ABSTRACT
Resilience is becoming an increasingly important trait for global and lean supply chains, especially due to recent external shocks or disruptions to supply chains, such as the Covid-19 pandemic, the Suez Canal blockage or the container shortage. Furthermore, megatrends and ongoing developments as for instance a rapidly growing global population, advancing urbanization, resource scarcity and digitalization intensify the question of resilient and sustainable production and logistics systems. Cities are a hotspot for value creation and are connected to specific opportunities and challenges at the same time for manufacturing industry in urban factories. To cope with these developments, a structured process has been designed to analyze the influence and linkages between the characteristics of urban factories and resilience indicators. With these connections, opportunities and challenges for urban factories and their supply chains regarding resilience indicators can be identified. The results can be utilized to create more resilient production sites and supply chains in an urban environment and in general. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.
ABSTRACT
Time-to-event analysis is an important statistical tool for allocating clinical resources such as ICU beds. However, classical techniques like the Cox model cannot directly incorporate images due to their high dimensionality. We propose a deep learning approach that naturally incorporates multiple, time-dependent imaging studies as well as non-imaging data into time-to-event analysis. Our techniques are bench-marked on a clinical dataset of 1,894 COVID-19 patients, and show that image sequences significantly improve predictions. For example, classical time-to-event methods produce a concordance error of around 30-40% for predicting hospital admission, while our error is 25% without images and 20% with multiple X-rays included. Ablation studies suggest that our models are not learning spurious features such as scanner artifacts and that models which use multiple images tend to perform better than those which only use one. While our focus and evaluation is on COVID-19, the methods we develop are broadly applicable.