Quantifying and leveraging predictive uncertainty for medical image assessment

The interpretation of medical images is a challenging task, often complicated by the presence of artifacts, occlusions, limited contrast and more. Most notable is the case of chest radiography, where there is a high inter-rater variability in the detection and classification of abnormalities. This is largely due to inconclusive evidence in the data or subjective definitions of disease appearance. An additional example is the classification of anatomical views based on 2D Ultrasound images. Often, the anatomical context captured in a frame is not sufficient to recognize the underlying anatomy. Current machine learning solutions for these problems are typically limited to providing probabilistic predictions, relying on the capacity of underlying models to adapt to limited information and the high degree of label noise. In practice, however, this leads to overconfident systems with poor generalization on unseen data. To account for this, we propose a system that learns not only the probabilistic estimate for classification, but also an explicit uncertainty measure which captures the confidence of the system in the predicted output. We argue that this approach is essential to account for the inherent ambiguity characteristic of medical images from different radiologic exams including computed radiography, ultrasonography and magnetic resonance imaging. In our experiments we demonstrate that sample rejection based on the predicted uncertainty can significantly improve the ROC-AUC for various tasks, e.g., by 8% to 0.91 with an expected rejection rate of under 25% for the classification of different abnormalities in chest radiographs. In addition, we show that using uncertainty-driven bootstrapping to filter the training data, one can achieve a significant increase in robustness and accuracy. Finally, we present a multi-reader study showing that the predictive uncertainty is indicative of reader errors.     

Comparison between tension band and cerclage with X-Plate and lag screws in treatment of comminuted patellar fractures

BackgroundComminuted patellar fractures are not rare, and the ideal treatment method remains controversial. The present study was conducted to evaluate effects and compare complications of two different methods used to treat comminuted patellar fractures.MethodsFrom March 2010 to August 2016, 102 cases of 34-C2 or 34-C3 comminuted patellar fractures were treated at our hospital, wherein patients received two different treatments: titanium cable tension band with cerclage method (group A) and intrafragmentary screws with X-shaped plating technique (group B). At follow-ups, articular step-off, range of motion (ROM), Lysholm scores, time of union, and complications were recorded and analyzed. Radiographic and clinical data as well as rate of complications were statistically analyzed.ResultsIn total, 87 patients were included in the final analysis (n = 47 in group A and n = 40 in group B). No significant differences were noted in terms of cost of implant, age, gender, rate of 34-C3 fractures, rate of layered inferior pole fractures, postoperative articular step-off and union time. At 2-year follow-up, average Lysholm scores, ROM and rate of complications were (89.0 ± 4.5), (122°±12°) and (27.7%) in group A and (90.2 ± 3.9), (124°±11°) and (17.5%) in group B, respectively, with no significant differences (p > 0.05). The mean time of surgery in group B was shorter than that in group A with significant difference (p < 0.05).ConclusionsTreatment using the intrafragmentary screws and plate method for amenable comminuted patellar fractures achieved similar complication rate and favorable functional outcomes at the 2-year follow-up, which was comparable to the titanium cable tension band with cerclage method. Thus, the intrafragmentary screws and plate method is effective, safe and convenient for 34-C2/C3 comminuted patellar fractures, especially appropriate for patients with layered fragments.     

Confounded association between proton pump inhibitor use and risk of biliary tract cancer: Result from three cohorts

Recent epidemiological studies suggested that proton pump inhibitor (PPI) use was associated with an increased risk of biliary tract cancer (BTC), however, confounders were not adequately controlled. Our study aimed to evaluate PPI use and subsequent risk of BTC and its subtypes in three well-established cohorts. We conducted a pooled analysis of the subjects free of cancers in UK Biobank (n = 463 643), Nurses' Health Study (NHS, n = 80 235) and NHS II (n = 95 869). Propensity score weighted Cox models were used to estimate marginal HRs of PPIs use on BTC risk, accounting for potential confounders. We documented 284 BTC cases in UK Biobank (median follow-up: 7.6 years), and 91 cases in NHS and NHS II cohorts (median follow-up: 15.8 years). In UK biobank, PPI users had a 96% higher risk of BTC compared to nonusers in crude model (HR 1.96, 95% CI 1.44-2.66), but the effect was attenuated to null after adjusting for potential confounders (HR 0.95, 95% CI 0.60-1.49). PPI use was not associated with risk of BTC in the pooled analysis of three cohorts (HR 0.93, 95% CI 0.60-1.43). We also observed no associations between PPI use with risk of intrahepatic (HR 1.00, 95% CI 0.49-2.04), extrahepatic bile duct (HR 1.09, 95% CI 0.52-2.27) and gallbladder cancers (HR 0.66, 95% CI 0.26-1.66) in UK Biobank. In summary, regular use of PPIs was not associated with the risk of BTC and its subtypes.     

A meta-analysis on salvage surgery as a potentially curative procedure in patients with isolated local recurrent or persistent esophageal cancer after chemoradiotherapy

BackgroundIsolated local recurrent or persistent esophageal cancer (EC) after curative intended definitive (dCRT) or neoadjuvant chemoradiotherapy (nCRT) with initially omitted surgery, is a potential indication for salvage surgery. We aimed to evaluate safety and efficacy of salvage surgery in these patients.Material and methodsA systematic literature search following PRISMA guidelines was performed using databases of PubMed/Medline. All included studies were performed in patients with persistent or recurrent EC after initial treatment with dCRT or nCRT, between 2007 and 2017. Survival analysis was performed with an inverse-variance weighting method.ResultsOf the 278 identified studies, 28 were eligible, including a total of 1076 patients. Postoperative complications after salvage esophagectomy were significantly more common among patients with isolated persistent than in those with locoregional recurrent EC, including respiratory (36.6% versus 22.7%; difference in proportion 10.9 with 95% confidence interval (CI) [3.1; 18.7]) and cardiovascular complications (10.4% versus 4.5%; difference in proportion 5.9 with 95% CI [1.5; 10.2]). The pooled estimated 30- and 90-day mortality was 2.6% [1.6; 3.6] and 8.0% [6.3; 9.8], respectively. The pooled estimated 3-year and 5-year overall survival (OS) were 39.0% (95% CI: [35.8; 42.2]) and 19.4% [95% CI:16.5; 22.4], respectively. Patients with isolated persistent or recurrent EC after initial CRT had similar 5-year OS (14.0% versus 19.7%, difference in proportion −5.7, 95% CI [-13.7; 2.3]).ConclusionsSalvage surgery is a potentially curative procedure in patients with locally recurrent or persistent esophageal cancer and can be performed safely after definitive or neoadjuvant chemoradiotherapy when surgery was initially omitted.     

Replacing HMG/FSH by low-dose HCG to complete corifollitropin alfa stimulation reduces cost per clinical pregnancy: a randomized pragmatic trial

Research questionThe cost of IVF treatment remains high, among other factors because of the medication needed for ovarian stimulation. This study investigated the effect of using low-dose human chorionic gonadotrophin (HCG) for the second phase of follicular maturation after corifollitropin alfa induction, to replace the more expensive, either recombinant or human menopausal gonadotrophin (HMG), on the cost of ovarian stimulation.DesignOne hundred and five patients were randomly divided into two groups: patients in the HCG group (n = 50) received low-dose HCG from Day 7 until the diameter of at least three follicles reached 17 mm or more, while patients in the FSH group (n = 55) received conventional ovarian stimulation with highly purified HMG injections.ResultsThe clinical pregnancy rate in the HCG group was 38% higher than in the FSH group (number needed to treat, NNT = 13). The cost per pregnancy needed for ovarian stimulation was reduced from €4902 in the FSH group to €2684 in the HCG group. Hence, the cost of ovarian stimulation medication to obtain 10 pregnancies using the conventional FSH protocol is sufficient to attain 18 pregnancies when applying the low-dose HCG protocol.ConclusionThis study provides evidence that using HCG instead of HMG/FSH for ovarian stimulation results in a significant reduction in the cost of IVF with, at least, an equivalent pregnancy rate.     

Clinical characteristics and surgical outcomes of resectable acinar cell carcinoma of the pancreas-propensity score matching analysis with pancreatic ductal adenocarcinoma

BackgroundSurgical resection is recommended for patients with resectable acinar cell carcinoma (ACC). The aim of this study was to investigate the clinical characteristics and surgical outcomes of resectable ACC in comparison to pancreatic ductal adenocarcinoma (PDAC).MethodA retrospective analysis was performed on all patients who consecutively underwent radical resection with pathologically confirmed ACC and PDAC from December 2011 to December 2018. Clinicopathologic characteristics and follow-up information were analyzed. A 1:3 propensity score matching (PSM) method was used to minimize the bias between ACC and PDAC.ResultsA total of 26 patients with ACC and 1351 with PDAC were included. Compared to PDAC, ACC tended to be larger (4.5 vs. 3.0 cm; p < 0.001) and more frequently located in the pancreatic body/tail (61.5% vs. 36.6%, p = 0.009), with lower total bilirubin levels, lower neutrophil lymphocyte ratio (NLR) levels and lower carbohydrate antigen 19-9 (CA19-9) levels and carcinoembryonic antigen (CEA) levels. There was no difference in postoperative morbidities in patients with ACC and PDAC. The median OS and RFS were longer in ACC when compared to PDAC (OS: 43.5 mo vs. 19.0 mo, p = 0.004; RFS: 24.5 mo vs. 11.6 mo, p = 0.023). After the 1:3 PSM, ACC remained to be a better histological type for OS (p = 0.024), but had comparable RFS with PDAC (p = 0.164).ConclusionPatients with ACC after radical resection had better OS than that with PDAC. However, ACC is also an aggressive tumor with a similar trend of RFS with PDAC after the matching, necessitating the multidisciplinary treatment for resectable ACC disease.