Mortality prediction with adaptive feature importance recalibration for peritoneal dialysis patients.

The study aims to develop AICare, an interpretable mortality prediction model, using electronic medical records (EMR) from follow-up visits for end-stage renal disease (ESRD) patients. AICare includes a multichannel feature extraction module and an adaptive feature importance recalibration module. It integrates dynamic records and static features to perform personalized health context representation learning. The dataset encompasses 13,091 visits and demographic data of 656 peritoneal dialysis (PD) patients spanning 12 years. An additional public dataset of 4,789 visits from 1,363 hemodialysis (HD) patients is also considered. AICare outperforms traditional deep learning models in mortality prediction while retaining interpretability. It uncovers mortality-feature relationships and variations in feature importance and provides reference values. An AI-doctor interaction system is developed for visualizing patients' health trajectories and risk indicators.

Monitoring autumn agriculture activities using Synthetic Aperture Radar (SAR) and coherence change detection.

Across Canada, farmers are encouraged to adopt beneficial management practices (BMPs) to protect soil heath, reduce green house gas emissions and mitigate off-site impacts from agriculture. Measuring the uptake of BMPs, including the implementation of conservation tillage, helps gauge the success of policies and programs to promote adoption. Satellites are one way to monitor BMP adoption and Synthetic Aperture Radars (SARs) are of particular interest given their all-weather data collection capability. This research investigated coherent change detection (CCD) to determine when farmers harvest and till their fields. A time series of both Sentinel-1 and RADARSAT Constellation Mission (RCM) images was acquired over a site in the Canadian Lake Erie basin, during the autumn of 2021, when farmers were harvesting and tilling fields of corn, soybeans and wheat. 16 CCD pairs were created and coherence values were interpreted based on observations collected for 101 fields. An m-chi decomposition was applied to the RCM data, and the Volume/Surface (V/S) ratio was calculated as an additional source of information to interpret results. Change events due to harvest, tillage, autumn seeding and chemical termination resulted in coherence values below 0.20. The mean and standard deviation for fields with observed change was 0.18 ± 0.03. Coherence values were 0.42 ± 0.15 for fields where no change was noted. Tests confirmed that the coherence associated with changed and unchanged fields was significantly different. Coherence values could also differentiate between some types of management events, including tillage and harvest. CCD could also separate harvest as a function of crop type (corn or soybeans). V/S ratios declined after tillage events but increased after both harvesting and chemical termination. Narrowing the date of harvest and tillage is as important as detecting change. To meet this requirement, Sentinel-1 and RCM CCD products with values below 0.20 (indicating change had occurred), were graphically overlaid. With this approach, the timing of corn harvest was identified as occurring within a 5-day window. The tilling of corn, soybeans and wheat was narrowed to a 4-day window. The results of this research confirmed that CCD can be used to capture change due to autumn agricultural activities, and this technique can also separate change due to harvest and tillage. Finally, this study demonstrated that when data from different SAR missions are combined in a virtual constellation, timing of harvest and tillage can be more precisely defined.