Interaction between clinicians and artificial intelligence to detect fetal atrioventricular septal defects on ultrasound: how can we optimize collaborative performance?

OBJECTIVES: Artificial intelligence (AI) has shown promise in improving the performance of fetal ultrasound screening in detecting congenital heart disease (CHD). The effect of giving AI advice to human operators has not been studied in this context. Giving additional information about AI model workings, such as confidence scores for AI predictions, may be a way of further improving performance. Our aims were to investigate whether AI advice improved overall diagnostic accuracy (using a single CHD lesion as an exemplar), and to determine what, if any, additional information given to clinicians optimized the overall performance of the clinician-AI team. METHODS: An AI model was trained to classify a single fetal CHD lesion (atrioventricular septal defect (AVSD)), using a retrospective cohort of 121 130 cardiac four-chamber images extracted from 173 ultrasound scan videos (98 with normal hearts, 75 with AVSD); a ResNet50 model architecture was used. Temperature scaling of model prediction probability was performed on a validation set, and gradient-weighted class activation maps (grad-CAMs) produced. Ten clinicians (two consultant fetal cardiologists, three trainees in pediatric cardiology and five fetal cardiac sonographers) were recruited from a center of fetal cardiology to participate. Each participant was shown 2000 fetal four-chamber images in a random order (1000 normal and 1000 AVSD). The dataset comprised 500 images, each shown in four conditions: (1) image alone without AI output; (2) image with binary AI classification; (3) image with AI model confidence; and (4) image with grad-CAM image overlays. The clinicians were asked to classify each image as normal or AVSD. RESULTS: A total of 20 000 image classifications were recorded from 10 clinicians. The AI model alone achieved an accuracy of 0.798 (95% CI, 0.760-0.832), a sensitivity of 0.868 (95% CI, 0.834-0.902) and a specificity of 0.728 (95% CI, 0.702-0.754), and the clinicians without AI achieved an accuracy of 0.844 (95% CI, 0.834-0.854), a sensitivity of 0.827 (95% CI, 0.795-0.858) and a specificity of 0.861 (95% CI, 0.828-0.895). Showing a binary (normal or AVSD) AI model output resulted in significant improvement in accuracy to 0.865 (P < 0.001). This effect was seen in both experienced and less-experienced participants. Giving incorrect AI advice resulted in a significant deterioration in overall accuracy, from 0.761 to 0.693 (P < 0.001), which was driven by an increase in both Type-I and Type-II errors by the clinicians. This effect was worsened by showing model confidence (accuracy, 0.649; P < 0.001) or grad-CAM (accuracy, 0.644; P < 0.001). CONCLUSIONS: AI has the potential to improve performance when used in collaboration with clinicians, even if the model performance does not reach expert level. Giving additional information about model workings such as model confidence and class activation map image overlays did not improve overall performance, and actually worsened performance for images for which the AI model was incorrect. © 2024 The Authors. Ultrasound in Obstetrics & Gynecology published by John Wiley & Sons Ltd on behalf of International Society of Ultrasound in Obstetrics and Gynecology.

The case for information fiduciaries: The implementation of a data ethics checklist at Seattle Children's Hospital.

There is little debate about the importance of ethics in health care, and clearly defined rules, regulations, and oaths help ensure patients' trust in the care they receive. However, standards are not as well established for the data professions within health care, even though the responsibility to treat patients in an ethical way extends to the data collected about them. Increasingly, data scientists, analysts, and engineers are becoming fiduciarily responsible for patient safety, treatment, and outcomes, and will require training and tools to meet this responsibility. We developed a data ethics checklist that enables users to consider the possible ethical issues that arise from the development and use of data products. The combination of ethics training for data professionals, a data ethics checklist as part of project management, and a data ethics committee holds potential for providing a framework to initiate dialogues about data ethics and can serve as an ethical touchstone for rapid use within typical analytic workflows, and we recommend the use of this or equivalent tools in deploying new data products in hospitals.

Demonstration of Scale-Invariant Rayleigh-Taylor Instability Growth in Laser-Driven Cylindrical Implosion Experiments.

Rayleigh-Taylor instability growth is shown to be hydrodynamically scale invariant in convergent cylindrical implosions for targets that varied in radial dimension and implosion timescale by a factor of 3. The targets were driven directly by laser irradiation providing a short impulse, and instability growth at an embedded aluminum interface occurs as it converges radially inward by a factor of 2.25 and decelerates on a central foam core. Late-time growth factors of 14 are observed for a single-mode m=20 azimuthal perturbation at both scales, despite the differences in laser drive conditions between the experimental facilities, consistent with predictions from radiation-hydrodynamics simulations. This platform enables detailed investigations into the limits of hydrodynamic scaling in high-energy-density systems.

But what do participants want? Comment on the "Data Sharing in Psychology" special section (2018).

This commentary addresses a recent special section on data sharing (i.e., open data) in the February-March 2018 American Psychologist. In 4 articles, the authors outline how open data can positively impact psychology and provide guidelines for adopting open data practices, which we believe is to be commended. However, this special issue has not acknowledged a crucial concern in the open data debate: the views and desires of participants. Participants are the backbone of psychological research and an important stakeholder in open data issues. We review research that has studied participants' opinions of open data and outline concerns regarding open data raised by some groups of participants. We conclude with recommendations, including a call to psychological researchers to move beyond opinion and instead to empirically examine the impact of open data. We believe psychology is a discipline uniquely poised to execute these recommendations and guide researchers' understandings of how to appropriately and ethically implement open data practices across multiple disciplines. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

Enteral feeding in duct-dependent congenital heart disease.

BACKGROUND: Infants with duct-dependent congenital heart lesions are treated with a prostaglandin E1 infusion. We aimed to describe the feeding strategies used at our institution in such infants, and to describe the incidence of necrotising enterocolitis (NEC) in this patient group, investigating whether enteral feeding is associated with a higher risk. METHODS: Patients diagnosed with hypoplastic left heart syndrome, coarctation of the aorta, pulmonary atresia, or transposition of the great arteries born over a defined period were identified. Premature infants, those with pre-existing gastrointestinal disease, and those who never received prostaglandin were excluded. Data were compared using univariable and multivariable logistic regression models. RESULTS: A total of 177 patients were identified, of them 18 received a diagnosis of suspected or confirmed NEC. There was no association between the diagnosis of NEC and enteral feeding (P = 0.9). CONCLUSIONS: Based on these data, there does not appear to be an association between enteral feeding and NEC in infants receiving prostaglandin.

Development of a survivorship needs assessment planning tool for head and neck cancer survivors and their caregivers: a preliminary study.

PURPOSE: The purpose of this study was to characterize primary end-of-treatment challenges in head and neck cancer (HNC) to drive the development of a survivorship needs assessment planning (SNAP) tool and evaluate its acceptability and feasibility. METHODS: Using qualitative methods (focus groups, interviews), we identified physical, emotional, and social post-treatment challenges from the perspectives of survivors (N = 17), caregivers (N = 14), and healthcare providers (N = 14) and pretested the SNAP tool. After Advisory Board ratings and consensus, the tool was finalized. RESULTS: Survivors, caregivers and clinicians consistently highlighted the importance of assessing symptoms and functional abilities (e.g., dry mouth, speech/swallowing difficulties, weight loss), health behaviors (e.g., smoking, alcohol), emotional concerns (e.g., depression, isolation, nutritional distress), and social challenges (e.g., support, finances). Caregivers were overwhelmed and intensely focused on survivors' nutrition and trach/feeding tube care while clinicians emphasized financial and access concerns. Most participants were enthusiastic about the tool and directed a flexible care plan design due to variability in dyad needs. Over 75% reported high comfort using and navigating questions on a tablet and were in strong agreement that the care plan would help families practically and emotionally. Coordination of survivorship visits with follow-up care was critical to address travel and time barriers. While survivors and clinicians recommended waiting 1-6 months after treatment, caregivers preferred earlier survivorship visits. CONCLUSIONS: Results pinpointed optimal end-of-treatment domains for routine assessment and support the feasibility of implementing a SNAP tool in the clinic. IMPLICATIONS FOR CANCER SURVIVORS: Capitalizing on technology to direct HNC survivorship care is promising.

Evaluating New Skill Sets for Performance Improvement: Human Factors as a Case Study.

New skill sets and improvement disciplines are constantly arising across the vast industrial and academic landscape of modern economies. Prescient hospital administrators are routinely searching for new and innovative ways to improve care, care delivery, safety, quality, and access. But, it can be challenging to identify those emerging skill sets, which will likely have lasting effect and will provide strong return on investment, from passing fads with little capacity to move performance benchmarks for a hospital. Here, we present a rubric for investigating new skill sets, using The Children's Hospital of Philadelphia's investigation into human factors engineering as a case study, and determining whether they can support hospital operations and improvements while providing sufficient return to justify the expense and challenge of incorporating ideas and methods into a quality and performance improvement environment.

Increased uptake and improved outcomes of bowel cancer screening with a faecal immunochemical test: results from a pilot study within the national screening programme in England.

BACKGROUND: The National Health Service Bowel Cancer Screening Programme (BCSP) in England uses a guaiac-based faecal occult blood test (gFOBt). A quantitative faecal immunochemical test (FIT) for haemoglobin (Hb) has many advantages, including being specific for human blood, detecting Hb at a much lower concentration with a single faecal sample and improved uptake. METHODS: In 2014, a large comparative pilot study was performed within BCSP to establish the acceptability and diagnostic performance of FIT. Over a 6-month period, 40 930 (1 in 28) subjects were sent a FIT (OC-SENSOR) instead of a gFOBt. A bespoke FIT package was used to mail FIT sampling devices to and from FIT subjects. All participants positive with either gFOBt or FIT (cut-off 20 µg Hb/g faeces) were referred for follow-up. Subgroup analysis included cut-off concentrations, age, sex, screening history and deprivation quintile. RESULTS: While overall uptake increased by over 7 percentage points with FIT (66.4% vs 59.3%, OR 1.35, 95% CI 1.33 to 1.38), uptake by previous non-responders almost doubled (FIT 23.9% vs gFOBt 12.5%, OR 2.20, 95% CI 2.10 to 2.29). The increase in overall uptake was significantly higher in men than women and was observed across all deprivation quintiles. With the conventional 20 µg/g cut-off, FIT positivity was 7.8% and ranged from 5.7% in 59-64-year-old women to 11.1% in 70-75-year-old men. Cancer detection increased twofold and that for advanced adenomas nearly fivefold. Detection rates remained higher with FIT for advanced adenomas, even at 180 µg Hb/g. CONCLUSIONS: Markedly improved participation rates were achieved in a mature gFOBt-based national screening programme and disparities between men and women were reduced. High positivity rates, particularly in men and previous non-respondents, challenge the available colonoscopy resource, but improvements in neoplasia detection are still achievable within this limited resource.

Managing Disruptions to Patient Flow Capacity: Rapid-Cycle Improvement in a Pediatric Cardiac Procedure Complex.

BACKGROUND: Managing service disruptions is a challenge in every health care environment. Discrete event simulation (DES)--a computer modeling tool used to build in silico (that is, in a digital computer) testbeds for potential changes in complex systems--has been deployed in health care for research and quality improvement (QI), specifically in surgical suite management. A strategy for managing a 6-week planned service disruption needed to be enacted 12 weeks after the announcement, in late October 2014, of the closure of the Hybrid Suite (operating room/catheterization laboratory) for renovation, at The Children's Hospital of Philadelphia's Cardiac Center's Cardiac Operative and Imaging Complex (COIC). METHODS: A previously developed DES was queried to determine theoretical system throughput capacity during the temporary disruption. On the basis of this analysis, a rapid improvement event (RIE) was enacted to address systemic challenges to meeting demand with diminished capacity. During the RIE, system stakeholders (physicians, nurses, and technicians) engaged with performance improvement personnel to identify potential improvements, test those changes in rapid succession, and then implement successful candidates for the disruption. RESULTS: First-case start time was 43 minutes earlier during the period of diminished capacity. Turnaround time between cases was reduced by 23 minutes. Length of day increased by 1 hour, in accordance with simulated predictions. System throughput was 138 patients during the disruption, compared with 135 patients during the same period the previous year. CONCLUSION: A combination of systems analysis and QI methodologies enabled the Cardiac Center to meet demand during a six-week period of diminished capacity. Planned, temporary service disruptions, which must be managed by clinical personnel, can be addressed proactively with promising results.

Simultaneous neutron and x-ray imaging of inertial confinement fusion experiments along a single line of sight at Omega.

Neutron and x-ray imaging provide critical information about the geometry and hydrodynamics of inertial confinement fusion implosions. However, existing diagnostics at Omega and the National Ignition Facility (NIF) cannot produce images in both neutrons and x-rays along the same line of sight. This leads to difficulty comparing these images, which capture different parts of the plasma geometry, for the asymmetric implosions seen in present experiments. Further, even when opposing port neutron and x-ray images are available, they use different detectors and cannot provide positive information about the relative positions of the neutron and x-ray sources. A technique has been demonstrated on implosions at Omega that can capture x-ray images along the same line of sight as the neutron images. The technique is described, and data from a set of experiments are presented, along with a discussion of techniques for coregistration of the various images. It is concluded that the technique is viable and could provide valuable information if implemented on NIF in the near future.

Impact of Open Data Policies on Consent to Participate in Human Subjects Research: Discrepancies between Participant Action and Reported Concerns.

Research outlets are increasingly adopting open data policies as a requisite for publication, including studies with human subjects data. We investigated whether open data policies influence participants' rate of consent by randomly assigning participants to view consent forms with and without discussion of open data policies. No participants declined to participate, regardless of condition, nor did rates of drop-out vs. completion vary between conditions. Furthermore, no significant change in potential consent rates was reported when participants were openly asked about the influence of open data policies on their likelihood of consent. However, follow-up analyses indicated possible poor attention to consent forms, consistent with previous research. Moreover, thematic analysis of participants' considerations of open data policy indicated multiple considerations such as concerns regarding confidentiality, anonymity, data security, and study sensitivity. The impact of open data policies on participation raises complex issues at the intersection of ethics and scientific innovation. We conclude by encouraging researchers to consider participants as stakeholders in open data policy and by providing recommendations for open data policies in human subjects research.

Four-wave mixing and octave-spanning supercontinuum generation in a small core hydrogenated amorphous silicon fiber pumped in the mid-infrared.

An octave-spanning supercontinuum is generated in a hydrogenated amorphous silicon core fiber when pumped in the mid-infrared regime. The broadband wavelength conversion which extends from the edge of the telecommunications band into the mid-infrared (1.64-3.37 µm) is generated by four-wave mixing (FWM) and subsequent pulse break-up, facilitated by the high material nonlinear figure of merit and the anomalous dispersion of the relatively small 1.7 µm diameter core fiber. The FWM sidebands and corresponding supercontinuum can be tuned through the pump parameters, and show good agreement with the predicted phase-matching curves for the fiber.

Influence of high altitude on cerebral blood flow and fuel utilization during exercise and recovery.

We examined the hypotheses that: (1) during incremental exercise and recovery following 4-6 days at high altitude (HA) global cerebral blood flow (gCBF) increases to preserve cerebral oxygen delivery (CDO2) in excess of that required by an increasing cerebral metabolic rate of oxygen ( CM RO2); (2) the trans-cerebral exchange of oxygen vs. carbohydrates (OCI; carbohydrates = glucose + ½lactate) would be similar during exercise and recovery at HA and sea level (SL). Global CBF, intra-cranial arterial blood velocities, extra-cranial blood flows, and arterial-jugular venous substrate differences were measured during progressive steady-state exercise (20, 40, 60, 80, 100% maximum workload (Wmax)) and through 30 min of recovery. Measurements (n = 8) were made at SL and following partial acclimatization to 5050 m. At HA, absolute Wmax was reduced by ∼50%. During submaximal exercise workloads (20-60% Wmax), despite an elevated absolute gCBF (∼20%, P < 0.05) the relative increases in gCBF were not different at HA and SL. In contrast, gCBF was elevated at HA compared with SL during 80 and 100% Wmax and recovery. Notwithstanding a maintained CDO2 and elevated absolute CM RO2 at HA compared with SL, the relative increase in CM RO2 was similar during 20-80% Wmax but half that of the SL response (i.e. 17 vs. 27%; P < 0.05 vs. SL) at 100% Wmax. The OCI was reduced at HA compared with SL during 20, 40, and 60% Wmax but comparable at 80 and 100% Wmax. At HA, OCI returned almost immediately to baseline values during recovery, whereas at SL it remained below baseline. In conclusion, the elevations in gCBF during exercise and recovery at HA serve to maintain CDO2. Despite adequate CDO2 at HA the brain appears to increase non-oxidative metabolism during exercise and recovery.

Coinfection and the evolution of drug resistance.

Recent experimental work in the rodent malaria model has shown that when two or more strains share a host, there is competitive release of drug-resistant strains upon treatment. In other words, the propagule output of a particular strain is repressed when competing with other strains and increases upon the removal of this competition. This within-host effect is predicted to have an important impact on the evolution and growth of resistant strains. However, how this effect translates to epidemiological parameters at the between-host level, the level at which disease and resistance spread, has yet to be determined. Here we present a general, between-host epidemiological model that explicitly takes into account the effect of coinfection and competitive release. Although our model does show that when there is coinfection competitive release may contribute to the emergence of resistance, it also highlights an additional between-host effect. It is the combination of these two effects, the between-host effect and the within-host effect, that determines the overall influence of coinfection on the emergence of resistance. Therefore, even when competitive release of drug-resistant strains occurs, within an infected individual, it is not necessarily true that coinfection will result in the increased emergence of resistance. These results have important implications for the control of the emergence and spread of drug resistance.

Sensitivity of diabetic retinopathy associated vision loss to screening interval in an agent-based/discrete event simulation model.

OBJECTIVE: To examine the effect of changes to screening interval on the incidence of vision loss in a simulated cohort of Veterans with diabetic retinopathy (DR). This simulation allows us to examine potential interventions without putting patients at risk. METHODS: Simulated randomized controlled trial. We develop a hybrid agent-based/discrete event simulation which incorporates a population of simulated Veterans--using abstracted data from a retrospective cohort of real-world diabetic Veterans--with a discrete event simulation (DES) eye clinic at which it seeks treatment for DR. We compare vision loss under varying screening policies, in a simulated population of 5000 Veterans over 50 independent ten-year simulation runs for each group. RESULTS: Diabetic Retinopathy associated vision loss increased as the screening interval was extended from one to five years (p<0.0001). This increase was concentrated in the third year of the screening interval (p<0.01). There was no increase in vision loss associated with increasing the screening interval from one year to two years (p=0.98). CONCLUSIONS: Increasing the screening interval for diabetic patients who have not yet developed diabetic retinopathy from 1 to 2 years appears safe, while increasing the interval to 3 years heightens risk for vision loss.

Randomized in vivo trial evaluating plaque inhibition benefits of an advanced stannous-containing sodium fluoride dentifrice used in conjunction with power brush technology.

OBJECTIVE: To compare the plaque inhibition efficacy of a novel stannous-containing sodium fluoride test dentifrice to a standard anticavity negative control dentifrice, when both were used in conjunction with an advanced oscillating-rotating (O/R) power toothbrush. METHODS: This was a randomized, two-treatment, three-period, double-blind crossover study conducted in a population using an O/R power brush. Subjects brushed twice per day with their assigned dentifrice during the three-treatment periods, each lasting for 17 consecutive days. Each period was separated by a 4-day washout period during which subjects continued to use their O/R power toothbrush. Plaque levels were assessed and averaged amongst three assessments taken on days 15, 16 and 17 at the end of each treatment period using digital plaque imaging analysis. Assessments were carried out on the facial anterior tooth surfaces in the morning before brushing (A.M. prebrush) following whole-mouth brushing (30 s per quadrant) with the assigned dentifrice (A.M. post-brush) and in the afternoon (P.M.). RESULTS: Twenty-seven subjects were randomized and completed the study. During the 17-day usage period, the stannous-containing test NaF dentifrice demonstrated a statistically significant lower mean plaque area versus the negative control dentifrice at each assessment timepoint; overnight A.M. prebrush was 33.8% lower (P < 0.0001), A.M. post-brush was 21.8% lower (P < 0.01), and P.M. was 29.2% lower (P < 0.0001). CONCLUSION: A population of O/R power toothbrush users had significantly less plaque coverage for all three measurements when using a stannous-containing NaF dentifrice than when using a negative control (fluoride) dentifrice.