Long term impact of transjugular intrahepatic portosystemic shunt (TIPS) creation on hepatic morphology.

PURPOSE: The purpose of this study was to evaluate long-term morphologic changes occurring in the liver after TIPS creation with correlation with hepatic function to gain insight on the physiologic impact of TIPS on the liver. METHODS: This retrospective study included patients who underwent TIPS creation between 2005 and 2022 and had contrasted CT or MRI studies prior to and between 1 and 2 years post procedure. Strict exclusion criteria were applied to avoid confounding. Parenchymal volume and vessel measurements were assessed on the pre- and post-TIPS CT or MRI and MELD scores calculated. RESULTS: Of 580 patients undergoing TIPS creation, 65 patients (mean age, 55 years; 36 males) had pre-TIPS and post-TIPS imaging meeting inclusion criteria at median 16.5 months. After TIPS, the mean MELD score increased (12.9 to 15.4; p = 0.008) and total liver volume decreased (1730 to 1432 mL; p < 0.001). However, the magnitude of volume change did not correlate with MELD change. Neither portosystemic gradient nor TIPS laterality correlated with total or lobar hepatic volume changes or MELD changes. The main portal vein diameter increased (15.0 to 18.7 mm; p < 0.001). Thrombosis of the hepatic vein used for TIPS creation resulted in a mean increase in MELD of +4.1 compared to -2.1 in patients who had a patent and normal hepatic vein (p = 0.007). CONCLUSIONS: Given lack of correlation between portosystemic gradient, hepatic atrophy, hepatic function, and TIPS laterality, the alterations in portal flow dynamics after TIPS may not be impactful to hepatic function. However, hepatic vein patency after TIPS correlated with improved hepatic function.

Tumor size and survival in intrahepatic cholangiocarcinoma treated with surgical resection or ablation.

OBJECTIVES: We performed a retrospective analysis within a national cancer registry on outcomes following resection or ablation for intrahepatic cholangiocarcinoma (iCCA). METHODS: The National Cancer Database was queried for patients with clinical stage I-III iCCA diagnosed during 2010-2018, who underwent resection or ablation. Overall survival (OS) was compared with Kaplan-Meier and multivariable Cox proportional hazards methods. RESULTS: Of 2140 patients, 1877 (87.7%) underwent resection and 263 (12.3%) underwent ablation, with median tumor sizes of 5.5 and 3 cm, respectively. Overall, resection was associated with greater median OS (41.2 months (95% confidence interval [95% CI]: 37.6-46.2) vs. 28 months (95% CI: 15.9-28.6) on univariable analysis (p < 0.0001). There was no significant difference on multivariable analysis (p = 0.42); however, there was a significant interaction between tumor size and management. On subgroup analysis of patients with tumors <3 cm, there was no difference in OS between resection versus ablation. However, ablation was associated with increased mortality for tumors ≥3 cm. CONCLUSION: Although resection is associated with improved OS for tumors ≥3 cm, we observed no difference in survival between management strategies for tumors < 3 cm. Ablation may be an alternative therapeutic strategy for small iCCA, particularly in patients at risk for high surgical morbidity.

Comparing Survival After Resection, Ablation, and Radiation in Small Intrahepatic Cholangiocarcinoma.

BACKGROUND: Hepatectomy is the cornerstone of curative-intent treatment for intrahepatic cholangiocarcinoma (ICC). However, in patients unable to be resected, data comparing efficacy of alternatives including thermal ablation and radiation therapy (RT) remain limited. Herein, we compared survival between resection and other liver-directed therapies for small ICC within a national cancer registry. PATIENTS AND METHODS: Patients with clinical stage I-III ICC < 3 cm diagnosed 2010-2018 who underwent resection, ablation, or RT were identified in the National Cancer Database. Overall survival (OS) was compared using Kaplan-Meier and multivariable Cox proportional hazards methods. RESULTS: Of 545 patients, 297 (54.5%) underwent resection, 114 (20.9%) ablation, and 134 (24.6%) RT. Median OS was similar between resection and ablation [50.5 months, 95% confidence interval (CI) 37.5-73.9; 39.5 months, 95% CI 28.7-58.4, p = 0.14], both exceeding that of RT (20.9 months, 95% CI 14.1-28.3). RT patients had high rates of stage III disease (10.4% RT vs. 1.8% ablation vs. 11.8% resection, p < 0.001), but the lowest rates of chemotherapy utilization (9.0% RT vs. 15.8% ablation vs. 38.7% resection, p < 0.001). In multivariable analysis, resection and ablation were associated with reduced mortality compared with RT [hazard ratio (HR) 0.44, 95% CI 0.33-0.58 and HR 0.53, 95% CI 0.38-0.75, p < 0.001, respectively]. CONCLUSION: Resection and ablation were associated with improved survival in patients with ICC < 3 cm compared with RT. Acknowledging confounders, anatomic constraints of ablation, limitations of available data, and need for prospective study, these results favor ablation in small ICC where resection is not feasible.

Utility of CBCT and AVD for intraprocedural diagnosis and treatment of lower GI bleeding.

BACKGROUND: Intraprocedural Cone Beam CT (CBCT) is assessed to examine if use improves diagnosis and embolization rates of acute lower GI bleed (LGIB) and if automatic vessel detection (AVD) software can identify feeding vessels (FV) for embolization. METHODS: Patients with inconclusive DSA findings had CBCT and retrospective analysis with AVD software (Innova 3100, GE Company, USA). Technical success was defined as the ability to detect a lower GIB site while clinical success was defined as successful embolization without evidence of rebleeding or death within 30 days. AVD technical success was defined by the ability to identify the FV on both CTA and CBCT upon independent review by 3 blinded IRs, who also assigned a degree of certainty on a 5-point Likert scale. RESULTS: 74 patients in total were treated for lower GI bleed of which 34 had indeterminate DSA. Of those, 10 patients received DSA only, of which 1 was super selective. 24 patients with GIB on pre-procedural CTA and inconclusive DSA underwent CBCT. Use of CBCT identified 9 bleeds not seen on DSA and an additional source artery in 1 case representing a 42% change in intraprocedural management as all findings were embolized. When a bleed could not be identified on CBCT, but the FV could be identified on CTA, the same suspected FV could be selected on AVD 62% of the time with an average certainty of 4.0. CONCLUSION: CBCT is useful in the intraprocedural detection of GIB when DSA is indeterminate. Furthermore, AVD software can feasibly be utilized to accurately identify FVs for empiric treatment when intraprocedural imaging is inconclusive. LEVEL OF EVIDENCE: Level III, therapeutic study.

Diagnostic accuracy of clinically reported adenomyosis on pelvic ultrasound and MRI compared to surgical pathology.

BACKGROUND: Adenomyosis is a common gynecologic pathology that relies on diagnostic imaging to guide treatment. Accuracy of both pelvic ultrasound and magnetic resonance imaging (MRI) when specifically evaluating for the presence of adenomyosis is high. However, the accuracy of reported rates in clinical practice is less well understood. PURPOSE: To demonstrate the accuracy in reporting of adenomyosis on pelvic ultrasound and MRI compared to histopathology in common clinical practice. BASIC PROCEDURES: An institutional database was searched for women with a pelvic ultrasound and a pelvic MRI with a subsequent hysterectomy. Findings were extracted from radiology and pathology reports, and the documented presence or absence of adenomyosis was recorded for each modality. Blinded radiologists viewed each imaging pair to directly evaluate for adenomyosis. MAIN FINDINGS: Compared to prior published data, imaging had lower accuracy in clinical practice when adenomyosis was not specifically evaluated for. For the finding of adenomyosis, pelvic ultrasound had a sensitivity of 10.9%, a specificity of 98.3%, positive predictive value (PPV) of 77.8%, negative predictive value (NPV) of 66.7%, an accuracy of 67.2%, and a diagnostic odds ratio (DOR) of 7. Pelvic MRI had a sensitivity of 29.7%, specificity of 85.3%, PPV of 52.8%, NPV of 68.8%, an accuracy of 65.6%, and DOR of 2.5. Overall accuracy of MRI improved when adenomyosis was directly evaluated for (82.4% vs 65.6%). PRINCIPLE CONCLUSIONS: Without direct communication to evaluate for adenomyosis, pelvic ultrasound and MRI may underestimate or misreport adenomyosis. Providers should be aware of these discrepancies when relying on radiology reports to guide treatment and potential interventions when diagnosing and managing adenomyosis.

Intravascular Ultrasound-Guided Transvenous Biopsy of Abdominal and Pelvic Targets Difficult to Access by Percutaneous Needle Biopsy: Technique and Initial Clinical Experience.

PURPOSE: To report initial clinical experience with intravascular ultrasound (US)-guided transvenous biopsy (TVB) for perivascular target lesions in the abdomen and pelvis using side-viewing phased-array intracardiac echocardiography catheters. MATERIALS AND METHODS: In this single-institution, retrospective study, 48 patients underwent 50 intravascular US-guided TVB procedures for targets close to the inferior vena cava or iliac veins deemed difficult to access by conventional percutaneous needle biopsy (PNB). In all procedures, side-viewing phased-array intracardiac echocardiography intravascular US catheters and transjugular liver biopsy sets were inserted through separate jugular or femoral vein access sheaths, and 18-gauge core needle biopsy specimens were obtained under real-time intravascular US guidance. Diagnostic yield, diagnostic accuracy, and complications were analyzed. RESULTS: Intravascular US-guided TVB was diagnostic of malignancy in 40 of 50 procedures for a diagnostic yield of 80%. There were 5 procedures in which biopsy was correctly negative for malignancy, with a per-procedure diagnostic accuracy of 90% (45/50). Among the 5 false negatives, 2 patients underwent repeat intravascular US-guided TVB, which was diagnostic of malignancy for a per-patient diagnostic accuracy of 94% (45/48). There were 1 (2%) mild, 2 (4%) moderate, and 1 (2%) severe adverse events, with 1 moderate severity adverse event (venous thrombosis) directly attributable to the intravascular US-guided TVB technique. CONCLUSIONS: Intravascular US-guided TVB performed on difficult-to-approach perivascular targets in the abdomen and pelvis resulted in a high diagnostic accuracy, similar to accepted thresholds for PNB. Complication rates may be slightly higher but should be weighed relative to the risks of difficult PNB, surgical biopsy, or clinical management without biopsy.

Proton pump inhibitor use is associated with increased rates of post-TIPS hepatic encephalopathy: Replication in an independent patient cohort.

PURPOSE: Proton pump inhibitor (PPI) use is a potential risk factor for hepatic encephalopathy (HE), but few studies have examined the effect on post-TIPS HE. The purpose of this study was to determine whether PPIs are associated with increased rates of post-TIPS HE in an independent patient cohort. MATERIALS AND METHODS: This single-institution retrospective study analyzed 86 patients (54 male, mean age 58.2) following TIPS from 1/1/2017 to 12/31/2019. Dates of PPI usage and episodes of new or worsening HE were recorded. Poisson regression with generalized estimating equations was used to test for association between PPI use and post-TIPS HE and to test for dose dependence. Post-TIPS HE was also analyzed using the Andersen-Gill survival model for recurrent events. RESULTS: There were 1.88 episodes of new or worsening post-TIPS HE per person-year among 35 patients on uninterrupted PPIs therapy, 1.95 on PPIs and 0.94 off PPIs among 35 patients on intermittent therapy, and 0.47 among 16 patients never on PPIs. PPI use was significantly associated with post-TIPS HE in both univariable (incidence rate ratio (IRR) = 2.62; CI = 1.41-4.84; p = 0.002) and multivariable (IRR = 2.31; CI = 1.37-3.89; p = 0.002) regression. Analysis of only those patients on PPIs showed increased rates of HE with higher doses (IRR = 1.17 per 10 mg omeprazole equivalent; CI = 1.04-1.33; p = 0.011). Recurrent events survival analysis supported the association between PPI use and HE in univariable (hazard ratio (HR) = 2.17; CI = 1.19-3.95; p = 0.011) and multivariable (HR = 1.87; CI = 1.12-3.13; p = 0.017) analysis. CONCLUSION: In an independent patient cohort PPI use was associated with increased rates of new or worsening post-TIPS HE.

Spinal Arterial Blood Supply Does Not Arise from the Bronchial Arteries: A Detailed Analysis of Angiographic Studies Performed for Hemoptysis.

PURPOSE: To determine the angiographic prevalence of spinal arteries originating directly from the bronchial arteries in the setting of embolization for hemoptysis. MATERIALS AND METHODS: Over a 14-year interval, 205 patients underwent angiography for hemoptysis. Twenty-five patients were excluded because their bronchial arteries were not visualized. The remaining 180 patients underwent a total of 254 angiographic procedures (range, 1-8 per patient). Images were reviewed jointly by 2 interventional radiologists with formal fellowship training in both peripheral and neurological interventional radiology. All catheterized arteries were evaluated for arterial contribution to the spinal cord. For patients with multiple studies, each unique artery was reported only once. Embolization was performed during at least 1 procedure in 158 patients (88%). Electronic record review was used to assess neurological sequelae after the procedure. RESULTS: One or 2 bronchial arteries originating from the aorta were identified in 57 patients (32%) on the right and in 75 patients (42%) on the left. Conjoined bronchial arteries were found in 76 patients (42%). Spinal arterial supply was absent in all. A total of 102 patients (57%) had at least 1 right and 11 patients (6%) at least 1 left intercostobronchial artery. Spinal arterial supply from the intercostal portion of an intercostobronchial artery was found in 6 patients (5 right, 1 left). Medical record review revealed no postprocedure symptoms referable to spinal cord injury in any patient. CONCLUSIONS: Spinal arterial supply does not originate directly from the bronchial artery but can originate from the intercostal portion of an intercostobronchial artery.

Comprehensive Word-Level Classification of Screening Mammography Reports Using a Neural Network Sequence Labeling Approach.

Radiology reports contain a large amount of potentially valuable unstructured data. Recently, neural networks have been employed to perform classification of radiology reports over a few classes at the document level. The success of neural networks in sequence-labeling problems such as named entity recognition and part of speech tagging suggests that they could be used to classify radiology report text with greater granularity. We employed a neural network architecture to comprehensively classify mammography report text at the word level using a sequence labeling approach. Two radiologists devised a comprehensive classification system for screening mammography reports. Each word in each report was manually categorized by a radiologist into one of 33 categories according to the classification system. Tagged words referencing the same finding were grouped into unique sets. We pre-labeled reports with a rule-based algorithm and then manually edited these annotations for 6705 screening mammography reports (25.1%, 66.8%, and 8.1% BI-RADS 0, 1, and 2, respectively). A combined convolutional and recurrent neural network model was used to label words in each sentence of the individual reports. A siamese recurrent neural network was then used to group findings into sets. Performance of the neural network-based method was compared to a rule-based algorithm and a conditional random field (CRF) model. Global accuracy (percentage of documents where all word tags were predicted correctly) and keyword accuracy (percentage of all words that were labeled correctly, excluding words tagged as unimportant) were calculated on an unseen 519 report test set. Two-tailed t tests were used to assess differences between algorithm performance, and p < 0.05 was used to determine statistical significance. The neural network-based approach showed significantly higher global accuracy compared to both the rule-based algorithm (88.3 vs 57.0%, p < 0.001) and the CRF model (88.3% vs. 75.8%, p < 0.001). The neural network also showed significantly higher keyword level accuracy compared to the rule-based algorithm (95.5% vs. 80.9% p < 0.001) and CRF model (95.5% vs. 76.9%, p < 0.001). We demonstrate the potential of neural networks to accurately perform word-level multilabel classification of free text radiology reports across 33 classes, thus showing the utility of a sequence labeling approach to NLP of radiology reports. We found that a neural network classifier outperforms a rule-based algorithm and a CRF classifier for comprehensive multilabel classification of free text screening mammography reports at the word level. By approaching radiology report classification as a sequence-labeling problem, we demonstrate the ability of neural networks to extract data from free text radiology reports at a level of granularity not previously reported.

A Normal Thyroid by Any Other Name: Linguistic Analysis of Statements Describing a Normal Thyroid Gland from Noncontrast Chest CT Reports.

PURPOSE: To determine the variability and readability of language used in chest CT reports to describe a "normal" thyroid gland. METHODS: Using a semi-automated process, we identified unique sentences or phrases describing a "normal" thyroid gland in 11,357 noncontrast chest CT reports. Readability metrics were computed for each descriptor, including sentence length and grade-level indices quantifying the education necessary for text comprehension. Grade-level indices included the Flesch-Kincaid (FK) grade level, Gunning Fog index (GF), Coleman-Liau (CL) index, automated readability index (ARI), and a computed composite grade level (CGL) calculated as (FK + GF + CL + ARI)/4. RESULTS: There were 342 unique "normal" thyroid descriptors identified among 6,957 noncontrast chest CT reports characterizing a normal thyroid gland. For these 342 unique descriptors, sentence length varied 23-fold, with a mean sentence length of 8.3 ± 5.1 words. CGL varied 4-fold, with a mean of 16.4 ± 4.5, suggesting that descriptors for a normal thyroid gland, on average, require an advanced college-level education for comprehension. CONCLUSIONS: The language used by radiologists to describe a normal thyroid gland in chest CT reports is variable and complex. The linguistic characteristics observed herein may be a surrogate for the broader readability of radiology reports. With the growing role of the radiology report in patient communication, further linguistic analysis of reporting language may provide valuable insight for optimizing radiology communication.

Patient-Centered Radiology Reporting: Using Online Crowdsourcing to Assess the Effectiveness of a Web-Based Interactive Radiology Report.

PURPOSE: The aim of this study was to evaluate the effectiveness of a patient-centered web-based interactive mammography report. METHODS: A survey was distributed on Amazon Mechanical Turk, an online crowdsourcing platform. One hundred ninety-three US women ≥18 years of age were surveyed and then randomized to one of three simulated BI-RADS® 0 report formats: standard report, Mammography Quality Standards Act-modeled patient letter, or web-based interactive report. Survey questions assessed participants' report comprehension, satisfaction with and perception of the interpreting radiologist, and experience with the presented report. Two-tailed t tests and χ2 tests were used to evaluate differences among groups. RESULTS: Participants in the interactive web-based group spent more than double the time viewing the report than the standard report group (160.0 versus 64.2 seconds, P < .001). Report comprehension scores were significantly higher for the interactive web-based and patient letter groups than the standard report group (P < .05). Scores of satisfaction with the interpreting radiologist were significantly higher for the web-based interactive report and patient letter groups than the standard report group (P < .01). There were no significant differences between the patient letter and web-based interactive report groups. CONCLUSIONS: Radiology report format likely influences communication effectiveness. For result communication to a non-medical patient audience, patient-centric report formats, such as a Mammography Quality Standards Act-modeled patient letter or web-based interactive report, may offer advantages over the standard radiology report. Future work is needed to determine if these findings are reproducible in patient care settings and to determine how best to optimize radiology result communication to patients.

Performance of a Machine Learning Classifier of Knee MRI Reports in Two Large Academic Radiology Practices: A Tool to Estimate Diagnostic Yield.

OBJECTIVE: The purpose of this study is to evaluate the performance of a natural language processing (NLP) system in classifying a database of free-text knee MRI reports at two separate academic radiology practices. MATERIALS AND METHODS: An NLP system that uses terms and patterns in manually classified narrative knee MRI reports was constructed. The NLP system was trained and tested on expert-classified knee MRI reports from two major health care organizations. Radiology reports were modeled in the training set as vectors, and a support vector machine framework was used to train the classifier. A separate test set from each organization was used to evaluate the performance of the system. We evaluated the performance of the system both within and across organizations. Standard evaluation metrics, such as accuracy, precision, recall, and F1 score (i.e., the weighted average of the precision and recall), and their respective 95% CIs were used to measure the efficacy of our classification system. RESULTS: The accuracy for radiology reports that belonged to the model's clinically significant concept classes after training data from the same institution was good, yielding an F1 score greater than 90% (95% CI, 84.6-97.3%). Performance of the classifier on cross-institutional application without institution-specific training data yielded F1 scores of 77.6% (95% CI, 69.5-85.7%) and 90.2% (95% CI, 84.5-95.9%) at the two organizations studied. CONCLUSION: The results show excellent accuracy by the NLP machine learning classifier in classifying free-text knee MRI reports, supporting the institution-independent reproducibility of knee MRI report classification. Furthermore, the machine learning classifier performed well on free-text knee MRI reports from another institution. These data support the feasibility of multiinstitutional classification of radiologic imaging text reports with a single machine learning classifier without requiring institution-specific training data.

Comparison of 4D-microSPECT and microCT for murine cardiac function.

PURPOSE: The objective of this study was to compare a new generation of four-dimensional micro-single photon emission computed tomography (microSPECT) with microCT for the quantitative in vivo assessment of murine cardiac function. PROCEDURES: Four-dimensional isotropic cardiac images were acquired from anesthetized normal C57BL/6 mice with either microSPECT (n = 6) or microCT (n = 6). One additional mouse with myocardial infarction (MI) was scanned with both modalities. Prior to imaging, mice were injected with either technetium tetrofosmin for microSPECT or a liposomal blood pool contrast agent for microCT. Segmentation of the left ventricle (LV) was performed using Vitrea (Vital Images) software, to derive global and regional function. RESULTS: Measures of global LV function between microSPECT and microCT groups were comparable (e.g., ejection fraction = 71 ± 6 % microSPECT and 68 ± 4 % microCT). Regional functional indices (wall motion, wall thickening, regional ejection fraction) were also similar for the two modalities. In the mouse with MI, microSPECT identified a large perfusion defect that was not evident with microCT. CONCLUSIONS: Despite lower spatial resolution, microSPECT was comparable to microCT in the quantitative evaluation of cardiac function. MicroSPECT offers an advantage over microCT in the ability to evaluate simultaneously myocardial radiotracer distribution and function, simultaneously. MicroSPECT should be considered as an alternative to microCT and magnetic resonance for preclinical cardiac imaging in the mouse.

Abdominopelvic and lower extremity deep venous thrombosis: evaluation with contrast-enhanced MR venography with a blood-pool agent.

OBJECTIVE: The purpose of this article is to evaluate contrast-enhanced (CE) MR venography (MRV) with a blood-pool agent for detection of abdominopelvic and lower extremity deep venous thrombosis (DVT) compared with a conventional unenhanced gradient-recalled echo (GRE) MRV technique. MATERIALS AND METHODS: This retrospective study was performed on 30 patients (mean age, 52.7 years; 15 men and 15 women) referred for MRV between March 2010 and November 2010 for evaluation of lower extremity or abdominopelvic DVT. All patients underwent a GRE sequence followed by a CE T1-weighted sequence with gadofosveset, a blood-pool agent. The abdominopelvic and lower extremity venous system was divided into 13 segments. The presence of acute or chronic DVT was assessed by six radiologists, as well as qualitative and quantitative assessments of each venous segment. Image acquisition and interpretation times were also tabulated. RESULTS: The sensitivity and specificity for acute DVT were 91.0% and 99.8%, respectively, on CE MRV compared with 80.8% and 95.8%, respectively, on GRE MRV (p = 0.077 and p < 0.001). The sensitivity and specificity for chronic DVT were 84.4% and 98.4%, respectively, on CE MRV and 64.5% and 95.6%, respectively, on GRE MRV (p < 0.001 for both). Subjective ratings of vein visualization, signal homogeneity, and confidence pertaining to DVT diagnosis were significantly higher with the CE images (p < 0.001). The contrast-to-noise ratio for CE images was similar or significantly higher for all venous segments. Image acquisition and radiologist interpretation times on the CE studies were decreased (p < 0.001). CONCLUSION: Gadofosveset-enhanced MRV had equal or higher sensitivity and specificity for detection of DVT than did GRE MRV, with decreased time for image acquisition and interpretation.