/or frontal lobe,with clear margin in 12 lesions,and mild to moderate peritumoral was noted in most cases.The solid portion of the lesions were hyperdense on CT,isointense or slightly hypointense on T1WI,and isointense or hyperintense on T2WI compared to the gray matter.Calcification (2/6),necrosis or cystic degeneration(1 3/1 5),hemorrhage (8/1 5)and flow void signal (1 1/1 3)were seen.Twelve lesions showed significantly heterogeneous enhancement and 7 lesions showed irregular ring-enhancement.The solid parts in 5 lesions showed hyperintensity on diffusion-weighted imaging. Preoperatively,the lesions were misdiagnosed as glioma in 10 cases,ependymoma in 2 cases,meningioma in 1 case and germ cell tumor in 1 case, respectively;and 1 case was not diagnosed definitely.Conclusion Supratentorial PNET in adults has characteristic CT and MRI features,and the main reasons accounting for misdiagnosis are its extraordinary low incidence and that radiologists do no master its imaging features.

Objective To investigate the diagnostic value of six MRI characteristic features for diagnosing central neurocytoma (CN).Methods MRI data of 30 cases of CN and 68 cases of non-CN located in lateral ventricles were retrospectively analyzed.Six characteristic MRI features,including scalloping sign,broad-based attachment sign,soapbubble sign,peripheral cyst sign,fluid-fluid level sign and gemstone sign were scored based on a five-point scale.ROC curve was used to assess the diagnostic value of each MRI sign.Results The scalloping sign showed the highest area under the curve (AUC) value (0.82) among all 6 signs (all P＜0.05),followed by broad-based attachment,soap-bubble andperipheral cyst signs (AUC 0.73-0.75),higher than that of fluid-fluid level sign and gemstone sign (all P＜0.05).The scalloping sign exhibited the highest specificity (84.56 ％),followed by fluid-fluid level (77.94 ％),gemstone (74.26 ％) and peripheral cyst (70.34％) sign.The soap-bubble sign (83.89％) was the most sensitive sign,followed by broad-based attachment sign (76.11％) and peripheral cyst sign (75.00％).Conclusion The scalloping sign is the most valuable indicator for CN among six characteristic MRI features.

Objective To illustrate the CT,MRI and PET-CT features of spinal osteoblastoma,and provide an important reference for the preoperative diagnosis and assessment.Methods The imaging and clinical data in 16 cases of spinal osteoblastoma confirmed by surgery and pathology were analyzed retrospectively.Results Age of onset in 11 cases ranged from 10 to 29 years old,5 cases from 30 to 51 years old.Tumors were originated from the spine appendage in 13 cases,and from the vertebral bodies in 3 cases.Tumors presented as expansive bone destruction in 6 cases,and as osteolytic bone destruction in 10 cases.13 cases were diagnosed as benign diseases and 3 cases as invasive ones.Calcification or ossification was seen in all cases,with sclerotic margin of variable thickness and peripheral patchy high density of bone sclerosis.The soft tissue mass was found in 15 cases and intraspinal involvement was showed in 13 cases.The tumors showed isointensity or hypointensity on T 1WI,isointensity or hyper-intensity on T2WI in 13 cases,and showed heterogenous enhancement after contrast administration in 13 cases.PET-CT depicted the nodular or lobulated distribution of radioactive tracer with high radioactive concentration,with increment in standardized uptake value in 5 cases.Conclusion The typical imaging signs of spinal osteoblastoma were expansile,osteolytic bone destruction of the spine appendage,accompanying the soft tissue mass,with a speckled or nodular pattern of calcification or ossification within the lesions.Inhomogeneous signal intensity and inhomogeneous enhancement on MRI and nodular or lobulated high radioactive concentration for the tumors on PET-CT are found.

Objective To analyse and summarize the MRI characteristics of tumor‐like inflammatory demyelinating diseases (TIDD) .Methods MRI findings of 10 cases with pathologically proved TIDD were analyzed retrospectively ,all patients received plain and enhanced MRI scan .Results 3 cases showed multiple lesions in bilateral frontal and parietal lobes ,and 7 cases showed soli‐tary mass ,in which 2 lesions located in the occipital lobe ,2 in the basal ganglia ,2 in the second to fifth segment of cervical cord ,and 1 lesion located in the left frontal parietal lobe .The shape of 7 lesions were irregular ,2 were stripped ,while 1 lesion was oval .6 cases had clear boundary ,while 4 showed ambiguous .Solitary lesions were 1 .7 cm -5 .6 cm in diameter(mean 4 .0 cm ± 1 .48 cm) .8 cases manifested mild edema while 2 had no edema .On MRI plain scan ,lesions were hypointensity on T1WI and hyperintensity on T2WI in 8 cases ,isointensity on T1 WI and hyperintensity on T2 WI in 2 cases ,and all lesions were hyperintensity on FLAIR sequence .DWI were performed in 3 cases ,all lesions showed hyperintensity ,2 of them had patchy hypointensity signal .Hemorrhage was detected in 3 lesions ,cystic lesions or calcification were not found in all cases .On post‐contrast scan ,3 lesions showed ring like enhancement ,2 showed obviously heterogeneous enhancement ,2 had mild enhancement and 2 cases had no enhancement ,1 presented with open ring like enhancement .Conclusion TIDD shows some specific MRI features ,and MRI is helpful to the diagnosis of TIDD .

Objective To investigate MR findings and analyze the misdiagnosed cases of pituitary macroadenoma.Methods MR features of 1 90 patients of pituitary macroadenoma confirmed by operation and pathology were reviewed.Results 1 6 cases were mis-diagnosed as craniopharyngioma,chordoma,or meningioma.Among the sixteen cases,eight cases with cyst degeneration and hemor-rhage were misdiagnosed as craniopharyngioma;five cases with clival and sphenoid sinus destruction were misdiagnosed as chordo-mas;three cases with suprasellar and anterior cranial fossa extension were misdiagnosed as meningiomas.Conclusion Craniophar-yngioma,chordoma and meningioma should be considered in the differential diagnosis of atypical pituitary macroadenoma.The com-prehensive analysis should be based on a variety of signs.