Objective To demonstrate enabling multi-institutional training without centralizing or sharing the underlying physical data via federated learning (FL). Materials and Methods Deep learning models were trained at each participating institution using local clinical data, and an additional model was trained using FL across all of the institutions. Results We found that the FL model exhibited superior performance and generalizability to the models trained at single institutions, with an overall performance level that was significantly better than that of any of the institutional models alone when evaluated on held-out test sets from each institution and an outside challenge dataset. Discussion The power of FL was successfully demonstrated across 3 academic institutions while avoiding the privacy risk associated with the transfer and pooling of patient data. Conclusion Federated learning is an effective methodology that merits further study to enable accelerated development of models across institutions, enabling greater generalizability in clinical use.

No abstract

No abstract

No abstract

Asia’s coastlines are choking in waste. The region is now home to many of the world’s most polluted beaches. The populous Indian Cities are growing economically but in an unsustainable manner. With Mumbai counted among topmost polluted beaches in the world, it is the need of the hour to take necessary steps for effective waste management by systematic data analysis for deriving useful information from waste generation patterns. The major objective of the study is pattern recognition and beach waste quantum prediction based on 5 years data, with a frequency of daily waste collection. The size of the training data set is 1,661 days and the validation data set is 335 days. The influence of population trend, waste generation during festivals, special days, weekends, and seasonal variations form the basis for the analysis. Using machine learning algorithms, the study identifies and investigates data patterns for the case study of Dadar-Mahim beach. Data frequency and weights are correlated with occurrence of events, festivals, weekends, and seasons. Exploratory Data Analysis (EDA) is employed for data preprocessing and wrangling, followed by a Random Forest algorithm-based model for the prediction of waste generated at Dadar-Mahim beach. The major challenges in data prediction are limited data availability and variation in the dates of festivals and holidays as well as lack of waste segregation information. Despite the above-mentioned challenges, the observations indicate the model’s average accuracy for making predictions of around 60%. The Graphic User Interface (GUI) developed based on the model provides a user-friendly application for predicting the total daily generation of beach waste with reasonable precision. On the basis of the model’s outcome and applicability, a schematic approach for efficient beach waste management is proposed. The recommendations would serve as guidelines for Urban Local Bodies (ULBs) to automate the collection, transport, and disposal of beach waste.