Development and Validation of a Deep Learning Model for Brain Tumor Diagnosis and Classification Using Magnetic Resonance Imaging.

Importance: Deep learning may be able to use patient magnetic resonance imaging (MRI) data to aid in brain tumor classification and diagnosis. Objective: To develop and clinically validate a deep learning system for automated identification and classification of 18 types of brain tumors from patient MRI data. Design, Setting, and Participants: This diagnostic study was conducted using MRI data collected between 2000 and 2019 from 37 871 patients. A deep learning system for segmentation and classification of 18 types of intracranial tumors based on T1- and T2-weighted images and T2 contrast MRI sequences was developed and tested. The diagnostic accuracy of the system was tested using 1 internal and 3 external independent data sets. The clinical value of the system was assessed by comparing the tumor diagnostic accuracy of neuroradiologists with vs without assistance of the proposed system using a separate internal test data set. Data were analyzed from March 2019 through February 2020. Main Outcomes and Measures: Changes in neuroradiologist clinical diagnostic accuracy in brain MRI scans with vs without the deep learning system were evaluated. Results: A deep learning system was trained among 37 871 patients (mean [SD] age, 41.6 [11.4] years; 18 519 women [48.9%]). It achieved a mean area under the receiver operating characteristic curve of 0.92 (95% CI, 0.84-0.99) on 1339 patients from 4 centers' data sets in diagnosis and classification of 18 types of tumors. Higher outcomes were found compared with neuroradiologists for accuracy and sensitivity and similar outcomes for specificity (for 300 patients in the Tiantan Hospital test data set: accuracy, 73.3% [95% CI, 67.7%-77.7%] vs 60.9% [95% CI, 46.8%-75.1%]; sensitivity, 88.9% [95% CI, 85.3%-92.4%] vs 53.4% [95% CI, 41.8%-64.9%]; and specificity, 96.3% [95% CI, 94.2%-98.4%] vs 97.9%; [95% CI, 97.3%-98.5%]). With the assistance of the deep learning system, the mean accuracy of neuroradiologists among 1166 patients increased by 12.0 percentage points, from 63.5% (95% CI, 60.7%-66.2%) without assistance to 75.5% (95% CI, 73.0%-77.9%) with assistance. Conclusions and Relevance: These findings suggest that deep learning system-based automated diagnosis may be associated with improved classification and diagnosis of intracranial tumors from MRI data among neuroradiologists.

Automatic quantitative evaluation of normal pancreas based on deep learning in a Chinese adult population.

OBJECTIVE: To develop a 3D U-Net-based model for the automatic segmentation of the pancreas using the diameters, volume, and density of normal pancreases among Chinese adults. METHODS: A total of 2778 pancreas images (dataset 1) were retrospectively collected and randomly divided into training (n = 2252), validation (n = 245), and test (n = 281) datasets. The segmentation model for the pancreas was constructed through cascaded application of two 3D U-Net networks. The segmentation efficiency for the pancreas was evaluated by the Dice similarity coefficient (DSC). Another dataset of 3189 normal pancreas CT images (dataset 2) was obtained for external validation, including 1063 non-contrast images, 1063 arterial phase images, and 1063 portal venous phase images. The pancreas segmentation in dataset 2 was assessed objectively and manually revised by two radiologists. Then, the pancreatic volume, diameters, and average CT value for each phase of pancreas images in dataset 2 were calculated. The relationships between pancreas volume and age, sex, height, and weight were analyzed. RESULTS: In dataset 1, a mean DSC of 0.94 for the test dataset was achieved. In dataset 2, the objective assessment yielded a 90% satisfaction rate for the automatic segmentation of the pancreas as external validation. The diameters of the pancreas were 43.71-44.28 mm, 67.40-68.15 mm, and 114.53-117.06 mm, respectively. The average pancreatic volume was 63,969.06-65,247.75 mm3, which was greatest at the age of 18-38 and then decreased to a minimum at the age of 69-85. The CT value of the pancreas also decreased with age, from a maximum value of 38.87 ± 9.70 HU to a minimum of 27.72 ± 10.85 HU. CONCLUSION: The pancreas segmentation tool based on deep learning can segment the pancreas on CT images and measure its normal diameter, volume, and CT value accurately and effectively.

Optimization of combined microwave and hot air drying technology for purple cabbage by Response Surface Methodology (RSM).

Purple cabbage is one of the world's most widely consumed vegetables with high nutritional values containing the antioxidants and anti-inflammatory activity of anthocyanins, vitamins, and minerals. But due to rapid postharvest quality decay, novel processing techniques including drying are required. In order to determine the conditions of combined microwave and hot air drying for purple cabbage, factors affecting the drying process including microwave density, hot air temperature, and the dry base water content at conversion point were investigated using the anthocyanin content, DPPH antioxidant capacity, chewiness, △E, rehydration ratio, and average drying rate as responses. The combined drying conditions were optimized considering three independent variables at three different levels by response surface methodology. The results showed that the processing parameters of purple cabbage with combined microwave and hot air drying technology were microwave density at 2.5 W/g, moisture content of conversion point at 4.0 g/g, and hot air temperature at 55°C. Under these conditions, the anthocyanin content, DPPH antioxidant capacity, chewiness, △E, rehydration ratio, average drying rate, and overall score of the dried purple cabbage were 175.87 mg/100 g, 87.59%, 4,521.468 g, 26.5, 4.3, 0.76 g/min, and 0.785, respectively. Therefore, combined microwave and hot air drying technology is an effective, suitable method for drying purple cabbage.

Preoperative prediction of clinically relevant postoperative pancreatic fistula after pancreaticoduodenectomy.

OBJECTIVES: To develop a radiomics model and a combined model for preoperative prediction of clinically relevant postoperative pancreatic fistula (CR-POPF) in patients undergoing pancreaticoduodenectomy and to compare the predictive performance of the two models with the traditional Fistula Risk Score system. METHODS: A total of 250 patients who underwent pancreaticoduodenectomy (PD) with preoperative computed tomography (CT) were divided into a training set (n = 175) and validation set (n = 75). The pancreatic area was automatically segmented on the portal venous phase CT images using a 3D U-Net segmentation model. A radiomics model was developed using radiomics features extracted from the volume of interest (VOI) and a combined model was developed using radiomics features, demographic information and radiological features. The FRS was also used to predict POPF. The predictive performance of the prediction models was assessed using receiver operating characteristic (ROC) curves, calibration curves and decision curve analysis (DCA). RESULTS: Eleven and 18 features were extracted for the radiomics model and combined model, respectively. The combined model showed excellent predictive value, with an AUC of 0.871 (95 %CI 0.816,0.926) and 0.869 (95 %CI 0.779,0.958) in the training cohort and validation cohort, respectively. Calibration curves and DCA showed that the combined model outperformed the traditional FRS system and radiomics model. CONCLUSION: The combined model exhibited excellent predictive performance and outperformed the traditional FRS system and radiomics model in the preoperative prediction of CR-POPF.

Accuracy of imaging markers on noncontrast computed tomography in predicting intracerebral hemorrhage expansion.

Objectives Hematoma expansion (HE) is an important factor of unfavorable outcome in patients with intracerebral hemorrhage (ICH). Imaging markers on noncontrast computed tomography (NCCT) provide increasing value in the prediction of HE due to fast and easy-to-use advantages; however, the accuracy of NCCT-based prediction of intracerebral HE remains unclear. We aimed to investigate the predictive accuracy of NCCT markers for the evaluation of HE using a well-characterized ICH cohort. Methods We retrospectively analyzed 414 patients with spontaneous ICH, who underwent baseline CT within 6 h after symptom onset and follow-up CT within 24 h after ICH. Hematoma volumes were measured on baseline and follow-up CT images, and imaging features that predicted HE were analyzed. The test characteristics for the NCCT predictors were calculated. Results Of the 414 patients investigated, 63 presented blend sign, 45 showed black hole sign, 36 had island sign and 34 had swirl sign. In the 414 patients, 88 presented HE, the incidence was 21.26%. Of the 88 patients with HE, 22 presented blend sign, 11 showed black hole sign, 8 had swirl sign and 7 had island sign. The blend sign showed highest sensitivity (25.00%) and swirl sign showed the highest specificity (92.02%) among the four predictors. We noted excellent interobserver agreement for the identification of HE. Conclusion The four NCCT markers can predict HE with limited sensitivity, high specificity and good accuracy. This may be useful for prompt identification of patients at high risk of active bleeding, and prevention of over-treatment associated with HE. Abbreviations HE, hematoma expansion; ICH, intracerebral hemorrhage; NCCT, noncontrast computed tomography.

Changing Ischemic Lesion Patterns and Hemodynamics of the Posterior Cerebral Artery in Moyamoya Disease.

OBJECTIVE: The aim of this study was to determine how hemodynamics of the posterior cerebral artery (PCA) are associated with cerebral ischemic lesions in moyamoya disease (MMD). METHODS: Thirty-six patients with ischemic MMD (Suzuki grade IV-V) were retrospectively analyzed. Hemodynamic parameters of the PCA were measured by transcranial color-coded sonography. We classified the range of ischemic lesions into 3 grades and perfusion levels into 3 grades according to computed tomography (CT) results. PCA steno-occlusion and leptomeningeal collaterals were confirmed by digital subtraction angiography. Ultrasonographic parameters in the PCA were compared with these radiographic findings. RESULTS: The velocity in the involved PCA (mean flow velocity [MFV] median, 42.00 [range, 34.50-58.00] cm/s) was significantly lower than that in the normal PCA (MFV median, 95.00 [range, 76.50-119.50] cm/s) (P < .001). The velocity in the PCA increased significantly as the leptomeningeal collateral stage advanced (MFV stage 1: median, 38.50 [range, 29.75-63.50] cm/s; stage 2: median, 55.00 [range, 44.00-96.00] cm/s; stage 3: median, 94.00 [range, 54.00-118.25] cm/s; stage 4: median, 85.50 [range, 70.50-117.75] cm/s, respectively) (P < .05). Decreased PCA velocities were associated with a larger ischemic area on CT (P ≤ .001). PCA velocity had no correlation with CT perfusion level of the temporal and frontal lobes. PCA velocity had significant correlations with perfusion level in the occipital (P < .001) and parietal lobes (P < .05). CONCLUSIONS: Our results suggest ischemic lesion patterns (as demonstrated on CT imaging) are associated with PCA velocity measurements in the advanced stage of MMD. Thus, monitoring PCA velocity in patients with advanced MMD may provide additional information to assist in managing these patients.

Supratentorial intracerebral cerebellar liponeurocytoma: A case report and literature review.

RATIONALE: Cerebellar liponeurocytoma is a rare tumor of the central nervous system (CNS) characterized by low proliferation but high likelihood of recurrence. Because of its rarity and the paucity of systematic follow-up, the biological behaviors and clinical features of this tumor are still poorly understood. We herein reported a case of cerebellar liponeurocytoma originating in the cerebral hemisphere. PATIENT CONCERNS: A 11-year-old male with intermittent headache, nausea, and vomiting. The first computed tomography revealed a large mass in the right cerebral hemisphere. He was transferred to our institution for neurosurgical treatment. DIAGNOSIS: Magnetic resonance imaging showed a large cystic-solid mass in the right frontal lobe with obvious contrast enhancement. Histopathological examinations showed sheets of isomorphic small neoplastic cells with clear cytoplasm and focal lipomatous differentiation. On immunohistochemistry, tumor cells were positive for synaptophysin, microtubule-associated protein 2, and neuronal nuclei antigen. INTERVENTIONS: The patient was performed a right fronto-parietal craniotomy, and gross total resection of the tumor was achieved without adjuvant therapy. OUTCOMES: No clinical or neuroradiological evidence of recurrence or residual of the tumor was found 6 years and 2 months after initial surgery. LESSONS: Cerebellar liponeurocytoma developing in supratentorial cerebral hemisphere was first reported in the present study. The radiological and histopathological features may be useful in differentiating this rare tumor from other tumors at similar locations. A change in the nomenclature of cerebellar liponeurocytomas should be considered in future World Health Organization (WHO) classifications.

Clinical features and prognosis for intraventricular liponeurocytoma.

Cerebellar liponeurocytoma is a rare central nervous system tumor, we investigate its biological behaviors and clinical prognosis to improve the understanding of this tumor. We retrospectively analyzed the clinical, radiological and histopathological findings as well as follow-up data of two patients with intraventricular liponeurocytomas in Beijing Tiantan Hospital between July 2000 and July 2016. The main clinical manifestations of the two patients were headache. The supratentorial intraventricular liponeurocytoma appeared as isodense to slight hyperdense on CT scan and heterogeneous intensity on T1-weighted imaging (T1WI) and T2-weighted imaging (T2WI). The plaque-like hypodense on CT images and hyperintensity on T1WI resembling fat could be seen inside the tumor. The liponeurocytoma located in the fourth ventricle showed isointensity on T1 and T2WI as well as slight enhancement on contrast. Two patients accepted gross total resection of tumors. Two intraventricular tumors demonstrated similarly histopathological features, such as isomorphic small tumor cells with clear cytoplasm, sheets of monomorphic round cells and focal lipomatous differentiation. In addition, expression of synaptophysin, neuron specific enolase, microtubule-associated protein 2 and S-100 were found. No radiological or clinical evidence of recurrence of the tumors was observed in their follow-up surveys. In conclusion, intraventricular liponeurocytoma has a favorable clinical course, radiological features may be useful in the diagnosis of this rare tumor before surgery.

Etiology and surgical treatment of hilar bile duct stricture.

OBJECTIVE: To improve the surgical effects of hilar duct stricture. METHODS: The clinical data of 76 patients with hilar bile duct stricture treated at our hospital from 1990 to 2000 were analyzed. The diagnosis was determined by triad signs of cholangitis, increase of ALP and gamma-GGT levels, dilation of intrahepatic and extrahepatic bile ducts confirmed by ultrasonography (US), computed tomography (CT), percutaneous transhepatic cholangiography (PTC) or endoscopic retrograde cholangiopancreatography (ERCP). The location of stricture was divided according to the Bismuth classification standard. RESULTS: Among the 76 patients, 46 (60.5%) suffered from injurious stricture, including 13% of Bismuth type I, 39% of type II, 19.4% of type III, and 28.2% of type IV. Inflammatory stricture was found in 28 patients, locating in the left hepatic duct (LHD) 46.4% (13/28), the right hepatic duct (RHD) 35.7% (10/28), and the common hepatic duct (CHD) 17.9% (5/28), respectively. The percentages of patients with stricture due to Mirizzi's syndrome, bile duct cyst, and sclerosing cholangitis were 9.2%, 3.9% and 2.6%, respectively. Bile duct repair procedures included biliary reconstruction with pedicled umbilical vein graft for 9.2% of the patients, and proximal cholangiojejunostomy combined with LHD and RHD plasticity for 92.2%. Seventy of the 76 patients were followed up for 2-10 years, and the excellent outcome rate was 94.7%. CONCLUSIONS: Injurious stricture is the major type of hilar bile duct stricture. Inflammatory stricture is mainly composed of RHD. Hilar bile duct stricture should be treated surgically according to various etiological features and technical principles of biliary repair.