CT Evaluation of Long-Term Changes in Common Bile Duct Diameter after Cholecystectomy.

Purpose: The present study aimed to investigate the frequency and extent of compensatory common bile duct (CBD) dilatation after cholecystectomy, assess the time between cholecystectomy and CBD dilatation, and identify potentially useful CT findings suggestive of obstructive CBD dilatation. Materials and Methods: This retrospective study included 121 patients without biliary obstruction who underwent multiple CT scans before and after cholecystectomy at a single center between 2009 and 2011. The maximum short-axis diameters of the CBD and intrahepatic duct (IHD) were measured on each CT scan. In addition, the clinical and CT findings of 11 patients who were initially excluded from the study because of CBD stones or periampullary tumors were examined to identify distinguishing features between obstructive and non-obstructive CBD dilatation after cholecystectomy. Results: The mean (standard deviation) short-axis maximum CBD diameter of 121 patients was 5.6 (± 1.9) mm in the axial plane before cholecystectomy but increased to 7.9 (± 2.6) mm after cholecystectomy (p < 0.001). Of the 106 patients with a pre-cholecystectomy axial CBD diameter of < 8 mm, 39 (36.8%) showed CBD dilatation of ≥ 8 mm after cholecystectomy. Six of the 17 patients with longterm (> 2 years) serial follow-up CT scans (35.3%) eventually showed a significant (> 1.5-fold) increase in the axial CBD diameter, all within two years after cholecystectomy. Of the 121 patients without obstruction or related symptoms, only one patient (0.1%) showed IHD dilatation > 3 mm after cholecystectomy. In contrast, all 11 patients with CBD obstruction had abdominal pain and abnormal laboratory indices, and 81.8% (9/11) had significant dilatation of the IHD and CBD. Conclusion: Compensatory non-obstructive CBD dilatation commonly occurs after cholecystectomy to a similar extent as obstructive dilatation. However, the presence of relevant symptoms, significant IHD dilatation, or further CBD dilatation 2-3 years after cholecystectomy should raise suspicion of CBD obstruction.

Distinct Characteristics and Changes in Liver Function of Patients with Hepatocellular Carcinoma Treated with Atezolizumab Plus Bevacizumab for more than 1 Year.

Purpose: Since 2020, Atezolizumab plus bevacizumab (Ate/Bev) has been the standard first-line therapy for unresectable hepatocellular carcinoma (HCC), but long-term treatment studies are limited. This study evaluated the clinical characteristics and effects of Ate/Bev for over 1 year. Materials and Methods: This study included patients with unresectable HCC treated with first-line Ate/Bev between May 2020 and April 2022. Those receiving Ate/Bev for 1 year or more were classified as the long-term treatment group. Results: Of 246 patients, 69 (28.0%) were in the long-term treatment group, which comprised more proportions of intrahepatic tumor burden <25%, ECOG 0, and a lower proportion of portal vein tumor thrombosis than the short-term treatment group. The long-term treatment group had a higher incidence of atezolizumab-related thyroid dysfunction (31.9% vs. 10.7%, p<0.001; median time to onset [mTTO]: 2.8 months), dermatologic toxicity (29.0% vs. 14.7%, p=0.017; mTTO: 3.3 months), bevacizumab-related hypertension (44.9% vs. 22.0%, p=0.001; mTTO: 4.2 months), and proteinuria (69.6% vs. 38.4%, p<0.001; mTTO: 6.8 months), compared to the short-term treatment group. Regarding liver function in the long-term treatment group, patients initially classified as Child-Pugh class A decreased from 87% to 75.4%, and albumin-bilirubin grade 1 decreased from 68.1% to 50.7% after 1 year of treatment. Conclusion: The Ate/Bev long-term treatment group had a lower intrahepatic tumor burden, less portal vein tumor thrombosis, and better performance status and liver function at baseline. Atezolizumab-related immunological adverse events emerged relatively early in treatment compared to the bevacizumab-related. Additionally, some patients demonstrated liver function deterioration during long-term Ate/Bev treatment.

Additional nodules detected using EOB-MRI in patients with resectable single hepatocellular carcinoma: an implication for active treatment strategy.

BACKGROUND/AIM: Gadolinium-ethoxybenzyl-diethylenetriamine pentaacetic acid-enhanced magnetic resonance imaging (EOBMRI) further enhances the identification of additional hepatic nodules compared with computed tomography (CT) alone; however, the optimal treatment for such additional nodules remains unclear. We investigated the long-term oncological effect of aggressive treatment strategies for additional lesions identified using EOB-MRI in patients with hepatocellular carcinoma (HCC). METHODS: Data from 522 patients diagnosed with solitary HCC using CT between January 2008 and December 2012 were retrospectively reviewed. Propensity score-matched (PSM) analysis was used to compare the oncologic outcomes between patients with solitary HCC and those with additional nodules on EOB-MRI after aggressive treatment (resection or radiofrequency ablation [RFA]). RESULTS: Among the 383 patients included, 59 had additional nodules identified using EOB-MRI. Compared with patients with solitary HCC, those with additional nodules on EOB-MRI had elevated total bilirubin, aspartate transaminase, and alanine transaminase; had a lower platelet count, higher MELD score, and highly associated with liver cirrhosis (P<0.05). Regarding long-term outcomes, 59 patients with solitary HCC and those with additional nodules after PSM were compared. Disease-free survival (DFS) and overall survival (OS) were comparable between the two groups (DFS, 60.4 vs. 44.3 months, P=0.071; OS, 82.8 vs. 84.8 months, P=0.986). CONCLUSION: The aggressive treatment approach, either resection or RFA, for patients with additional nodules identified on EOBMRI was associated with long-term survival comparable with that for solitary HCC. However, further studies are required to confirm these findings.

Suspicious findings observed retrospectively on CT imaging performed before the diagnosis of pancreatic cancer.

Background: Few studies have focused on computed tomography findings before a pancreatic cancer diagnosis. We aimed to investigate the prediagnostic computed tomography findings of patients who had undergone computed tomography within the prediagnostic period of their pancreatic cancer diagnosis. Methods: Between January 2008 and December 2019, 27 patients who underwent contrast-enhanced abdominal or chest computed tomography including the pancreas within 1 year of a pancreatic cancer diagnosis were enrolled in this retrospective study. The prediagnostic computed tomography imaging findings were divided into pancreatic parenchyma and pancreatic duct findings. Results: All patients underwent computed tomography for reasons unrelated to pancreatic cancer. The pancreatic parenchyma and ducts showed normal findings in seven patients and abnormal findings in 20 patients. Hypoattenuating mass-like lesions were detected in nine patients with a median size of 1.2 cm. Six patients had focal pancreatic duct dilatations, and two patients had distal parenchymal atrophy. In three patients, two of these findings were found simultaneously. Taken together, 14 (51.9%) of 27 patients had findings suggestive of pancreatic cancer in prediagnostic computed tomography. Conclusions: In contrast-enhanced computed tomography performed for other purposes, attention should be paid to the presence of a hypoattenuating mass, focal pancreatic duct dilatation, or distal parenchymal atrophy of the pancreas. These features may be clues for an early diagnosis of pancreatic cancer.

Fully automated radiomics-based machine learning models for multiclass classification of single brain tumors: Glioblastoma, lymphoma, and metastasis.

BACKGROUND AND PURPOSE: To investigate the diagnostic performance of fully automated radiomics-based models for multiclass classification of a single enhancing brain tumor among glioblastoma, central nervous system lymphoma, and metastasis. MATERIALS AND METHODS: The training and test sets were comprised of 538 cases (300 glioblastomas, 73 lymphomas, and 165 metastases) and 169 cases (101 glioblastomas, 29 lymphomas, and 39 metastases), respectively. After fully automated segmentation, radiomic features were extracted. Three conventional machine learning classifiers, including least absolute shrinkage and selection operator (LASSO), adaptive boosting (Adaboost), and support vector machine with the linear kernel (SVC), combined with one of four feature selection methods, including forward sequential feature selection, F score, mutual information, and LASSO, were trained. Additionally, one ensemble classifier based on the three classifiers was used. The diagnostic performance of the optimized models was tested in the test set using the accuracy, F1-macro score, and the area under the receiver operating characteristic curve (AUCROC). RESULTS: The best performance was achieved when the LASSO was used as a feature selection method. In the test set, the best performance was achieved by the ensemble classifier, showing an accuracy of 76.3% (95% CI, 70.0-82.7), a F1-macro score of 0.704, and an AUCROC of 0.878. CONCLUSION: Our fully automated radiomics-based models for multiclass classification might be useful for differential diagnosis of a single enhancing brain tumor with a good diagnostic performance and generalizability.

Comparison of the performance of conventional and spectral-based tagged stool cleansing algorithms at CT colonography.

OBJECTIVES: To compare the performance of conventional versus spectral-based electronic stool cleansing for iodine-tagged CT colonography (CTC) using a dual-layer spectral detector scanner. METHODS: We retrospectively evaluated iodine contrast stool-tagged CTC scans of 30 consecutive patients (mean age: 69 ± 8 years) undergoing colorectal cancer screening obtained on a dual-layer spectral detector CT scanner. One reader identified locations of electronic cleansing artifacts (n = 229) on conventional and spectral cleansed images. Three additional independent readers evaluated these locations using a conventional cleansing algorithm (Intellispace Portal) and two experimental spectral cleansing algorithms (i.e., fully transparent and translucent tagged stool). For each cleansed image set, readers recorded the severity of over- and under-cleansing artifacts on a 5-point Likert scale (0 = none to 4 = severe) and readability compared to uncleansed images. Wilcoxon's signed-rank tests were used to assess artifact severity, type, and readability (worse, unchanged, or better). RESULTS: Compared with conventional cleansing (66% score ≥ 2), the severity of overall cleansing artifacts was lower in transparent (60% score ≥ 2, p = 0.011) and translucent (50% score ≥ 2, p < 0.001) spectral cleansing. Under-cleansing artifact severity was lower in transparent (49% score ≥ 2, p < 0.001) and translucent (39% score ≥ 2, p < 0.001) spectral cleansing compared with conventional cleansing (60% score ≥ 2). Over-cleansing artifact severity was worse in transparent (17% score ≥ 2, p < 0.001) and translucent (14% score ≥ 2, p = 0.023) spectral cleansing compared with conventional cleansing (9% score ≥ 2). Overall readability was significantly improved in transparent (p < 0.001) and translucent (p < 0.001) spectral cleansing compared with conventional cleansing. CONCLUSIONS: Spectral cleansing provided more robust electronic stool cleansing of iodine-tagged stool at CTC than conventional cleansing. KEY POINTS: • Spectral-based electronic cleansing of tagged stool at CT colonography provides higher quality images with less perception of artifacts than does conventional cleansing. • Spectral-based electronic cleansing could potentially advance minimally cathartic approach for CT colonography. Further clinical trials are warranted.

Positive Versus Neutral Oral Contrast Material for Detection of Malignant Deposits in Intraabdominal Nonsolid Organs on CT.

BACKGROUND. Data are limited regarding utility of positive oral contrast material for peritoneal tumor detection on CT. OBJECTIVE. The purpose of this article is to compare positive versus neutral oral contrast material for detection of malignant deposits in nonsolid intraabdominal organs on CT. METHODS. This retrospective study included 265 patients (133 men, 132 women; median age, 61 years) who underwent an abdominopelvic CT examination in which the report did not suggest presence of malignant deposits and a subsequent CT examination within 6 months in which the report indicated at least one unequivocal malignant deposit. Examinations used positive (iohexol; n = 100) or neutral (water; n = 165) oral agents. A radiologist reviewed images to assess whether the deposits were visible (despite clinical reports indicating no deposits) on unblinded comparison with the follow-up examinations; identified deposits were assigned to one of seven intraabdominal compartments. The radiologist also assessed adequacy of bowel filling with oral contrast material. Two additional radiologists independently reviewed examinations in blinded fashion for malignant deposits. NPV was assessed of clinical CT reports and blinded retrospective readings for detection of malignant deposits visible on unblinded comparison with follow-up examinations. RESULTS. Unblinded review identified malignant deposits in 58.1% (154/265) of examinations. In per-patient analysis of clinical reports, NPV for malignant deposits was higher for examinations with adequate bowel filling with positive oral contrast material (65.8% [25/38]) than for examinations with inadequate bowel filling with positive oral contrast material (45.2% [28/62], p = .07) or with neutral oral contrast material regardless of bowel filling adequacy (35.2% [58/165], p = .002). In per-compartment analysis of blinded interpretations, NPV was higher for examinations with adequate and inadequate bowel filling with positive oral contrast material than for examinations with neutral oral contrast regardless of bowel filling adequacy (reader 1: 94.7% [234/247] and 92.5% [382/413] vs 88.3% [947/1072], both p = .045; reader 2: 93.1% [228/245] and 91.6% [361/394] vs 85.9% [939/1093], both p = .01). CONCLUSION. CT has suboptimal NPV for malignant deposits in intraabdominal nonsolid organs. Compared with neutral material, positive oral contrast material improves detection, particularly with adequate bowel filling. CLINICAL IMPACT. Optimization of bowel preparation for oncologic CT may help avoid potentially severe clinical consequences of missed malignant deposits.

Development and validation of a prognostic model for early triage of patients diagnosed with COVID-19.

We developed a tool to guide decision-making for early triage of COVID-19 patients based on a predicted prognosis, using a Korean national cohort of 5,596 patients, and validated the developed tool with an external cohort of 445 patients treated in a single institution. Predictors chosen for our model were older age, male sex, subjective fever, dyspnea, altered consciousness, temperature ≥ 37.5 °C, heart rate ≥ 100 bpm, systolic blood pressure ≥ 160 mmHg, diabetes mellitus, heart disease, chronic kidney disease, cancer, dementia, anemia, leukocytosis, lymphocytopenia, and thrombocytopenia. In the external validation, when age, sex, symptoms, and underlying disease were used as predictors, the AUC used as an evaluation metric for our model's performance was 0.850 in predicting whether a patient will require at least oxygen therapy and 0.833 in predicting whether a patient will need critical care or die from COVID-19. The AUCs improved to 0.871 and 0.864, respectively, when additional information on vital signs and blood test results were also used. In contrast, the protocols currently recommended in Korea showed AUCs less than 0.75. An application for calculating the prognostic score in COVID-19 patients based on the results of this study is presented on our website ( https://nhimc.shinyapps.io/ih-psc/ ), where the results of the validation ongoing in our institution are periodically updated.

Radiomics-based prediction of multiple gene alteration incorporating mutual genetic information in glioblastoma and grade 4 astrocytoma, IDH-mutant.

PURPOSE: In glioma, molecular alterations are closely associated with disease prognosis. This study aimed to develop a radiomics-based multiple gene prediction model incorporating mutual information of each genetic alteration in glioblastoma and grade 4 astrocytoma, IDH-mutant. METHODS: From December 2014 through January 2020, we enrolled 418 patients with pathologically confirmed glioblastoma (based on the 2016 WHO classification). All selected patients had preoperative MRI and isocitrate dehydrogenase (IDH) mutation, O-6-methylguanine-DNA methyltransferase (MGMT) promoter methylation, epidermal growth factor receptor amplification, and alpha-thalassemia/mental retardation syndrome X-linked (ATRX) loss status. Patients were randomly split into training and test sets (7:3 ratio). Enhancing tumor and peritumoral T2-hyperintensity were auto-segmented, and 660 radiomics features were extracted. We built binary relevance (BR) and ensemble classifier chain (ECC) models for multi-label classification and compared their performance. In the classifier chain, we calculated the mean absolute Shapley value of input features. RESULTS: The micro-averaged area under the curves (AUCs) for the test set were 0.804 and 0.842 in BR and ECC models, respectively. IDH mutation status was predicted with the highest AUCs of 0.964 (BR) and 0.967 (ECC). The ECC model showed higher AUCs than the BR model for ATRX (0.822 vs. 0.775) and MGMT promoter methylation (0.761 vs. 0.653) predictions. The mean absolute Shapley values suggested that predicted outcomes from the prior classifiers were important for better subsequent predictions along the classifier chains. CONCLUSION: We built a radiomics-based multiple gene prediction chained model that incorporates mutual information of each genetic alteration in glioblastoma and grade 4 astrocytoma, IDH-mutant and performs better than a simple bundle of binary classifiers using prior classifiers' prediction probability.

Radiomics machine learning study with a small sample size: Single random training-test set split may lead to unreliable results.

This study aims to determine how randomly splitting a dataset into training and test sets affects the estimated performance of a machine learning model and its gap from the test performance under different conditions, using real-world brain tumor radiomics data. We conducted two classification tasks of different difficulty levels with magnetic resonance imaging (MRI) radiomics features: (1) "Simple" task, glioblastomas [n = 109] vs. brain metastasis [n = 58] and (2) "difficult" task, low- [n = 163] vs. high-grade [n = 95] meningiomas. Additionally, two undersampled datasets were created by randomly sampling 50% from these datasets. We performed random training-test set splitting for each dataset repeatedly to create 1,000 different training-test set pairs. For each dataset pair, the least absolute shrinkage and selection operator model was trained and evaluated using various validation methods in the training set, and tested in the test set, using the area under the curve (AUC) as an evaluation metric. The AUCs in training and testing varied among different training-test set pairs, especially with the undersampled datasets and the difficult task. The mean (±standard deviation) AUC difference between training and testing was 0.039 (±0.032) for the simple task without undersampling and 0.092 (±0.071) for the difficult task with undersampling. In a training-test set pair with the difficult task without undersampling, for example, the AUC was high in training but much lower in testing (0.882 and 0.667, respectively); in another dataset pair with the same task, however, the AUC was low in training but much higher in testing (0.709 and 0.911, respectively). When the AUC discrepancy between training and test, or generalization gap, was large, none of the validation methods helped sufficiently reduce the generalization gap. Our results suggest that machine learning after a single random training-test set split may lead to unreliable results in radiomics studies especially with small sample sizes.

Identification of magnetic resonance imaging features for the prediction of molecular profiles of newly diagnosed glioblastoma.

PURPOSE: We predicted molecular profiles in newly diagnosed glioblastoma patients using magnetic resonance (MR) imaging features and explored the associations between imaging features and major molecular alterations. METHODS: This retrospective study included patients with newly diagnosed glioblastoma and available next-generation sequencing results. From preoperative MR imaging, Visually AcceSAble Rembrandt Images (VASARI) features, volumetric parameters, and apparent diffusion coefficient (ADC) values were obtained. First, univariate random forest was performed to identify gene abnormalities that could be predicted by imaging features with high accuracy and stability. Next, multivariate random forest was trained to predict the selected genes in the discovery cohort and was validated in the external cohort. Univariable logistic regression was performed to further explore the associations between imaging features and genes. RESULTS: Univariate random forest identified nine genes predicted by imaging features, with high accuracy and stability. The multivariate random forest model showed excellent performance in predicting IDH and PTPN11 mutations in the discovery cohort, which were validated in the external validation cohorts (areas under the receiver operator characteristic curve [AUCs] of 0.855 for IDH and 0.88 for PTPN11). ATRX loss and EGFR mutation were predicted with AUCs of 0.753 and 0.739, respectively, whereas PTEN could not be reliably predicted. Based on univariable logistic regression analyses, IDH, ATRX, and TP53 were clustered according to their shared imaging features, whereas EGFR and CDKN2A/B were clustered in the opposite direction. CONCLUSIONS: MR imaging features are related to specific molecular alterations and can be used to predict molecular profiles in patients with newly diagnosed glioblastoma.

Prediction of the risk of developing hepatocellular carcinoma in health screening examinees: a Korean cohort study.

BACKGROUND: Almost all Koreans are covered by mandatory national health insurance and are required to undergo health screening at least once every 2 years. We aimed to develop a machine learning model to predict the risk of developing hepatocellular carcinoma (HCC) based on the screening results and insurance claim data. METHODS: The National Health Insurance Service-National Health Screening database was used for this study (NHIS-2020-2-146). Our study cohort consisted of 417,346 health screening examinees between 2004 and 2007 without cancer history, which was split into training and test cohorts by the examination date, before or after 2005. Robust predictors were selected using Cox proportional hazard regression with 1000 different bootstrapped datasets. Random forest and extreme gradient boosting algorithms were used to develop a prediction model for the 9-year risk of HCC development after screening. After optimizing a prediction model via cross validation in the training cohort, the model was validated in the test cohort. RESULTS: Of the total examinees, 0.5% (1799/331,694) and 0.4% (390/85,652) in the training cohort and the test cohort were diagnosed with HCC, respectively. Of the selected predictors, older age, male sex, obesity, abnormal liver function tests, the family history of chronic liver disease, and underlying chronic liver disease, chronic hepatitis virus or human immunodeficiency virus infection, and diabetes mellitus were associated with increased risk, whereas higher income, elevated total cholesterol, and underlying dyslipidemia or schizophrenic/delusional disorders were associated with decreased risk of HCC development (p < 0.001). In the test, our model showed good discrimination and calibration. The C-index, AUC, and Brier skill score were 0.857, 0.873, and 0.078, respectively. CONCLUSIONS: Machine learning-based model could be used to predict the risk of HCC development based on the health screening examination results and claim data.

Robust performance of deep learning for automatic detection and segmentation of brain metastases using three-dimensional black-blood and three-dimensional gradient echo imaging.

OBJECTIVES: To evaluate whether a deep learning (DL) model using both three-dimensional (3D) black-blood (BB) imaging and 3D gradient echo (GRE) imaging may improve the detection and segmentation performance of brain metastases compared to that using only 3D GRE imaging. METHODS: A total of 188 patients with brain metastases (917 lesions) who underwent a brain metastasis MRI protocol including contrast-enhanced 3D BB and 3D GRE were included in the training set. DL models based on 3D U-net were constructed. The models were validated in the test set consisting of 45 patients with brain metastases (203 lesions) and 49 patients without brain metastases. RESULTS: The combined 3D BB and 3D GRE model yielded better performance than the 3D GRE model (sensitivities of 93.1% vs 76.8%, p < 0.001), and this effect was significantly stronger in subgroups with small metastases (p interaction < 0.001). For metastases < 3 mm, ≥ 3 mm and < 10 mm, and ≥ 10 mm, the sensitivities were 82.4%, 93.2%, and 100%, respectively. The combined 3D BB and 3D GRE model showed a false-positive per case of 0.59 in the test set. The combined 3D BB and 3D GRE model showed a Dice coefficient of 0.822, while 3D GRE model showed a lower Dice coefficient of 0.756. CONCLUSIONS: The combined 3D BB and 3D GRE DL model may improve the detection and segmentation performance of brain metastases, especially in detecting small metastases. KEY POINTS: • The combined 3D BB and 3D GRE model yielded better performance for the detection of brain metastases than the 3D GRE model (p < 0.001), with sensitivities of 93.1% and 76.8%, respectively. • The combined 3D BB and 3D GRE model showed a false-positive rate per case of 0.59 in the test set. • The combined 3D BB and 3D GRE model showed a Dice coefficient of 0.822, while the 3D GRE model showed a lower Dice coefficient of 0.756.

Development and validation of a nomogram to predict pulmonary function and the presence of chronic obstructive pulmonary disease in a Korean population.

BACKGROUND: Early suspicion followed by assessing lung function with spirometry could decrease the underdiagnosis of chronic obstructive pulmonary disease (COPD) in primary care. We aimed to develop a nomogram to predict the FEV1/FVC ratio and the presence of COPD. METHODS: We retrospectively reviewed the data of 4241 adult patients who underwent spirometry between 2013 and 2019. By linear regression analysis, variables associated with FEV1/FVC were identified in the training cohort (n = 2969). Using the variables as predictors, a nomogram was created to predict the FEV1/FVC ratio and validated in the test cohort (n = 1272). RESULTS: Older age (ß coefficient [95% CI], - 0.153 [- 0.183, - 0.122]), male sex (- 1.904 [- 2.749, - 1.056]), current or past smoking history (- 3.324 [- 4.200, - 2.453]), and the presence of dyspnea (- 2.453 [- 3.612, - 1.291]) or overweight (0.894 [0.191, 1.598]) were significantly associated with the FEV1/FVC ratio. In the final testing, the developed nomogram showed a mean absolute error of 8.2% between the predicted and actual FEV1/FVC ratios. The overall performance was best when FEV1/FVC < 70% was used as a diagnostic criterion for COPD; the sensitivity, specificity, and balanced accuracy were 82.3%, 68.6%, and 75.5%, respectively. CONCLUSION: The developed nomogram could be used to identify potential patients at risk of COPD who may need further evaluation, especially in the primary care setting where spirometry is not available.

Diffusion tensor and postcontrast T1-weighted imaging radiomics to differentiate the epidermal growth factor receptor mutation status of brain metastases from non-small cell lung cancer.

PURPOSE: To assess whether the radiomic features of diffusion tensor imaging (DTI) and conventional postcontrast T1-weighted (T1C) images can differentiate the epidermal growth factor receptor (EGFR) mutation status in brain metastases from non-small cell lung cancer (NSCLC). METHODS: A total of 99 brain metastases in 51 patients who underwent surgery or biopsy with underlying NSCLC and known EGFR mutation statuses (57 from EGFR wild type, 42 from EGFR mutant) were allocated to the training (57 lesions in 31 patients) and test (42 lesions in 20 patients) sets. Radiomic features (n = 526) were extracted from preoperative MR images including T1C and DTI. Radiomics classifiers were constructed by combinations of five feature selectors and four machine learning algorithms. The trained classifiers were validated on the test set, and the classifier performance was assessed by determining the area under the curve (AUC). RESULTS: EGFR mutation status showed an overall discordance rate of 12% between the primary tumors and corresponding brain metastases. The best performing classifier was a combination of the tree-based feature selection and linear discriminant algorithm and 5 features were selected (1 from ADC, 2 from fractional anisotropy, and 2 from T1C images), resulting in an AUC, accuracy, sensitivity, and specificity of 0.73, 78.6%, 81.3%, and 76.9% in the test set, respectively. CONCLUSIONS: Radiomics classifiers integrating multiparametric MRI parameters may have potential in differentiating the EGFR mutation status in brain metastases from NSCLC.

Machine learning prediction for mortality of patients diagnosed with COVID-19: a nationwide Korean cohort study.

The rapid spread of COVID-19 has resulted in the shortage of medical resources, which necessitates accurate prognosis prediction to triage patients effectively. This study used the nationwide cohort of South Korea to develop a machine learning model to predict prognosis based on sociodemographic and medical information. Of 10,237 COVID-19 patients, 228 (2.2%) died, 7772 (75.9%) recovered, and 2237 (21.9%) were still in isolation or being treated at the last follow-up (April 16, 2020). The Cox proportional hazards regression analysis revealed that age > 70, male sex, moderate or severe disability, the presence of symptoms, nursing home residence, and comorbidities of diabetes mellitus (DM), chronic lung disease, or asthma were significantly associated with increased risk of mortality (p ≤ 0.047). For machine learning, the least absolute shrinkage and selection operator (LASSO), linear support vector machine (SVM), SVM with radial basis function kernel, random forest (RF), and k-nearest neighbors were tested. In prediction of mortality, LASSO and linear SVM demonstrated high sensitivities (90.7% [95% confidence interval: 83.3, 97.3] and 92.0% [85.9, 98.1], respectively) and specificities (91.4% [90.3, 92.5] and 91.8%, [90.7, 92.9], respectively) while maintaining high specificities > 90%, as well as high area under the receiver operating characteristics curves (0.963 [0.946, 0.979] and 0.962 [0.945, 0.979], respectively). The most significant predictors for LASSO included old age and preexisting DM or cancer; for RF they were old age, infection route (cluster infection or infection from personal contact), and underlying hypertension. The proposed prediction model may be helpful for the quick triage of patients without having to wait for the results of additional tests such as laboratory or radiologic studies, during a pandemic when limited medical resources must be wisely allocated without hesitation.

Estimation of Fractional Extracellular Space at CT for Predicting Chemotherapy Response and Survival in Pancreatic Ductal Adenocarcinoma.

OBJECTIVE. The purpose of this study was to investigate the association between primary pancreatic ductal adenocarcinoma fractional extracellular space (fECS) estimated from pretreatment CT and tumor response to chemotherapy and patient outcome. MATERIALS AND METHODS. A database search identified the records of patients with locally advanced or metastatic pancreatic ductal adenocarcinoma treated with systemic therapies who had undergone pretreatment CT that included both unenhanced and equilibrium phase images. An ROI was placed on the primary tumor and aorta, and the tumor fECS was calculated as follows: (tumor attenuation in the equilibrium phase - tumor attenuation in the unenhanced phase) / (aortic attenuation in the equilibrium phase - aortic attenuation in the unenhanced phase) × (1 - hematocrit). Response to therapy was assessed in subsequent CT examinations according to the Response Evaluation Criteria in Solid Tumors version 1.1. Relevant clinical variables, including carbohydrate antigen 19-9 level, chemotherapy regimen, and survival were recorded. Multivariate analyses were performed to determine the predictors of treatment response and patient survival. RESULTS. The median primary tumor fECS was 0.41 (range, 0.02-0.69). When dichotomized to high (> 0.41) versus low fECS (≤ 0.41) values, a larger proportion of patients with high tumor fECS values achieved disease control after chemotherapy than did those with low tumor fECS values: full cohort, 27 of 30 versus 19 of 30 (p = 0.030); cohort with locally advanced disease, 23 of 24 versus 10 of 15 (p = 0.024). The mean progression-free survival among patients with high primary tumor fECS values was significantly longer than that among those with low fECS values (191 versus 115 days, p = < 0.0001). Primary tumor fECS was an independent predictor of progression-free survival (p = 0.003) in multivariate analysis. CONCLUSION. High primary tumor fECS value estimated from staging CT was associated with chemotherapy response and progression-free survival of patients with advanced pancreatic ductal adenocarcinoma.

Robust performance of deep learning for distinguishing glioblastoma from single brain metastasis using radiomic features: model development and validation.

We evaluated the diagnostic performance and generalizability of traditional machine learning and deep learning models for distinguishing glioblastoma from single brain metastasis using radiomics. The training and external validation cohorts comprised 166 (109 glioblastomas and 57 metastases) and 82 (50 glioblastomas and 32 metastases) patients, respectively. Two-hundred-and-sixty-five radiomic features were extracted from semiautomatically segmented regions on contrast-enhancing and peritumoral T2 hyperintense masks and used as input data. For each of a deep neural network (DNN) and seven traditional machine learning classifiers combined with one of five feature selection methods, hyperparameters were optimized through tenfold cross-validation in the training cohort. The diagnostic performance of the optimized models and two neuroradiologists was tested in the validation cohort for distinguishing glioblastoma from metastasis. In the external validation, DNN showed the highest diagnostic performance, with an area under receiver operating characteristic curve (AUC), sensitivity, specificity, and accuracy of 0.956 (95% confidence interval [CI], 0.918-0.990), 90.6% (95% CI, 80.5-100), 88.0% (95% CI, 79.0-97.0), and 89.0% (95% CI, 82.3-95.8), respectively, compared to the best-performing traditional machine learning model (adaptive boosting combined with tree-based feature selection; AUC, 0.890 (95% CI, 0.823-0.947)) and human readers (AUC, 0.774 [95% CI, 0.685-0.852] and 0.904 [95% CI, 0.852-0.951]). The results demonstrated deep learning using radiomic features can be useful for distinguishing glioblastoma from metastasis with good generalizability.

Improved Sensitivity and Reader Confidence in CT Colonography Using Dual-Layer Spectral CT: A Phantom Study.

Background Limited cathartic preparations for CT colonography with fecal tagging can improve patient comfort but may result in nondiagnostic examinations from poorly tagged stool. Dual-energy CT may overcome this limitation by improving the conspicuity of the contrast agent, but more data are needed. Purpose To investigate whether dual-energy CT improves polyp detection in CT colonography compared with conventional CT at different fecal tagging levels in vitro. Materials and Methods In this HIPAA-compliant study, between December 2017 and August 2019, a colon phantom 30 cm in diameter containing 60 polyps of different shapes (spherical, ellipsoid, flat) and size groups (5-9 mm, 11-15 mm) was constructed and serially filled with simulated feces tagged with four different iodine concentrations (1.26, 2.45, 4.88, and 21.00 mg of iodine per milliliter), then it was scanned with dual-energy CT with and without an outer fat ring to simulate large body size (total diameter, 42 cm). Two readers independently reviewed conventional 120-kVp CT and 40-keV monoenergetic dual-energy CT images to record the presence of polyps and confidence (three-point scale.) Generalized estimating equations were used for sensitivity comparisons between conventional CT and dual-energy CT, and a Wilcoxon signed-rank test was used for reader confidence. Results Dual-energy CT had higher overall sensitivity for polyp detection than conventional CT (58.8%; 95% confidence interval [CI]: 49.7%, 67.3%; 564 of 960 polyps vs 42.1%; 95% CI: 32.1%, 52.8%; 404 of 960 polyps; P < .001), including with the fat ring (48% and 31%, P < .001). Reader confidence improved with dual-energy CT compared with conventional images on all tagging levels (P < .001). Interrater agreement was substantial (κ = 0.74; 95% CI: 0.70, 0.77). Conclusion Compared with conventional 120-kVp CT, dual-energy CT improved polyp detection and reader confidence in a dedicated dual-energy CT colonography phantom, especially with suboptimal fecal tagging. © RSNA, 2020.

Diagnostic performance of Liver Imaging Reporting and Data System in patients at risk of both hepatocellular carcinoma and metastasis.

OBJECTIVE: The purpose of this study was to evaluate the diagnostic performance of the Liver Imaging Reporting and Data System (LI-RADS) in patients with both chronic liver disease and a history of extrahepatic malignancy. MATERIALS AND METHODS: This retrospective study included 59 hepatocellular carcinomas (HCCs) and 45 metastases pathologically confirmed between 2008 and 2017 in 104 patients with chronic liver disease (cirrhosis or chronic hepatitis B) and a history of extrahepatic malignancy. Two radiologists blinded to the final diagnosis independently reviewed MRI (95 patients) or CT (9 patients) images, and their consensus data were used to calculate the diagnostic performance of LI-RADS categories. Serum tumor markers, tumor multiplicity, and suspected metastatic lymph nodes were also evaluated. RESULTS: The sensitivity, specificity, and accuracy of LR-5 for diagnosing HCC were 69% (95% confidence intervals [CI] 56-81), 98% (95% CI 88-99), and 82% (95% CI 73-89), respectively, and those of LR-M for diagnosing metastasis were 89% (95% CI 76-96), 88% (95% CI 77-95), and 88% (95% CI 81-94), respectively. Elevation of serum carcinoembryonic antigen (P = 0.01) or carbohydrate antigen 19-9 levels (P = 0.02) and tumor multiplicity (P = 0.004) were more frequently observed in metastasis than in HCC. Three of four metastases categorized as LR-4 or LR-5 were smaller than 2 cm. CONCLUSIONS: The LI-RADS provides high specificity (98%) for differentiating HCC from metastases in patients with both chronic liver disease and a history of extrahepatic malignancy.