Convolutional Neural Network for Automated FLAIR Lesion Segmentation on Clinical Brain MR Imaging

BACKGROUND AND PURPOSE

Most brain lesions are characterized by hyperintense signal on FLAIR. We sought to develop an automated deep learning-based method for segmentation of abnormalities on FLAIR and volumetric quantification on clinical brain MRIs across many pathologic entities and scanning parameters. We evaluated the performance of the algorithm compared with manual segmentation and existing automated methods.

MATERIALS AND METHODS

We adapted a U-Net convolutional neural network architecture for brain MRIs using 3D volumes. This network was retrospectively trained on 295 brain MRIs to perform automated FLAIR lesion segmentation. Performance was evaluated on 92 validation cases using Dice scores and voxelwise sensitivity and specificity, compared with radiologists' manual segmentations. The algorithm was also evaluated on measuring total lesion volume.

RESULTS

Our model demonstrated accurate FLAIR lesion segmentation performance (median Dice score, 0.79) on the validation dataset across a large range of lesion characteristics. Across 19 neurologic diseases, performance was significantly higher than existing methods (Dice, 0.56 and 0.41) and approached human performance (Dice, 0.81). There was a strong correlation between the predictions of lesion volume of the algorithm compared with true lesion volume (ρ = 0.99). Lesion segmentations were accurate across a large range of image-acquisition parameters on >30 different MR imaging scanners.

CONCLUSIONS

A 3D convolutional neural network adapted from a U-Net architecture can achieve high automated FLAIR segmentation performance on clinical brain MR imaging across a variety of underlying pathologies and image acquisition parameters. The method provides accurate volumetric lesion data that can be incorporated into assessments of disease burden or into radiologic reports.

Social Media Research Strategy to Understand Clinician and Public Perception of Health Care Messages

Introduction:

Despite the potential of social media to influence public health and generate insights, the process of monitoring and analyzing the dissemination of health care messages on social media has been described as difficult and in need of improvement.

Objectives:

The objective of this study was to describe and demonstrate a reproducible methodology for cataloging and analyzing health care–related social media comments and provide insight into how clinicians and members of the general public respond to health care messaging on social media.

Methods:

We collected social media comments related to the American Dental Association’s 2016 “Evidence-Based Clinical Practice Guideline for the Use of Pit-and-Fissure Sealants” between April 10, 2017, and October 31, 2017, from Facebook, Twitter, LinkedIn, Reddit, and online message boards for the New York Times, FiveThirtyEight, and Dentaltown. Using data provided in the comments, we conducted engagement analysis as well as content, network, and sentiment analysis across 8 categories.

Results:

We collected 671 comments. Among our findings, Facebook (472 of 671) was the most popular platform among commentators; almost half of all comments (335 of 671) aligned with the recommendations of the 2016 American Dental Association sealants guideline; clinicians were more likely than the general public to like a comment that suggested an improvement to the guideline; and >75% of comments (521 of 671) were supported by anecdotal evidence.

Conclusion:

As the prevalence of anecdotes on social media suggests, the likelihood of falsehoods spreading on social media is high. Insights gleaned from the methodology described in this research could help combat the spread of such misinformation by providing disseminators of health care messaging with insight into their target audiences. Armed with this knowledge, disseminators can craft health care messages that more effectively engage clinicians and the general public.

Knowledge Transfer Statement:

The methodology used in this research provides a reproducible strategy for tracking social media engagement with health care messages. Engagement results can assist future delivery of health care messages to key stakeholders and ensure better implementation and adoption of these communications.

Highly Pathogenic Avian Influenza A(H5N1) Viruses at the Animal-Human Interface in Vietnam, 2003-2010

Mutation and reassortment of highly pathogenic avian influenza A(H5N1) viruses at the animal-human interface remain a major concern for emergence of viruses with pandemic potential. To understand the relationship of H5N1 viruses circulating in poultry and those isolated from humans, comprehensive phylogenetic and molecular analyses of viruses collected from both hosts in Vietnam between 2003 and 2010 were performed. We examined the temporal and spatial distribution of human cases relative to H5N1 poultry outbreaks and characterized the genetic lineages and amino acid substitutions in each gene segment identified in humans relative to closely related viruses from avian hosts. Six hemagglutinin clades and 8 genotypes were identified in humans, all of which were initially identified in poultry. Several amino acid mutations throughout the genomes of viruses isolated from humans were identified, indicating the potential for poultry viruses infecting humans to rapidly acquire molecular markers associated with mammalian adaptation and antiviral resistance.

Assessing evidence for avian-to-human transmission of influenza A/H9N2 virus in rural farming communities in northern Vietnam

Rural farming communities in northern Vietnam do not routinely practice vaccination for influenza A viruses (IAV) for either humans or poultry, which enables us to study transmission intensity via seroepidemiology. Using samples from a longitudinal cohort of farming households, we determined the number of symptomatic and asymptomatic human infections for seasonal IAV and avian A/H9 over 2 years. As expected, we detected virologically confirmed acute cases of seasonal IAV in humans, as well as large numbers of subclinical seroconversions to A/H1pdm [55/265 (21 %)], A/H3 [95/265 (36 %)] and A/H9 [24/265 (9 %)]. Five of the A/H9 human seroconverters likely represented true infections rather than heterosubtypic immunity, because the individuals seroconverted solely to A/H9. Among co-located poultry, we found significantly higher seroprevalance for A/H5 compared to A/H9 in both chickens and ducks [for northern study sites overall, 337/1105 (30.5 %) seropositive for A/H5 and 123/1105 (11.1 %) seropositive for A/H9].

[D-Pen2,4'-125I-Phe4,D-Pen5]enkephalin: a selective high affinity radioligand for delta opioid receptors with exceptional specific activity

[D-Pen2,4'-125I-Phe4,D-Pen5]enkephalin ([125I]DPDPE) is a highly selective radioligand for the delta opioid receptor with a specific activity (2200 Ci/mmol) that is over 50-fold greater than that of tritium-labeled DPDPE analogs. [125I]DPDPE binds to a single site in rat brain membranes with an equilibrium dissociation constant (Kd) value of 421 +/- 67 pM and a receptor density (Bmax) value of 36.4 +/- 2.7 fmol/mg protein. The high affinity of this site for delta opioid receptor ligands and its low affinity for mu or kappa receptor-selective ligands are consistent with its being a delta opioid receptor. The distribution of these sites in rat brain, observed by receptor autoradiography, is also consistent with that of delta opioid receptors. Association and dissociation binding kinetics of 1.0 nM [125I] DPDPE are monophasic at 25 degrees C. The association rate (k + 1 = 5.80 +/- 0.88 X 10(7) M-1 min-1) is about 20- and 7-fold greater than that measured for 1.0 nM [3H DPDPE and 0.8 nM [3H] [D-Pen2,4'-Cl-Phe4, D-Pen5]enkephalin, respectively. The dissociation rate of [125I]DPDPE (0.917 +/- 0.117 X 10(-2) min-1) measured at 1.0 nM is about 3-fold faster than is observed for either of the other DPDPE analogs. The rapid binding kinetics of [125I]DPDPE is advantageous because binding equilibrium is achieved with much shorter incubation times than are required for other cyclic enkephalin analogs. This, in addition to its much higher specific activity, makes [125I]DPDPE a valuable new radioligand for studies of delta opioid receptors.