A systematic review on the use of explainability in deep learning systems for computer aided diagnosis in radiology: Limited use of explainable AI?

OBJECTIVES: This study aims to contribute to an understanding of the explainability of computer aided diagnosis studies in radiology that use end-to-end deep learning by providing a quantitative overview of methodological choices and by discussing the implications of these choices for their explainability. METHODS: A systematic review was executed using the preferred reporting items for systemic reviews and meta-analysis guidelines. Primary diagnostic test accuracy studies using end-to-end deep learning for radiology were identified from the period January 1st, 2016, to January 20th, 2021. Results were synthesized by identifying the explanation goals, measures, and explainable AI techniques. RESULTS: This study identified 490 primary diagnostic test accuracy studies using end-to-end deep learning for radiology, of which 179 (37%) used explainable AI. In 147 out of 179 (82%) of studies, explainable AI was used for the goal of model visualization and inspection. Class activation mapping is the most common technique, being used in 117 out of 179 studies (65%). Only 1 study used measures to evaluate the outcome of their explainable AI. CONCLUSIONS: A considerable portion of computer aided diagnosis studies provide a form of explainability of their deep learning models for the purpose of model visualization and inspection. The techniques commonly chosen by these studies (class activation mapping, feature activation mapping and t-distributed stochastic neighbor embedding) have potential limitations. Because researchers generally do not measure the quality of their explanations, we are agnostic about how effective these explanations are at addressing the black box issues of deep learning in radiology.

Risk communication about work-related stress disorders in healthcare workers: a scoping review.

PURPOSES: Healthcare workers are at risk of stress-related disorders. Risk communication can be an effective preventive health measure for some health risks, but is not yet common in the prevention of stress-related disorders in an occupational healthcare setting. The overall aim is to examine whether risk communication was part of interventions aimed at the prevention of stress-related disorders in healthcare workers. METHOD: We performed a scoping review using the framework of Arksey and O'Malley. We searched in Medline, Web of Science and PsychInfo for studies reporting on preventive interventions of stress-related disorders in healthcare workers between 2005 and December 2020. Studies were included when the intervention reported on at least one element of risk communication and one goal. We predefined four elements of risk communication: risk perception, communication of early stress symptoms, risk factors and prevention; and three goals: inform, stimulate informed decision-making and motivate action. RESULTS: We included 23 studies that described 17 interventions. None of the included interventions were primarily developed as risk communication interventions, but all addressed the goals. Two interventions used all four elements of risk communication. The prominent mode of delivery was face to face, mostly delivered by researchers. Early stress symptoms and risk factors were measured by surveys. CONCLUSIONS: Risk communication on risk factors and early signs of stress-related disorders is not that well studied and evaluated in an occupational healthcare setting. Overall, the content of the communication was not based on the risk perception of the healthcare workers, which limited the likelihood of them taking action.

Online Indicated Preventive Mental Health Interventions for Youth: A Scoping Review.

Objective: Between the ages of 12 and 25 the onset of mental disorders typically occurs, and the burden of mental health problems is greatest for this group. Indicated preventive interventions to target individuals with subclinical symptoms to prevent the transition to clinical levels of disorders have gained considerable traction. However, the threshold to seek help appears to be high even when help is needed. Online interventions could offer a solution, especially during the COVID-19 pandemic. This scoping review will present an overview of the recent research of indicated online preventive interventions for youth (12-25 years) experiencing the early stages of mental health complaints with the aim of identifying the nature and extent of the research evidence. Methods: The 5-stage framework by Arksey and O'Malley was used. Academic literature published from 2013 onwards in printed or electronic format was included from Scopus, PsychINFO, and Ovid MEDLINE(R) ALL. Results: The search yielded 11,122 results, with the final selection resulting in inclusion of 30 articles for this review. In total, the articles included 4,950 participants. 26.7% of the selected articles focused on youth between 12 and 25 years. Of the articles 60% did not screen for, nor exclude participants with clinical levels of symptoms. Most studies used a common evidence-based therapy for the disorder-category targeted. More than half of the online interventions included some form of human support. Adherence levels ranged between 27.9 and 98%. The results indicate general effectiveness, usability and acceptability of online indicated preventive interventions. The most commonly used approach was CBT (n = 12 studies). Studies varied in their size, rigor of study, effectiveness and outcome measures. Online interventions with a combination of clinical and peer moderation (n = 3 studies) appear to result in the most stable and highest effect sizes. Conclusion: Online indicated preventive mental health interventions for youth with emerging mental health issues show promise in reducing various mental health complaints, and increasing positive mental health indicators such as well-being and resilience. Additionally, high levels of usability and acceptability were found. However, the included studies show important methodological shortcomings. Also, the research has mainly focused on specific diagnostic categories, meaning there is a lack of transdiagnostic approaches. Finally, clear definitions of- as well as instruments to measure- emerging or subclinical mental health symptoms in youth remain are missing.