Artificial Swarm Intelligence employed to Amplify Diagnostic Accuracy in Radiology

Rosenberg, Louis; Lungren, Matthew P.; Halabi, Safwan; Willcox, Gregg; Baltaxe, David; Lyons, Mimi

doi:10.1109/iemcon.2018.8614883

Cited by 48 publications

(23 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Similarly to the use of the National Lung Screening Trial, bootstrapping was used to assess individual radiologists' results compared to a collective "swarm" intelligence model of multiple radiologists for ground truth production. By bootstrapping 10,000 times, it was found that the collective "swarm" model outperformed the individual radiologists by 9.1% at a p < 0.01 (29). Bootstrapping has also been used along with feature engineering to evaluate radiomics features predicting cancer in ultrasound breast tumor classification (30) and other malignancies (31).…”

Section: Booting With the Bootstrapmentioning

confidence: 99%

Machine Learning Principles for Radiology Investigators

Borstelmann

2020

Academic Radiology

View full text Add to dashboard Cite

Section: Booting With the Bootstrapmentioning

confidence: 99%

Machine Learning Principles for Radiology Investigators

Borstelmann

2020

Academic Radiology

View full text Add to dashboard Cite

“…The group's decision is dynamic, representing real‐time negotiation among group members collectively exploring the decision space and converging upon the most agreeable answer. Artificial swarm intelligence outperforms medical experts and machine‐learning algorithms and makes relatively accurate predictions for financial markets and the outcomes of sporting events (Rosenberg & Pescetelli, , ).…”

Section: Distributed Processingmentioning

confidence: 99%

Untapped potential of collective intelligence in conservation and environmental decision making

Vercammen

Burgman

2019

Conservation Biology

View full text Add to dashboard Cite

Environmental decisions are often deferred to groups of experts, committees, or panels to develop climate policy, plan protected areas, or negotiate trade‐offs for biodiversity conservation. There is, however, surprisingly little empirical research on the performance of group decision making related to the environment. We examined examples from a range of different disciplines, demonstrating the emergence of collective intelligence (CI) in the elicitation of quantitative estimates, crowdsourcing applications, and small‐group problem solving. We explored the extent to which similar tools are used in environmental decision making. This revealed important gaps (e.g., a lack of integration of fundamental research in decision‐making practice, absence of systematic evaluation frameworks) that obstruct mainstreaming of CI. By making judicious use of interdisciplinary learning opportunities, CI can be harnessed effectively to improve decision making in conservation and environmental management. To elicit reliable quantitative estimates an understanding of cognitive psychology and to optimize crowdsourcing artificial intelligence tools may need to be incorporated. The business literature offers insights into the importance of soft skills and diversity in team effectiveness. Environmental problems set a challenging and rich testing ground for collective‐intelligence tools and frameworks. We argue this creates an opportunity for significant advancement in decision‐making research and practice.

show abstract

“…Examples of other industries actively using trace-based communication include the military to improve SEAL (SEa, Air, and Land) team training, 26 the construction industry to increase the efficiency in interior wall building, 27 and the tech industry to optimise forecasts and decision-making. 28 These examples share the same philosophy: without directly talking to each other, individuals become aware of traces left in their work environment and use them to carry out their part of the work. In the case of healthcare, the introduction of electronic health records has opened one avenue for the use of traces such as digital ‘flags’ for asynchronous communication in acute care settings.…”

Section: Introductionmentioning

confidence: 99%

Qualitative investigation of trace-based communication: how are traces conceptualised in healthcare teamwork?

Cristancho

Field

2020

BMJ Open

View full text Add to dashboard Cite

ObjectivesThis interview-based qualitative study aims to explore how healthcare providers conceptualise trace-based communication and considers its implications for how teams work. In the biological literature, trace-based communication refers to the non-verbal communication that is achieved by leaving ‘traces’ in the environment and other members sensing them and using them to drive their own behaviour. Trace-based communication is a key component of swam intelligence and has been described as a critical process that enables superorganisms to coordinate work and collectively adapt. This paper brings awareness to its existence in the context of healthcare teamwork.DesignInterview-based study using Constructivist Grounded Theory methodology.SettingThis study was conducted in multiple team contexts at one of Canada’s largest acute-care teaching hospitals.Participants25 clinicians from across professions and disciplines. Specialties included surgery, anesthesiology, psychiatry, internal medicine, geriatrics, neonatology, paramedics, nursing, intensive care, neurology and emergency medicine.InterventionNot relevant due to the qualitative nature of the study.Primary and secondary outcomeNot relevant due to the qualitative nature of the study.ResultsThe dataset was analysed using the sensitising concept of ‘traces’ from Swarm Intelligence. This study brought to light novel and unique elements of trace-based communication in the context of healthcare teamwork including focused intentionality, successful versus failed traces and the contextually bounded nature of the responses to traces. While participants initially felt ambivalent about the idea of using traces in their daily teamwork, they provided a variety of examples. Through these examples, participants revealed the multifaceted nature of the purposes of trace-based communication, including promoting efficiency, preventing mistakes and saving face.ConclusionsThis study demonstrated that clinicians pervasively use trace-based communication despite differences in opinion as to its implications for teamwork and safety. Other disciplines have taken up traces to promote collective adaptation. This should serve as inspiration to at least start exploring this phenomenon in healthcare.

show abstract

Artificial Swarm Intelligence employed to Amplify Diagnostic Accuracy in Radiology

Cited by 48 publications

References 12 publications

Machine Learning Principles for Radiology Investigators

Machine Learning Principles for Radiology Investigators

Untapped potential of collective intelligence in conservation and environmental decision making

Qualitative investigation of trace-based communication: how are traces conceptualised in healthcare teamwork?

Contact Info

Product

Resources

About