Andrew G. Taylor

Pneumothorax can precipitate a life-threatening emergency due to lung collapse and respiratory or circulatory distress. Pneumothorax is typically detected on chest X-ray; however, treatment is reliant on timely review of radiographs. Since current imaging volumes may result in long worklists of radiographs awaiting review, an automated method of prioritizing X-rays with pneumothorax may reduce time to treatment. Our objective was to create a large human-annotated dataset of chest X-rays containing pneumothorax and to train deep convolutional networks to screen for potentially emergent moderate or large pneumothorax at the time of image acquisition. In all, 13,292 frontal chest X-rays (3,107 with pneumothorax) were visually annotated by radiologists. This dataset was used to train and evaluate multiple network architectures. Images showing large- or moderate-sized pneumothorax were considered positive, and those with trace or no pneumothorax were considered negative. Images showing small pneumothorax were excluded from training. Using an internal validation set (n = 1,993), we selected the 2 top-performing models; these models were then evaluated on a held-out internal test set based on area under the receiver operating characteristic curve (AUC), sensitivity, specificity, and positive predictive value (PPV). The final internal test was performed initially on a subset with small pneumothorax excluded (as in training; n = 1,701), then on the full test set (n = 1,990), with small pneumothorax included as positive. External evaluation was performed using the National Institutes of Health (NIH) ChestX-ray14 set, a public dataset labeled for chest pathology based on text reports. All images labeled with pneumothorax were considered positive, because the NIH set does not classify pneumothorax by size. In internal testing, our “high sensitivity model” produced a sensitivity of 0.84 (95% CI 0.78–0.90), specificity of 0.90 (95% CI 0.89–0.92), and AUC of 0.94 for the test subset with small pneumothorax excluded. Our “high specificity model” showed sensitivity of 0.80 (95% CI 0.72–0.86), specificity of 0.97 (95% CI 0.96–0.98), and AUC of 0.96 for this set. PPVs were 0.45 (95% CI 0.39–0.51) and 0.71 (95% CI 0.63–0.77), respectively. Internal testing on the full set showed expected decreased performance (sensitivity 0.55, specificity 0.90, and AUC 0.82 for high sensitivity model and sensitivity 0.45, specificity 0.97, and AUC 0.86 for high specificity model). External testing using the NIH dataset showed some further performance decline (sensitivity 0.28–0.49, specificity 0.85–0.97, and AUC 0.75 for both). Due to labeling differences between internal and external datasets, these findings represent a preliminary step towards external validation. We trained automated classifiers to detect moderate and large pneumothorax in frontal chest X-rays at high levels of performance on held-out test data. These models may provide a high specificity screening solution to detect moderate or large pneumothorax on images collected when human review might be delayed, such as overnight. They are not intended for unsupervised diagnosis of all pneumothoraces, as many small pneumothoraces (and some larger ones) are not detected by the algorithm. Implementation studies are warranted to develop appropriate, effective clinician alerts for the potentially critical finding of pneumothorax, and to assess their impact on reducing time to treatment.

Purpose To evaluate the radiation exposure of pediatric patients during exchange of percutaneous gastrojejunostomy tubes. Materials and Methods Retrospective review of consecutive pediatric patients undergoing percutaneous gastrojejunostomy tube exchanges from January 1, 2010 to August 8, 2016 was performed. Procedural records were evaluated for total fluoroscopy time, cumulative air kerma, and cumulative dose area product (DAP). The total number of procedures, indications, and interval time for repeat procedures were also recorded. Patients were subcategorized into an increased cumulative procedure group if three or more procedures were performed and an increased frequency group if the interval between any two consecutive procedures occurred within a 6-month period. Results In the 77-month study period, a total of 175 gastrojejunostomy tube exchanges were performed on 62 patients, median age 62 months (range: 1 month–206 months). The mean fluoroscopy time was 7.36 minutes (range: 0.20–40 minutes), mean cumulative air kerma was 41.13 mGy (range: 0.40–507 mGy), and mean cumulative DAP was 3,717.19 mGy-cm² (range: 102–72,479 mGy-cm²). Twenty-six patients were subcategorized into an “increased cumulative procedure” group with an average of five procedures during the study period. The mean fluoroscopy time, cumulative air kerma, and cumulative DAP for the individual procedures of this group was 8.13 minutes, 49.07 mGy, and 4250.73 mGy-cm², respectively. Thirty-three patients were subcategorized into an “increased frequency” group with average 2 months interval between procedures. The mean fluoroscopy time, cumulative air kerma, and cumulative DAP for the individual procedures of this group was 7.87 minutes, 46.77 mGy, and 4147.52 mGy-cm², respectively. Twenty-three patients underwent single procedures during the study period. The mean fluoroscopy time, cumulative air kerma, and cumulative DAP for the individual procedures of this group was 3.02 minutes, 6.11 mGy, and 1232.13 mGy-cm², respectively. A statistically significant difference in means (p < 0.05) exists in all radiation metrics for a given procedure while comparing either the “increased cumulative procedure” group or the “increased frequency group” to the single procedure group. Conclusion Chronically ill pediatric patients may require serial percutaneous gastrojejunostomy tube exchanges, subjecting them to high doses of radiation exposure, particularly if repeat procedures are required. Such patients may benefit from alternative methods of enteral feeding, such as through surgical interventions.

Nonalcoholic fatty liver disease (NAFLD) has become highly prevalent, now considered the most common liver disease in the western world. Approximately one-third of patients with NASH develop non-alchoholic steatohepatitis (NASH), histologically defined by lobular and portal inflammation, and accompanied by marked oxidative stress. Patients with NASH are at increased risk for cirrhosis and hepatocellular carcinoma, and diagnosis currently requires invasive biopsy. In animal models of NASH, particularly the methionine-choline deficient (MCD) model, profound changes are seen in redox enzymes and key intracellular antioxidants. To study antioxidant status in NASH non-invasively, we applied the redox probe hyperpolarized [1-¹³C] dehydroascorbic acid (HP DHA), which is reduced to Vitamin C (VitC) rapidly in the normal liver. In MCD mice, we observed a significant decrease in HP DHA to VitC conversion that accompanied hepatic fat deposition. When these animals were subsequently placed on a normal diet, resonance ratios reverted to those seen in control mice. These findings suggest that HP DHA, a potentially clinically translatable imaging agent, holds special promise in imaging NASH and other metabolic syndromes, to monitor disease progression and response to targeted therapies.

Objective: As a safe and effective local treatment for hepatocellular carcinoma (HCC), percutaneous ethanol injection (PEI) is commonly performed with other locoregional treatments. In patients with coagulopathy, ascites, and thrombocytopenia, PEI is a low-risk option to help maintain local control. The purpose of this study was to evaluate the safety and efficacy of PEI for transplant-candidate patients with HCC who had marginal hepatic reserve. Materials and Methods: Retrospective review was conducted on 29 transplant-candidate patients with marginal hepatic reserve who underwent PEI at our institution from December 2008 to June 2016. Marginal hepatic reserve was defined as platelets < 60,000/mL of blood, total bilirubin ≥ 3.0 mg/dL, international normalized ratio ≥ 1.5, or the presence of ascites. The mRECIST criteria were used to assess radiographic response to treatment. Complications were graded based on NCI common terminology criteria for adverse events. Results: Twenty-nine patients (62% male) with median age of 59 years (range: 50–74 years) were followed for a median time of 85 days (range: 0–720 days). The median size of the target lesion prior to PEI was 2.0 cm (range: 0.8–3.1 cm). 93% received TACE and 17% underwent RFA prior to PEI. 45% ultimately were transplanted with median wait of 158 days (range 12–720 days). 34% died with median time to death of 56 days (range: 29–403 days). Based on mRECIST criteria, 54% (13/24) had complete radiographic response while only 8% (2/24) had progressive disease within a month follow-up. After PEI, hepatotoxicity occurred in 6 patients (21%), with liver tests returning to baseline in 3 patients. Conclusion: PEI is an effective and safe locoregional treatment for patients with HCC who have marginal hepatic reserve to help bridge them to transplant.

Objective: To compare the visibility of hepatocellular carcinoma (HCC) at angiography during transarterial chemoembolization (TACE) in patients with and without transjugular intrahepatic portosystemic shunt (TIPS). Materials and Methods: A retrospective review of patients with HCC and TIPS who underwent TACE between January 2004 and June 2014 was performed resulting in 26 patients. A group of 23 patients with HCC treated with TACE but without TIPS were used as comparison that encompassed the same time period. Both groups had comparable model for end-stage liver disease (MELD) scores. The angiographic tumor blush of the HCC at TACE was graded as poorly visualized, easily visualized, or hyperenhancing by two independent board-certified interventional radiologists. Results: Twenty-six patients with TIPS and 23 patients without TIPS who underwent TACE for HCC were included resulting in a total of 98 grades by the two reviewers. The tumor blush of the HCC was more visible in patients without TIPS than those with TIPS during TACE for both reviewers, (p = 0.0003). In the TIPS group, only 5 of 26 patients (19%) had a hyperenhancing blush at TACE for both reviewers. Hyperenhancing or easily visualized tumor blush was present in 22 of 23 patients (96%) for reviewer 1 and in 17 of 23 patients (74%) for reviewer 2 without TIPS. Of the HCC’s with a poorly visualized tumor blush, 12 of 13 patients (92%) and 14 of 20 patients (70%) had TIPS for reviewer 1 and reviewer 2, respectively. Conclusion: The tumor stain of HCC at TACE may be more difficult to see in patients with TIPS compared with patients without TIPS in place. Patients with TIPS with poor vascular blush may benefit from nonvascular interventions such as radiofrequency or alcohol ablation.