Search

Diagnostic performance of urine biomarkers in discriminating pancreatic adenocarcinoma all stages (A-C) and stage I-II (D-F) from chronic pancreatitis patients. A, ROC curves of PDAC (n=143) versus CP (n=62) patients for individual urine biomarkers in the training set (70% of the data). B, ROC curves of PDAC versus CP patients for the panel in the training set and in the independent validation set (30% of the data, PDAC n=49, CP n=30). C, Summary table. D, ROC curves of individual urine biomarkers in training dataset (70%, PDAC stage I-II n= 56, CP=66). E, ROC curves of the panel in training and validation (PDAC stage I-II n=15, CP n=26) dataset. F, Summary table. Cnorm, creatinine-normalised, creat, creatinine, AUC: area under the curve SN: sensitivity, SP: specificity with 95% Confidence Interval (CI). SN and SP in the validation set were derived for optimal cutpoint determined in the training dataset.

Detection of CA1-9 by ELISA in (A) individual time to diagnosis groups and (B) diagnosed samples and controls.

Peptides and their fragment ions used in MRM assays for TSP-1.

PDF file - 206K, Table S1 Numbers of test positive cases and controls using CA19-9 37 U/mL and CA125 30 U/mL cut-offs Figure S1 Scatter plots showing distribution of CA19-9, CA125, CEACAM1 and REG3A levels against time to diagnosis for discovery set. Zero represents the point of clinical diagnosis. Figure S2 Examples of CA19-9 and CA125 levels in individual cases with serial/longitudinal samples. Figure S3 Box and whisker plots showing serum levels of CA19-9 and CA125 for case control validation samples grouped into different time to diagnosis groups. Whisker limits represent the 5th and 95th percentiles, the box limits represent interquartile range, the horizontal line the median and the cross the mean. Case and control groups were compared using the Mann-Whitney test; significant P values (

Expression of the protein biomarkers in pancreatic cancer tissues. A, Immunohistochemical analysis of REG1A: i) REG1A in poorly differentiated PDAC, ii) luminal REG1A in malignant glands. B, TFF1: i) heterogenous expression in cancer,, ii) luminal TFF1 in malignant gland. C, LYVE1 expression in the scattered lymphatic vessels i) in the muscle layer and ii) in the stroma surrounding malignant gland. D, The biomarker levels during monitoring of pancreatic adenocarcinoma patients: LYVE1, REG1A and TFF1 were measured using ELISA in urine samples collected before surgery and during the patient's follow up. Each point represents log-transformed ELISA values at a particular time point (x-axis).

We initially investigated antigen-specific CTL responses generated upon short-term culture of T cells from PBMCs of 13 PC patients (supplemental table S1) and 10 healthy controls upon stimulation with 3 CEA HLA-A2-restricted peptides (CEA605-613 YLSGANLNL, CEA691-699 IMIGVLVGV, and CEA694-702 GVLVGVALI) as well as one irrelevant peptide Telomerase540-548 ILAKFLHWL by short-term cultures (9-10 days). This table shows the clinical information of the 13 PC patients. Antigen-specific IFN-γ producing CD8 T cells responses are shown in Supplemental Figure S1.

Exploratory comparison of plasma CA19.9 and the urine biomarker panel in discriminating early pancreatic adenocarcinoma from chronic pancreatitis patients. A, ROC curves of the biomarker panel with corresponding plasma CA19.9 alone and in combination comparing CP urine (n=50), and urines from PDAC stages I-II (n=71) and, I-IIA (n=16) (B). C, Summary table. AUC: area under the curve, SN: sensitivity, SP: specificity with 95% Confidence Interval (CI). SN and SP in the validation set were derived for optimal cutpoint determined in the training dataset.

Supplementary Figure Legends from Identification of a Three-Biomarker Panel in Urine for Early Detection of Pancreatic Adenocarcinoma

Figure S2 shows CEA691-specific cytotoxic activity of CTL lines isolated from PC patients. A. T cell line from CA11 were measured for their capability to recognise and kill T2 cells coated with 1mM (10-6 M) of CEA691, compared to control peptide, in 4-h CFSE cytotoxicity assay . B. The percentage specific cytotoxicity of CEA 691 loaded T2 cells (1mM peptide) at various E:T ratio, using CTL lines from 3 PC patients (CA07, 11, and 18). C. The cytotoxicity of T cells from CA11 against T2 cells coated with 10-5 to 10-12 M CEA691 peptide and control peptide, at E:T ratio 20:1 (left). The cytotoxicity curve was summarised on the bottom (D).

Urine biomarker concentrations in different tumours. Scatter dot plots of urine LYVE1, REG1A and plasma CA19.9 in different hepatobiliary pathologies and early stages of pancreatic adenocarcinoma (I-IIA, n=16) and I-II (n=71). The level of TFF1 protein was not measured in these samples due to substantial modifications made to the original ELISA assay by the source company at the moment of this analysis. IPMN (n=33): intraductal papillary mucinous neoplasm, AMP (n=26): ampullary cancer, NET (n=18): neuroendocrine tumour, CHL (n=24): cholangiocarcinoma, DuCA (n=16): duodenal cancer. Bars indicate median and IQR values. Upper bars: Kruskal-Wallis/Dunn's post test, *: p0.05 (no adjustment for multiplicity was made).

Correlation of the three urinary biomarkers and plasma CA19.9 (CA19.9p). A, Correlation plots (Navy blue: Healthy; Turquoise: chronic pancreatitis (CP); Purple: pancreatic adenocarcinoma (PDAC). B, Pearson correlation coefficients and corresponding significance (NS: non-significant, *: p

Expression of non-redundant proteins identified by GeLC-MS/MS from males and females in healthy controls, chronic pancreatitis and pancreatic adenocarcinoma patients.

Purpose: Biomarkers for the early detection of pancreatic cancer are urgently needed. The primary objective of this study was to evaluate whether increased levels of serum CA19-9, CA125, CEACAM1, and REG3A are present before clinical presentation of pancreatic cancer and to assess the performance of combined markers for early detection and prognosis.Experimental Design: This nested case–control study within the UKCTOCS included 118 single and 143 serial serum samples from 154 postmenopausal women who were subsequently diagnosed with pancreatic cancer and 304 matched noncancer controls. Samples were split randomly into independent training and test sets. CA19-9, CA125, CEACAM1, and REG3A were measured using ELISA and/or CLIA. Performance of markers to detect cancers at different times before diagnosis and for prognosis was evaluated.Results: At 95% specificity, CA19-9 (>37 U/mL) had a sensitivity of 68% up to 1 year, and 53% up to 2 years before diagnosis. Combining CA19-9 and CA125 improved sensitivity as CA125 was elevated (>30 U/mL) in approximately 20% of CA19-9–negative cases. CEACAM1 and REG3A were late markers adding little in combined models. Average lead times of 20 to 23 months were estimated for test-positive cases. Prediagnostic levels of CA19-9 and CA125 were associated with poor overall survival (HR, 2.69 and 3.15, respectively).Conclusions: CA19-9 and CA125 have encouraging sensitivity for detecting preclinical pancreatic cancer, and both markers can be used as prognostic tools. This work challenges the prevailing view that CA19-9 is upregulated late in the course of pancreatic cancer development. Clin Cancer Res; 21(3); 622–31. ©2014 AACR.

Urine proteome analysis. A, schematic outline of the study. B, classification of total identified proteins according to sub-cellular localisation and C, functional activity determined by Ingenuity Pathway Analysis. H: healthy, CP: chronic pancreatitis, PDAC: pancreatic ductal adenocarcinoma, GeLC/MS/MS: SDS-PAGE-Liquid Chromatography-Tandem Mass Spectrometry.

Supplementary Legends

Purpose: Carcinoembryonic antigen (CEA) is a candidate target for cellular immunotherapy of pancreatic cancer. In this study, we have characterized the antigen-specific function of autologous cytotoxic T lymphocytes (CTL) specific for the HLA-A2–restricted peptide, pCEA691-699, isolated from the peripheral T-cell repertoire of pancreatic cancer patients and sought to determine if ex vivo PD-L1 and TIM-3 blockade could enhance CTL function.Experimental Design: CD8+ T-cell lines were generated from peripheral blood mononuclear cells of 18 HLA-A2+ patients with pancreatic cancer and from 15 healthy controls. In vitro peptide-specific responses were evaluated by flow cytometry after staining for intracellular cytokine production and carboxy fluorescein succinimydyl ester cytotoxicity assays using pancreatic cancer cell lines as targets.Results: Cytokine-secreting functional CEA691-specific CTL lines were successfully generated from 10 of 18 pancreatic cancer patients, with two CTL lines able to recognize and kill both CEA691 peptide–loaded T2 cells and CEA+ HLA-A2+ pancreatic cancer cell lines. In the presence of ex vivo PD-L1 blockade, functional CEA691-specific CD8+ T-cell responses, including IFNγ secretion and proliferation, were enhanced, and this effect was more pronounced on Ag-specific T cells isolated from tumor draining lymph nodes.Conclusions: These data demonstrate that CEA691-specific CTL can be readily expanded from the self-restricted T-cell repertoire of pancreatic cancer patients and that their function can be enhanced by PD-L1 blockade. Clin Cancer Res; 23(20); 6178–89. ©2017 AACR.

Purpose: Identification of serum biomarkers enabling earlier diagnosis of pancreatic ductal adenocarcinoma (PDAC) could improve outcome. Serum protein profiles in patients with preclinical disease and at diagnosis were investigated.Experimental Design: Serum from cases up to 4 years prior to PDAC diagnosis and controls (UKCTOCS, n = 174) were studied, alongside samples from patients diagnosed with PDAC, chronic pancreatitis, benign biliary disease, type 2 diabetes mellitus, and healthy subjects (n = 298). Isobaric tags for relative and absolute quantification (iTRAQ) enabled comparisons of pooled serum from a test set (n = 150). Validation was undertaken using multiple reaction monitoring (MRM) and/or Western blotting in all 472 human samples and samples from a KPC mouse model.Results: iTRAQ identified thrombospondin-1 (TSP-1) as reduced preclinically and in diagnosed samples. MRM confirmed significant reduction in levels of TSP-1 up to 24 months prior to diagnosis. A combination of TSP-1 and CA19-9 gave an AUC of 0.86, significantly outperforming both markers alone (0.69 and 0.77, respectively; P < 0.01). TSP-1 was also decreased in PDAC patients compared with healthy controls (P < 0.05) and patients with benign biliary obstruction (P < 0.01). Low levels of TSP-1 correlated with poorer survival, preclinically (P < 0.05) and at clinical diagnosis (P < 0.02). In PDAC patients, reduced TSP-1 levels were more frequently observed in those with confirmed diabetes mellitus (P < 0.01). Significantly lower levels were also observed in PDAC patients with diabetes compared with individuals with type 2 diabetes mellitus (P = 0.01).Conclusions: Circulating TSP-1 levels decrease up to 24 months prior to diagnosis of PDAC and significantly enhance the diagnostic performance of CA19-9. The influence of diabetes mellitus on biomarker behavior should be considered in future studies. Clin Cancer Res; 22(7); 1734–43. ©2015 AACR.

Purpose The registration of Laparoscopic Ultrasound (LUS) to CT can enhance the safety of laparoscopic liver surgery by providing the surgeon with awareness on the relative positioning between critical vessels and a tumour. In an effort to provide a translatable solution for this poorly constrained problem, Content-based Image Retrieval (CBIR) based on vessel information has been suggested as a method for obtaining a global coarse registration without using tracking information. However, the performance of these frameworks is limited by the use of non-generalisable handcrafted vessel features. Methods We propose the use of a Deep Hashing (DH) network to directly convert vessel images from both LUS and CT into fixed size hash codes. During training, these codes are learnt from a patient-specific CT scan by supplying the network with triplets of vessel images which include both a registered and a mis-registered pair. Once hash codes have been learnt, they can be used to perform registration with CBIR methods. Results We test a CBIR pipeline on 11 sequences of untracked LUS distributed across 5 clinical cases. Compared to a handcrafted feature approach, our model improves the registration success rate significantly from 48% to 61%, considering a 20 mm error as the threshold for a successful coarse registration. Conclusions We present the first DH framework for interventional multi-modal registration tasks. The presented approach is easily generalisable to other registration problems, does not require annotated data for training, and may promote the translation of these techniques.

Pancreatic cystic lesions are an increasingly common clinical finding. Current diagnostic techniques cannot reliably differentiate patients with high-risk lesions requiring surgical resection from those that can be safely surveyed or discharged. As a result, some patients may undergo unnecessary surgery with associated morbidity while others enter long-term surveillance with associated healthcare costs. Needle-based confocal laser endomicroscopy enables real time microscopic examination of the epithelial lining of a cyst wall at the time of a standard endoscopic ultrasound examination. The procedure is associated with low rates of adverse events, especially when the probe is loaded into the fine-needle aspiration needle before the procedure and examination times are limited. Needle-based confocal laser endomicroscopy has consistently been shown to have better diagnostic accuracy than cytology, which is often paucicellular and non-diagnostic in pancreatic cystic lesions. Studies have shown that diagnostic accuracy in needle-based confocal laser endomicroscopy is 84–95% in mucinous lesions and 39–99% in serous lesions. However, this technology is expensive and its place in diagnostic algorithms remains uncertain. Despite this, health economic analyses in certain health systems have been favourable, largely because of its potential to be able to discharge patients with benign lesions, such as serous cystic neoplasms, from long-term surveillance. Widespread adoption of this technology is unlikely but it has the potential to have an important role in indeterminate pancreatic cystic lesions.

Background Endoscopic ultrasound guided gastrojejunostomy (EUS-GJ) with lumen apposing metal stents has recently emerged as a viable option, as an alternative to surgical gastrojejunostomy and endoscopic enteral stenting, for managing gastric outlet obstruction (GOO). We aim to perform a retrospective analysis of the efficacy, safety and outcomes of EUS-GJ performed at three tertiary institutions in the United Kingdom. Methods Consecutive patients who underwent EUS-GJ between August 2018 and March 2021 were identified from a prospectively maintained database. Data were obtained from interrogation of electronic health records. Results Twenty five patients (15 males) with a median age of 63 years old (range 29–80) were included for analysis. 88% (22/25) of patients had GOO due to underlying malignant disease. All patients were deemed surgically inoperable or at high surgical risk. Both technical and clinical success were achieved in 92% (23/25) of patients. There was an improvement in the mean Gastric Outlet Obstruction Scoring System scores following a technically successful EUS-GJ (2.52 vs 0.68, p Conclusion EUS-GJ in carefully selected patients is an effective and safe procedure when performed by experienced endoscopists.

Background and Aims Locally advanced pancreatic cancer (LAPC) often causes obstruction. Verteporfin photodynamic therapy (PDT) can feasibly “debulk” the tumor more safely than noncurative surgery and has multiple advantages over older PDT agents. We aimed to assess the feasibility of EUS-guided verteporfin PDT in ablating nonresectable LAPC. Methods Adults with LAPC with adequate biliary drainage were prospectively enrolled. Exclusion criteria were significant metastatic disease burden, disease involving >50% duodenal or major artery circumference, and recent treatment with curative intent. CT was obtained between days –28 to 0. On day 0, verteporfin .4 mg/kg was infused 60 to 90 minutes before EUS, during which a diffuser was positioned in the tumor and delivered light at 50 J/cm for 333 seconds. CT was obtained on day 2, with adverse event monitoring occurring on days 1, 2, and 14. The primary outcome was presence of necrosis. Results Of 8 patients (62.5% men, mean age 65 ± 7.9 years) included in the study, 5 were staged at T3, 2 at T2, and 1 at T1. Most (n = 4) had primary lesions in the pancreatic head. Mean pretrial tumor diameter was 33.3 ± 13.4 mm. On day 2 CT, 5 lesions demonstrated a zone of necrosis measuring a mean diameter of 15.7 ± 5.5 mm; 3 cases did not develop necrosis. No adverse events were noted during the procedure or postprocedure observation period (days 1-3), and no changes in patient-reported outcomes were noted. Conclusions In this pilot study, EUS-guided verteporfin PDT is feasible and shows promise as a minimally invasive ablative therapy for LAPC in select patients. Tumor necrosis is visible within 48 hours after treatment. Patient enrollment and data collection are ongoing. (Clinical trial registration number: NCT03033225.)

Although emerging data evidences that EUS-guided needle-based confocal laser endomicroscopy (nCLE) accurately diagnoses pancreatic cystic lesions (PCLs), there are a lack of interobserver agreement (IOA) studies utilizing reference histopathological diagnosis and for specific PCL subtypes. Hence, we sought to assess the IOA, intra-observer reliability (IOR), and diagnostic performance of EUS-nCLE using a large cohort of patients with histopathological diagnosis amongst a broad panel of international observers.EUS-nCLE videos (n = 76) of subjects with PCLs [intraductal papillary mucinous neoplasm (IPMN), mucinous cystic neoplasm (MCN), serous cystadenoma (SCA), pseudocyst, and cystic-neuroendocrine tumors/solid pseudopapillary neoplasm (cystic-NET/SPN)], simulating clinical prevalence rates were obtained from 3 prospective studies. An international panel of 13 endosonographers with nCLE experience, blinded to all PCL data, evaluated the video library twice with a two-week washout for PCL differentiation (mucinous vs. non-mucinous) and subtype diagnosis.The IOA (κ = 0.82, 95% CI 0.77-0.87) and IOR (κ = 0.82, 95% CI 0.78-0.85) were "almost perfect" to differentiate mucinous vs. non-mucinous PCLs. For PCL subtype, IOA was highest for SCA (almost perfect; κ = 0.85), followed by IPMN (substantial, κ = 0.72), and cystic-NET/SPN (substantial, κ = 0.73). The IOA was moderate for MCN (κ = 0.47), and pseudocyst (κ = 0.57). Compared to histopathology, observers differentiated mucinous vs. non-mucinous PCLs with high accuracy (94.8%, 95% CI 93.3-96.1). For detecting specific PCLs subtypes, EUS-nCLE was highly accurate in diagnosing non-mucinous cysts (SCA: 98%; cystic-NET/SPN: 96%; pseudocyst: 96%) and slightly less accurate for mucinous lesions (IPMN: 86%; MCN: 84%).Diagnosis of PCLs by EUS-nCLE guided virtual biopsy is very accurate and reliable for the most prevalent pancreatic cysts in clinical practice.

Background: Pancreatic cystic lesions (PCL) represent an increasingly diagnosed condition with significant burden to patients’ lives and medical resources. Endoscopic ultrasound (EUS) ablation techniques have been utilized to treat focal pancreatic lesions. This systematic review with meta-analysis aims to assess the efficacy of EUS ablation on PCL in terms of complete or partial response and safety. Methods: A systematic search in Medline, Cochrane and Scopus databases was performed in April 2023 for studies assessing the performance of the various EUS ablation techniques. The primary outcome was complete cyst resolution, defined as cyst disappearance in follow-up imaging. Secondary outcomes included partial resolution (reduction in PCL size), and adverse events rate. A subgroup analysis was planned to evaluate the impact of the available ablation techniques (ethanol, ethanol/paclitaxel, radiofrequency ablation (RFA), and lauromacrogol) on the results. Meta-analyses using a random effects model were conducted and the results were reported as percentages with 95% confidence intervals (95%CI). Results: Fifteen studies (840 patients) were eligible for analysis. Complete cyst resolution after EUS ablation was achieved in 44% of cases (95%CI: 31–57; 352/767; I2 = 93.7%), and the respective partial response rate was 30% (95%CI: 20–39; 206/767; I2 = 86.1%). Adverse events were recorded in 14% (95%CI: 8–20; 164/840; I2 = 87.2%) of cases, rated as mild in 10% (95%CI: 5–15; 128/840; I2 = 86.7%), and severe in 4% (95%CI: 3–5; 36/840; I2 = 0%). The subgroup analysis for the primary outcome revealed rates of 70% (95%CI: 64–76; I2 = 42.3%) for ethanol/paclitaxel, 44% (95%CI: 33–54; I2= 0%) for lauromacrogol, 32% (95%CI: 27–36; I2 = 88.4%) for ethanol, and 13% (95%CI: 4–22; I2 = 95.8%) for RFA. Considering adverse events, the ethanol-based subgroup rated the highest percentage (16%; 95%CI: 13–20; I2 = 91.0%). Conclusion: EUS ablation of pancreatic cysts provides acceptable rates of complete resolution and a low incidence of severe adverse events, with chemoablative agents yielding higher performance rates.

There is an increasing demand and availability of bariatric surgery, with a range of procedures performed, some leading to altered upper gastrointestinal anatomy. The patient population undergoing bariatric surgery is also at increased risk of gallstones and biliary stone disease. Endoscopy (ie, endoscopic retrograde cholangiopancreatography) is the cornerstone of management of biliary stone disease, but may be challenging after bariatric surgery. In this review the endoscopic, surgery assisted, or percutaneous options that may be considered are discussed, based on the details of surgical anatomy and available expertise.

Summary of statementsIn Europe at present, but also in 2040, 1 in 3 cancer-related deaths are expected to be caused by digestive cancers. Endoscopic technologies enable diagnosis, with relatively low invasiveness, of precancerous conditions and early cancers, thereby improving patient survival. Overall, endoscopy capacity must be adjusted to facilitate both effective screening programs and rigorous control of the quality assurance and surveillance systems required. 1 For average-risk populations, ESGE recommends the implementation of organized population-based screening programs for colorectal cancer, based on fecal immunochemical testing (FIT), targeting individuals, irrespective of gender, aged between 50 and 75 years. Depending on local factors, namely the adherence of the target population and availability of endoscopy services, primary screening by colonoscopy or sigmoidoscopy may also be recommendable. 2 In high-risk populations, endoscopic screening for gastric cancer should be considered for individuals aged more than 40 years. Its use in countries/regions with intermediate risk may be considered on the basis of local settings and availability of endoscopic resources. 3 For esophageal and pancreatic cancer, endoscopic screening may be considered only in high-risk individuals:– For squamous cell carcinoma, in those with a personal history of head/neck cancer, achalasia, or previous caustic injury; – For Barrett’s esophagus (BE)-associated adenocarcinoma, in those with long-standing gastroesophageal reflux disease symptoms (i. e., > 5 years) and multiple risk factors (age ≥ 50 years, white race, male sex, obesity, first-degree relative with BE or esophageal adenocarcinoma [EAC]). – For pancreatic cancer screening, endoscopic ultrasound may be used in selected high-risk patients such as those with a strong family history and/or genetic susceptibility.

International audience; Cholangiocarcinoma (CCA) is a malignant tumour arising from the biliary system. In Europe, this tumour frequently presents as a sporadic cancer in patients without defined risk factors and is usually diagnosed at advanced stages with a consequent poor prognosis. Therefore, the identification of biomarkers represents an utmost need for patients with CCA. Numerous studies proposed a wide spectrum of biomarkers at tissue and molecular levels. With the present paper, a multidisciplinary group of experts within the European Network for the Study of Cholangiocarcinoma discusses the clinical role of tissue biomarkers and provides a selection based on their current relevance and potential applications in the framework of CCA. Recent advances are proposed by dividing biomarkers based on their potential role in diagnosis, prognosis and therapy response. Limitations of current biomarkers are also identified, together with specific promising areas (ie, artificial intelligence, patient-derived organoids, targeted therapy) where research should be focused to develop future biomarkers.

BackgroundPancreatic cancer continues to have an extremely poor prognosis in part due to late diagnosis. 25% of pancreatic cancer patients have a prior diagnosis of diabetes, and hence identifying individuals at risk of pancreatic cancer in those with recently diagnosed type 2 diabetes may be a useful opportunity to identify candidates for screening and early detection. In this study, we will comparatively evaluate regression and machine learning-based clinical prediction models for estimating individual risk of developing pancreatic cancer two years after type 2 diabetes diagnosis.MethodsIn the development dataset, we will include adults aged 30-84 years with incident type-2 diabetes registered with QResearch primary care database. Patients will be followed up from type-2 diabetes diagnosis to first diagnosis of pancreatic cancer as recorded in any one of primary care records, hospital episode statistics, cancer registry data, or death records. Cox-proportional hazards models will be used to develop a risk prediction model for estimating individual risk of developing pancreatic cancer during up to 2 years of follow-up. We will perform variable selection using a combination of clinical and statistical significance approach i.e. HR 1.1 and pInternal-external cross-validation (IECV) framework across geographical regions within England will be used to assess performance and pooled using random effects meta-analysis using: (i) model fit in terms of variation explained by the model Royston & Sauerbrei’s R2D, (ii) calibration slope and calibration-in-the-large, and (iii) discrimination measured in terms of Harrell’s C and Royston & Sauerbrei’s D-statistic.Further, we will evaluate machine learning (ML) approaches for the clinical prediction model using neural networks (NN) and XGBoost. The model predictors and performance of these will be compared with the results of those derived from the regression-based strategy.DiscussionThe proposed study will develop and validate a novel risk prediction model to aid early diagnosis of pancreatic cancer in patients with new-onset diabetes in primary care. With an enhanced decision-risk tool for use at point-of care by general practitioners to assess pancreatic cancer risk, it may improve decision-making so that at-risk patients are rapidly prioritised to aid early diagnosis of pancreatic cancer in patients with newly diagnosed diabetes.

Earlier detection of pancreatic ductal adenocarcinoma (PDAC) is key to improving patient outcomes, as it is mostly detected at advanced stages which are associated with poor survival. Developing non-invasive blood tests for early detection would be an important breakthrough. The primary objective of the work presented here was to use a unique dataset, that is both large and prospectively collected, to quantify a set of 96 cancer-associated proteins and construct multi-marker models with the capacity to accurately predict PDAC years before diagnosis. The data is part of a nested case control study within UK Collaborative Trial of Ovarian Cancer Screening and is comprised of 219 samples, collected from a total of 143 post-menopausal women who were diagnosed with pancreatic cancer within 70 months after sample collection, and 248 matched non-cancer controls. We developed a stacked ensemble modelling technique to achieve robustness in predictions and, therefore, improve performance in newly collected datasets. With a pool of 10 base-learners and a Bayesian averaging meta-learner, we can predict PDAC status with an AUC of 0.91 (95% CI 0.75 - 1.0), sensitivity of 92% (95% CI 0.54 - 1.0) at 90% specificity, up to 1 year to diagnosis, and at an AUC of 0.85 (95% CI 0.74 - 0.93) up to 2 years to diagnosis (sensitivity of 61%, 95 % CI 0.17 - 0.83, at 90% specificity). These models also use clinical covariates such as hormone replacement therapy use (at randomization), oral contraceptive pill use (ever) and diabetes and outperform biomarker combinations cited in the literature.

Ultrasound imaging is a commonly used technology for visualising patient anatomy in real-time during diagnostic and therapeutic procedures. High operator dependency and low reproducibility make ultrasound imaging and interpretation challenging with a steep learning curve. Automatic image classification using deep learning has the potential to overcome some of these challenges by supporting ultrasound training in novices, as well as aiding ultrasound image interpretation in patient with complex pathology for more experienced practitioners. However, the use of deep learning methods requires a large amount of data in order to provide accurate results. Labelling large ultrasound datasets is a challenging task because labels are retrospectively assigned to 2D images without the 3D spatial context available in vivo or that would be inferred while visually tracking structures between frames during the procedure. In this work, we propose a multi-modal convolutional neural network (CNN) architecture that labels endoscopic ultrasound (EUS) images from raw verbal comments provided by a clinician during the procedure. We use a CNN composed of two branches, one for voice data and another for image data, which are joined to predict image labels from the spoken names of anatomical landmarks. The network was trained using recorded verbal comments from expert operators. Our results show a prediction accuracy of 76% at image level on a dataset with 5 different labels. We conclude that the addition of spoken commentaries can increase the performance of ultrasound image classification, and eliminate the burden of manually labelling large EUS datasets necessary for deep learning applications., Submitted to IEEE TMI