An impact model to understand and improve work-life balance in early-career researchers in radiation oncology.

Purpose: The COVID-19 pandemic had a substantial effect on mental health and work productivity of early-career researchers working in Radiation Oncology (RO). However, the underlying mechanisms of these effects are unclear. The aim of the current qualitative study was therefore to achieve a better understanding of how these effects arose and could be managed in the future. Methods: This study was conducted jointly by RO and qualitative health researchers. Data was collected in four online Focus Groups with 6-11 RO researchers (total N = 31) working in Europe. The transcripts were analysed through a qualitative cross-impact analysis. Results: Causal relations were identified between seventeen variables that depict the impact of disrupted working conditions. Mental health and work productivity were indeed the most important affected variables, but relations between variables towards these impacts were complex. Relations could either be positive or negative and direct or indirect, leading to a cascade of interrelated events which are highly personal and could change over time. We developed the model 'impact of disrupted working conditions' depicting the identified variables and their relations, to allow more individual assessment and personalised solutions. Conclusion: The impacts of disrupted working conditions on RO researchers varied due to the complexity of interrelated variables. Consequently, collective actions are not sufficient, and a more personal approach is needed. Our impact model is recommended to help guide conversations and reflections with the aim of improving work/life balance. The participants showed high levels of personal responsibility towards their own mental health and work productivity. Although being an individual issue, a collective responsibility in developing such approaches is key due to the dependency on organizational variables.

Automatic coronavirus disease 2019 diagnosis based on chest radiography and deep learning - Success story or dataset bias?

PURPOSE: Over the last 2 years, the artificial intelligence (AI) community has presented several automatic screening tools for coronavirus disease 2019 (COVID-19) based on chest radiography (CXR), with reported accuracies often well over 90%. However, it has been noted that many of these studies have likely suffered from dataset bias, leading to overly optimistic results. The purpose of this study was to thoroughly investigate to what extent biases have influenced the performance of a range of previously proposed and promising convolutional neural networks (CNNs), and to determine what performance can be expected with current CNNs on a realistic and unbiased dataset. METHODS: Five CNNs for COVID-19 positive/negative classification were implemented for evaluation, namely VGG19, ResNet50, InceptionV3, DenseNet201, and COVID-Net. To perform both internal and cross-dataset evaluations, four datasets were created. The first dataset Valencian Region Medical Image Bank (BIMCV) followed strict reverse transcriptase-polymerase chain reaction (RT-PCR) test criteria and was created from a single reliable open access databank, while the second dataset (COVIDxB8) was created through a combination of six online CXR repositories. The third and fourth datasets were created by combining the opposing classes from the BIMCV and COVIDxB8 datasets. To decrease inter-dataset variability, a pre-processing workflow of resizing, normalization, and histogram equalization were applied to all datasets. Classification performance was evaluated on unseen test sets using precision and recall. A qualitative sanity check was performed by evaluating saliency maps displaying the top 5%, 10%, and 20% most salient segments in the input CXRs, to evaluate whether the CNNs were using relevant information for decision making. In an additional experiment and to further investigate the origin of potential dataset bias, all pixel values outside the lungs were set to zero through automatic lung segmentation before training and testing. RESULTS: When trained and evaluated on the single online source dataset (BIMCV), the performance of all CNNs is relatively low (precision: 0.65-0.72, recall: 0.59-0.71), but remains relatively consistent during external evaluation (precision: 0.58-0.82, recall: 0.57-0.72). On the contrary, when trained and internally evaluated on the combinatory datasets, all CNNs performed well across all metrics (precision: 0.94-1.00, recall: 0.77-1.00). However, when subsequently evaluated cross-dataset, results dropped substantially (precision: 0.10-0.61, recall: 0.04-0.80). For all datasets, saliency maps revealed the CNNs rarely focus on areas inside the lungs for their decision-making. However, even when setting all pixel values outside the lungs to zero, classification performance does not change and dataset bias remains. CONCLUSIONS: Results in this study confirm that when trained on a combinatory dataset, CNNs tend to learn the origin of the CXRs rather than the presence or absence of disease, a behavior known as short-cut learning. The bias is shown to originate from differences in overall pixel values rather than embedded text or symbols, despite consistent image pre-processing. When trained on a reliable, and realistic single-source dataset in which non-lung pixels have been masked, CNNs currently show limited sensitivity (<70%) for COVID-19 infection in CXR, questioning their use as a reliable automatic screening tool.

Conducting research in Radiation Oncology remotely during the COVID-19 pandemic: Coping with isolation.

INTRODUCTION: With the COVID-19 pandemic, individuals have been forced to follow strict social isolation guidelines. While crucial to control the pandemic, isolation might have a significant impact on productivity and mental health. Especially for researchers working in healthcare, the current situation is complex. We therefore carried out a survey amongst researchers in the field of radiation oncology to gain insights on the impact of social isolation and working from home and to guide future work. MATERIALS AND METHODS: An online survey was conducted between March 27th and April 5th, 2020. The first part contained 14 questions intended to capture an overview of the specific aspects related to research while in isolation. The second (optional) part of the questionnaire was the validated Hospital Anxiety and Depression Scale (HADS), a self-reported measure used to assess levels of anxiety and depressive symptoms. RESULTS: From 543 survey participants, 48.8% reported to work full-time from home. The impact on perceived productivity, with 71.2% of participants feeling less productive, caused 58% of participants to feel some level of guilt.Compared to normative data, relatively high levels of anxiety and depressive symptoms were recorded for the 335 participants who filled out the HADS questionnaire. Group comparisons found the presence of a supportive institutional program as the sole factor of statistical significance in both anxiety and depressive symptom levels. People having to work full-time on location showed higher depressive symptom levels than those working from home. Anxiety scores were negatively correlated with the number of research years. CONCLUSION: Results of the survey showed there is a non-negligible impact on both productivity and mental health. As the radiation oncology research community was forced to work from home during the COVID-19 pandemic, lessons can be learned to face future adverse situations but also to improve work-life balance in general.