Search

Competing Interests: R. H. Flowers does/has performed investigational work for Concert and Venthera. The remaining authors declare no conflict of interest.

Competing Interests: R. H. Flowers does/has done investigational work for Concert and Venthera. The other authors declare no conflict of interest.

Pathologists are responsible for rapidly providing a diagnosis on critical health issues. Challenging cases benefit from additional opinions of pathologist colleagues. In addition to on-site colleagues, there is an active worldwide community of pathologists on social media for complementary opinions. Such access to pathologists worldwide has the capacity to improve diagnostic accuracy and generate broader consensus on next steps in patient care. From Twitter we curate 13,626 images from 6,351 tweets from 25 pathologists from 13 countries. We supplement the Twitter data with 113,161 images from 1,074,484 PubMed articles. We develop machine learning and deep learning models to (i) accurately identify histopathology stains, (ii) discriminate between tissues, and (iii) differentiate disease states. Area Under Receiver Operating Characteristic (AUROC) is 0.805-0.996 for these tasks. We repurpose the disease classifier to search for similar disease states given an image and clinical covariates. We report precision@k = 1 = 0.7618 ± 0.0018 (chance 0.397 ± 0.004, mean ±stdev ). The classifiers find that texture and tissue are important clinico-visual features of disease. Deep features trained only on natural images (e.g., cats and dogs) substantially improved search performance, while pathology-specific deep features and cell nuclei features further improved search to a lesser extent. We implement a social media bot (@pathobot on Twitter) to use the trained classifiers to aid pathologists in obtaining real-time feedback on challenging cases. If a social media post containing pathology text and images mentions the bot, the bot generates quantitative predictions of disease state (normal/artifact/infection/injury/nontumor, preneoplastic/benign/low-grade-malignant-potential, or malignant) and lists similar cases across social media and PubMed. Our project has become a globally distributed expert system that facilitates pathological diagnosis and brings expertise to underserved regions or hospitals with less expertise in a particular disease. This is the first pan-tissue pan-disease (i.e., from infection to malignancy) method for prediction and search on social media, and the first pathology study prospectively tested in public on social media. We will share data through http://pathobotology.org . We expect our project to cultivate a more connected world of physicians and improve patient care worldwide.

Purpose:  Clinically significant portal hypertension (CSPH) is responsible for most of the complications in patients with cirrhosis. Liver stiffness (LS) measurement by vibration-controlled transient elastography (VCTE) is currently used to evaluate CSPH. Bi-dimensional shear wave elastography from General Electric (2D-SWE.GE) has not yet been validated for the diagnosis of PHT. Our aims were to test whether 2D-SWE.GE-LS is able to evaluate CSPH, to determine the reliability criteria of the method and to compare its accuracy with that of VCTE-LS in this clinical setting., Materials and Methods:  Patients with chronic liver disease referred to hepatic catheterization (HVPG) were consecutively enrolled. HVPG and LS by both VCTE and 2D-SWE.GE were performed on the same day. The diagnostic performance of each LS method was compared against HVPG and between each other., Results:  2D-SWE.GE-LS was possible in 123/127 (96.90 %) patients. The ability to record at least 5 LS measurements by 2D-SWE.GE and IQR < 30 % were the only features associated with reliable results. 2D-SWE.GE-LS was highly correlated with HVPG (r = 0.704; p < 0.0001), especially if HVPG < 10 mmHg and was significantly higher in patients with CSPH (15.52 vs. 8.14 kPa; p < 0.0001). For a cut-off value of 11.3 kPa, the AUROC of 2D-SWE.GE-LS to detect CSPH was 0.91, which was not inferior to VCTE-LS (0.92; p = 0.79). The diagnostic accuracy of LS by 2D-SWE.GE-LS to detect CSPH was similar with the one of VCTE-LS (83.74 % vs. 85.37 %; p = 0.238). The diagnostic accuracy was not enhanced by using different cut-off values which enhanced the sensitivity or the specificity. However, in the subgroup of compensated patients with alcoholic liver disease, 2D-SWE.GE-LS classified CSPH better than VCTE-LS (93.33 % vs. 85.71 %, p = 0.039)., Conclusion:  2D-SWE.GE-LS has good accuracy, not inferior to VCTE-LS, for the diagnosis of CSPH., Competing Interests: The authors declare that they have no conflict of interest., (Thieme. All rights reserved.)

Background & Aims: Several non-invasive tests (NITs) have been developed to diagnose oesophageal varices (EV), including the recent Baveno VI criteria to rule out high-risk varices (HRV). Spleen stiffness measurement (SSM) with the standard FibroScan® (SSM@50Hz) has been evaluated. However, the EV grading could be underestimated because of a ceiling threshold (75 kPa) of the SSM@50Hz. The aims were to evaluate SSM by a novel spleen-dedicated FibroScan® (SSM@100Hz) for EV diagnosis compared with SSM@50Hz, other validated NITs and Baveno VI criteria., Methods: This prospective multicentre study consecutively enrolled patients with chronic liver disease; blood data, endoscopy, liver stiffness measurement (LSM), SSM@50Hz and SSM@100Hz were collected., Results: Two hundred and sixty patients met inclusion criteria. SSM@100Hz success rate was significantly higher than that of SSM@50Hz (92.5% vs 76.0%, P < .001). SSM@100Hz accuracy for the presence of EV (AUC = 0.728) and HRV (AUC = 0.756) was higher than in other NITs. SSM@100Hz AUC for large EV (0.782) was higher than SSM@50Hz (0.720, P = .027). AUC for HRV with SSM@100Hz (0.780) was higher than with LSM (0.615, P < .001). The spared endoscopy rate of Baveno VI criteria (8.1%) was significantly increased by the combination to SSM@50Hz (26.5%) or SSM@100Hz (38.9%, P < .001 vs others). The missed HRV rate was, respectively, 0% and 4.7% for combinations., Conclusions: SSM@100Hz is a new performant non-invasive marker for EV and HRV providing a higher accuracy than SSM@50Hz and other NITs. The combination of Baveno VI criteria and SSM@100Hz significantly increased the spared endoscopy rate compared to Baveno VI criteria alone or combined with SSM@50Hz. Clinical trial number: NCT02180113., (© 2019 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd.)

Background: In resource-limited settings, genotype testing at virologic failure on first-line antiretroviral therapy (ART) may identify patients with wild-type (WT) virus. After adherence counseling, these patients may safely and effectively continue first-line ART, thereby delaying more expensive second-line ART., Methods: We used the Cost-Effectiveness of Preventing AIDS Complications International model of human immunodeficiency virus (HIV) disease to simulate a South African cohort of HIV-infected adults at first-line ART failure. Two strategies were examined: no genotype vs genotype, assuming availability of protease inhibitor-based second-line ART. Model inputs at first-line ART failure were mean age 38 years, mean CD4 173/µL, and WT virus prevalence 20%; genotype cost was $300 per test and delay to results, 3 months. Outcomes included life expectancy, per-person costs (2010 US dollars), and incremental cost-effectiveness ratios (dollars per years of life saved [YLS])., Results: No genotype had a projected life expectancy of 106.1 months, which with genotype increased to 108.3 months. Per-person discounted lifetime costs were $16 360 and $16 540, respectively. Compared to no genotype, genotype was very cost-effective, by international guidance, at $900/YLS. The cost-effectiveness of genotype was sensitive to prevalence of WT virus (very cost-effective when prevalence ≥ 12%), CD4 at first-line ART failure, and ART efficacy. Genotype-associated delays in care ≥ 5 months decreased survival and made no genotype the preferred strategy. When the test cost was <$100, genotype became cost-saving., Conclusions: Genotype resistance testing at first-line ART failure is very cost-effective in South Africa. The cost-effectiveness of this strategy will depend on prevalence of WT virus and timely response to genotype results.

Background: In settings with high tuberculosis (TB) prevalence, 15-30% of HIV-infected individuals initiating antiretroviral therapy (ART) have undiagnosed TB. Such patients are usually screened by symptoms and sputum smear, which have poor sensitivity., Objective: To project the clinical and economic outcomes of using Xpert MTB/RIF(Xpert), a rapid TB/rifampicin-resistance diagnostic, to screen individuals initiating ART., Design: We used a microsimulation model to evaluate the clinical impact and cost-effectiveness of alternative TB screening modalities - in all patients or only symptomatic patients - for hypothetical cohorts of individuals initiating ART in South Africa (mean CD4 cell count = 171 cells/μl; TB prevalence 22%). We simulated no active screening and four diagnostic strategies, smear microscopy (sensitivity 23%); smear and culture (sensitivity, 100%); one Xpert sample (sensitivity in smear-negative TB: 43%); two Xpert samples (sensitivity in smear-negative TB: 62%). Outcomes included projected life expectancy, lifetime costs (2010 US$), and incremental cost-effectiveness ratios (ICERs). Strategies with ICERs less than $7100 (South African gross domestic product per capita) were considered very cost-effective., Results: Compared with no screening, life expectancy in TB-infected patients increased by 1.6 months using smear in symptomatic patients and by 6.6 months with two Xpert samples in all patients. At 22% TB prevalence, the ICER of smear for all patients was $2800 per year of life saved (YLS), and of Xpert (two samples) for all patients was $5100/YLS. Strategies involving one Xpert sample or symptom screening were less efficient., Conclusion: Model-based analysis suggests that screening all individuals initiating ART in South Africa with two Xpert samples is very cost-effective.