Evaluation of ensemble strategy on the development of multiple view ankle fracture detection algorithm.

OBJECTIVE: To identify the feasibility and efficiency of deep convolutional neural networks (DCNNs) in the detection of ankle fractures and to explore ensemble strategies that applied multiple projections of radiographs.Ankle radiographs (AXRs) are the primary tool used to diagnose ankle fractures. Applying DCNN algorithms on AXRs can potentially improve the diagnostic accuracy and efficiency of detecting ankle fractures. METHODS: A DCNN was trained using a trauma image registry, including 3102 AXRs. We separately trained the DCNN on anteroposterior (AP) and lateral (Lat) AXRs. Different ensemble methods, such as "sum-up," "severance-OR," and "severance-Both," were evaluated to incorporate the results of the model using different projections of view. RESULTS: The AP/Lat model's individual sensitivity, specificity, positive-predictive value, accuracy, and F1 score were 79%/84%, 90%/86%, 88%/86%, 83%/85%, and 0.816/0.850, respectively. Furthermore, the area under the receiver operating characteristic curve (AUROC) of the AP/Lat model was 0.890/0.894 (95% CI: 0.826-0.954/0.831-0.953). The sum-up method generated balanced results by applying both models and obtained an AUROC of 0.917 (95% CI: 0.863-0.972) with 87% accuracy. The severance-OR method resulted in a better sensitivity of 90%, and the severance-Both method obtained a high specificity of 94%. CONCLUSION: Ankle fracture in the AXR could be identified by the trained DCNN algorithm. The selection of ensemble methods can depend on the clinical situation which might help clinicians detect ankle fractures efficiently without interrupting the current clinical pathway. ADVANCES IN KNOWLEDGE: This study demonstrated different ensemble strategies of AI algorithms on multiple view AXRs to optimize the performance in various clinical needs.

Comparison of outcomes of proximal versus distal and combined splenic artery embolization in the management of blunt splenic injury: a report of 202 cases from a single trauma center.

BACKGROUND: To compare the outcomes of blunt splenic injuries (BSI) managed with proximal (P) versus distal (D) versus combined (C) splenic artery embolization (SAE). METHODS: This retrospective study included patients with BSI who demonstrated vascular injuries on angiograms and were managed with SAE between 2001 and 2015. The success rate and major complications (Clavien-Dindo classification ≥ III) were compared between the P, D, and C embolizations. RESULTS: In total, 202 patients were enrolled (P, n = 64, 31.7%; D, n = 84, 41.6%; C, n = 54, 26.7%). The median injury severity score was 25. The median times from injury to SAE were 8.3, 7.0, and 6.6 h for the P, D, and C embolization, respectively. The overall haemostasis success rates were 92.6%, 93.8%, 88.1%, and 98.1% in the P, D, and C embolizations, respectively, with no significant difference (p = 0.079). Additionally, the outcomes were not significantly different between the different types of vascular injuries on angiograms or the materials used in the location of embolization. Splenic abscess occurred in six patients (P, n = 0; D, n = 5; C, n = 1), although it occurred more commonly in those who underwent D embolization with no significant difference (p = 0.092). CONCLUSIONS: The success rate and major complications of SAE were not significantly different regardless of the location of embolization. The different types of vascular injuries on angiograms and agents used in different embolization locations also did not affect the outcomes.

Comparing the Clinical Efficacy of Coil Embolization in GDA Stump versus Common Hepatic Artery in Postoperative Hemorrhage after Pancreatoduodenectomy.

BACKGROUND: Hemorrhage after pancreaticoduodenectomy is an uncommon but fatal complication. In this retrospective study, the different treatment modalities and outcomes for treating post-pancreaticoduodenectomy hemorrhage are analyzed. METHODS: Our hospital imaging database was queried to identify patients who had undergone pancreaticoduodenectomy during the period of 2004-2019. The patients were retrospectively split into three groups, according to their treatment: conservative treatment without embolization (group A: A1, negative angiography; A2, positive angiography), hepatic artery sacrifice/embolization (group B: B1, complete; B2, incomplete), and gastroduodenal artery (GDA) stump embolization (group C). RESULTS: There were 24 patients who received angiography or transarterial embolization (TAE) treatment 37 times (cases). In group A, high re-bleeding rates (60%, 6/10 cases) were observed, with 50% (4/8 cases) for subgroup A1 and 100% (2/2 cases) for subgroup A2. In group B, the re-bleeding rates were lowest (21.1%, 4/19 cases) with 0% (0/16 cases) for subgroup B1 and 100% (4/4 cases) for subgroup B2. The rate of post-TAE complications (such as hepatic failure, infarct, and/or abscess) in group B was not low (35.3%, 6/16 patients), especially in patients with underlying liver disease, such as liver cirrhosis and post-hepatectomy (100% (3/3 patients), vs. 23.1% (3/13 patients); p = 0.036, p < 0.05). The highest rate of re-bleeding (62.5%, 5/8 cases) was observed for group C. There was a significant difference in the re-bleeding rates of subgroup B1 and group C (p = 0.00017). The more iterations of angiography, the higher the mortality rate (18.2% (2/11 patients), <3 times vs. 60% (3/5 patients), ≥3 times; p = 0.245). CONCLUSIONS: The complete sacrifice of the hepatic artery is an effective first-line treatment for pseudoaneurysm or for the rupture of the GDA stump after pancreaticoduodenectomy. Hepatic complications are not uncommon and are highly associated with underlying liver disease. Conservative treatment, the selective embolization of the GDA stump, and incomplete hepatic artery embolization do not provide enduring treatment effects.

Timely Antecedent CT or MRI Can Help Predict Hemorrhage Site of Posttreatment Head and Neck Cancer, With Digital Subtraction Angiography Used as the Reference Standard.

OBJECTIVE: We investigated the timing of CT and MRI performed before digital subtraction angiography (DSA) in the prediction of hemorrhage sites in patients with head and neck cancers who present with acute oral or neck bleeding after receiving treatment. MATERIALS AND METHODS: A total of 123 DSA examinations that evaluated 123 oral or neck bleeding events in 85 patients were analyzed. The last CT or MRI examinations performed within a time frame of 0-337 days before transarterial embolization were reviewed retrospectively, with three findings (pseudoaneurysm, air-containing necrotic tissue, and residual tumor) used to predict hemorrhage sites. DSA findings of pseudoaneurysm or active contrast extravasation were used as a reference standard. The sensitivity of CT and MRI for correctly predicting hemorrhage sites was used to determine the optimal timing of CT or MRI examinations performed before DSA. RESULTS: A total of 8.9% of the DSA examinations (11/123) had equivocal findings but were followed by another bleeding event for which DSA findings were positive. CT or MRI was statistically significantly better at predicting hemorrhage sites in patients with bleeding events associated with nonhypopharyngeal cancers (p = 0.019) than in those with bleeding events associated with hypopharyngeal cancers. The sensitivity of CT or MRI in the prediction of hemorrhage sites was statistically significantly higher for the common carotid artery and the internal carotid artery when CT or MRI was performed less than 30 days before bleeding events occurred. Prediction of hemorrhagic sites was better with the use of CT angiography than with the use of enhanced CT or MRI, although it was not statistically significant. CONCLUSION: DSA findings can temporarily be equivocal. CT or MRI examinations performed within 30 days of bleeding events can predict the site of hemorrhage. If no CT or MRI findings from the past 30 days are available, we suggest performing emergent CT angiography for the sake of obtaining better arterial detail.

Pretreatment Dynamic Contrast-Enhanced MRI Improves Prediction of Early Distant Metastases in Patients With Nasopharyngeal Carcinoma.

The identification of early distant metastases (DM) in patients with newly diagnosed, previously untreated nasopharyngeal carcinoma (NPC) plays an important role in selecting the most appropriate treatment approach. Here, we sought to investigate the predictive value of distinct MRI parameters for the detection of early DM.Between November 2010 and June 2011, a total of 51 newly diagnosed NPC patients were included. All of the study participants were followed until December 2014 at a single institution after completion of therapy. DM was defined as early when they were detected on pretreatment FDG-PET scans or within 6 months after initial diagnosis. The following parameters were tested for their ability to predict early DM: pretreatment FDG-PET standardized uptake value (SUV), MRI-derived AJCC tumor staging, tumor volume, and dynamic contrast-enhanced (DCE) values. The DCE-derived ve was defined as the volume fraction of the extravascular, extracellular space.Compared with patients without early DM, patients with early DM had higher SUV, tumor volume, DCE mean (median) ve, ve skewness, ve kurtosis, and the largest mean ve selected among sequential slices (P < 0.05). No differences were identified when early DM were defined only according to the results of pretreatment FDG-PET. Among different quantitative DCE parameters, the mean ve had the highest area under curve (AUC, 0.765). However, the AUCs of SUV, tumor volume, mean ve, ve skewness, ve kurtosis, or the largest mean ve selected among the sequential slices did not differ significantly from one another (P = 0.82).Taken together, our results suggest that DCE-derived ve may be a useful parameter in combination with SUV and tumor volume for predicting early DM. Dynamic contrast-enhanced MRI may be complementary to FDG-PET for selecting the most appropriate treatment approach in NPC patients.

Computed tomography arterial portography for assessment of portal vein injury after blunt hepatic trauma.

PURPOSE: Intrahepatic portal vein injuries secondary to blunt abdominal trauma are difficult to diagnose and can result in insidious bleeding. We aimed to compare computed tomography arterial portography (CTAP), reperfusion CTAP (rCTAP), and conventional computed tomography (CT) for diagnosing portal vein injuries after blunt hepatic trauma. METHODS: Patients with blunt hepatic trauma, who were eligible for nonoperative management, underwent CTAP, rCTAP, and CT. The number and size of perfusion defects observed using the three methods were compared. RESULTS: A total of 13 patients (seven males/six females) with a mean age of 34.5±14.1 years were included in the study. A total of 36 hepatic segments had perfusion defects on rCTAP and CT, while there were 47 hepatic segments with perfusion defects on CTAP. The size of perfusion defects on CT (239 cm3; interquartile range [IQR]: 129.5, 309.5) and rCTAP (238 cm3; IQR: 129.5, 310.5) were significantly smaller compared with CTAP (291 cm3; IQR: 136, 371) (both, P = 0.002). CONCLUSION: Perfusion defects measured by CTAP were significantly greater than those determined by either rCTAP or CT in cases of blunt hepatic trauma. This finding suggests that CTAP is superior to rCTAP and CT in evaluating portal vein injuries after blunt liver trauma.

Endovascular covered stent reconstruction improved the outcomes of acute carotid blowout syndrome. Experiences at a single institute.

SUMMARY: Carotid blowout is a devastating complication in patients with head and neck cancer, commonly encountered as a delayed complication of radiation therapy. The clinical outcomes in patients with carotid blowout are discouraging; even transarterial embolization has been performed to control the acute massive bleeding. In recent years, covered stents have been reported as an alternative treatment producing favorable results. In this study, 13 consecutive patients with acute carotid blowout syndrome were treated at our institute by covered-stent reconstruction between December 2005 and December 2007. The median posthemorrhagic survival period after reconstruction (187 days) was more than that reported in patients treated only with transarterial embolization (26 days). Though the estimated mortality was about 54%, those who survived showed favorable outcomes, and only one transit complication of acute in-stent thrombosis occurred. Thus, endovascular covered-stent reconstruction is a safe and effective approach to manage acute carotid blowout syndrome.