Inconsistency between Human Observation and Deep Learning Models: Assessing Validity of Postmortem Computed Tomography Diagnosis of Drowning.

Drowning diagnosis is a complicated process in the autopsy, even with the assistance of autopsy imaging and the on-site information from where the body was found. Previous studies have developed well-performed deep learning (DL) models for drowning diagnosis. However, the validity of the DL models was not assessed, raising doubts about whether the learned features accurately represented the medical findings observed by human experts. In this paper, we assessed the medical validity of DL models that had achieved high classification performance for drowning diagnosis. This retrospective study included autopsy cases aged 8-91 years who underwent postmortem computed tomography between 2012 and 2021 (153 drowning and 160 non-drowning cases). We first trained three deep learning models from a previous work and generated saliency maps that highlight important features in the input. To assess the validity of models, pixel-level annotations were created by four radiological technologists and further quantitatively compared with the saliency maps. All the three models demonstrated high classification performance with areas under the receiver operating characteristic curves of 0.94, 0.97, and 0.98, respectively. On the other hand, the assessment results revealed unexpected inconsistency between annotations and models' saliency maps. In fact, each model had, respectively, around 30%, 40%, and 80% of irrelevant areas in the saliency maps, suggesting the predictions of the DL models might be unreliable. The result alerts us in the careful assessment of DL tools, even those with high classification performance.

System dynamics analysis on the effectiveness of vaccination and social mobilization policies for COVID-19 in the United States.

The COVID-19 pandemic has presented significant public health and economic challenges worldwide. Various health and non-pharmaceutical policies have been adopted by different countries to control the spread of the virus. To shed light on the impact of vaccination and social mobilization policies during this wide-ranging crisis, this paper applies a system dynamics analysis on the effectiveness of these two types of policies on pandemic containment and the economy in the United States. Based on the simulation of different policy scenarios, the findings are expected to help decisions and mitigation efforts throughout this pandemic and beyond.

Correction: A dynamic generalized fuzzy multi-criteria group decision making approach for green supplier segmentation.

[This corrects the article DOI: 10.1371/journal.pone.0245187.].

A dynamic generalized fuzzy multi-criteria croup decision making approach for green supplier segmentation.

Supplier selection and segmentation are crucial tasks of companies in order to reduce costs and increase the competitiveness of their goods. To handle uncertainty and dynamicity in the supplier segmentation problem, this research thus proposes a new dynamic generalized fuzzy multi-criteria group decision making (MCGDM) approach from the aspects of capability and willingness and with respect to environmental issues. The proposed approach defines the aggregated ratings of alternatives, the aggregated weights of criteria, and the weighted ratings by using generalized fuzzy numbers with the effect of time weight. Next, we determine the ranking order of alternatives via a popular centroid-index ranking approach. Finally, two case studies demonstrate the efficiency of the proposed dynamic approach. The applications show that the proposed appoach is effective in solving the MCGDM in vague environment.

Dynamic Space Allocation Based on Internal Demand for Optimizing Release of Shared Parking.

The size of cities has been continuously increasing because of urbanization. The number of public and private transportation vehicles is rapidly increasing, thus resulting in traffic congestion, traffic accidents, and environmental pollution. Although major cities have undergone considerable development in terms of transportation infrastructure, problems caused by a high number of moving vehicles cannot be completely resolved through the expansion of streets and facilities. This paper proposes a solution for the parking problem in cities that entails a shared parking system. The primary concept of the proposed shared parking system is to release parking lots that are open to specific groups for public usage without overriding personal usage. Open-to-specific-groups parking lots consist of parking spaces provided for particular people, such as parking buildings at universities for teachers, staff, and students. The proposed shared parking system comprises four primary steps: collecting and preprocessing data by using an Internet of Things system, predicting internal demand by using a recurrent neural network algorithm, releasing several unoccupied parking lots based on prediction results, and continuously updating the real-time data to improve future internal usage prediction. Data collection and data forecasting are performed to ensure that the system does not override personal usage. This study applied several forecasting algorithms, including seasonal ARIMA, support vector regression, multilayer perceptron, convolutional neural network, long short-term memory recurrent neural network with a many-to-one structure, and long short-term memory recurrent neural network with a many-to-many structure. The proposed system was evaluated using artificial and real datasets. Results show that the recurrent neural network with the many-to-many structure generates the most accurate prediction. Furthermore, the proposed shared parking system was evaluated for some scenarios in which different numbers of parking spaces were released. Simulation results show that the proposed shared parking system can provide parking spaces for public usage without overriding personal usage. Moreover, this system can generate new income for parking management and/or parking lot owners.