Chronic Disease Management with Personalized Lab Test Response Prediction

Abstract

Monitoring vegetation productivity at extremely fine resolutions is valuable for real-world agricultural applications, such as detecting crop stress and providing early warning of food insecurity. Solar-Induced Chlorophyll Fluorescence (SIF) provides a promising way to directly measure plant productivity from space. However, satellite SIF observations are only available at a coarse spatial resolution, making it impossible to monitor how individual crop types or farms are doing. This poses a challenging coarsely-supervised regression (or downscaling) task; at training time, we only have SIF labels at a coarse resolution (3km), but we want to predict SIF at much finer spatial resolutions (e.g. 30m, a 100x increase). We also have additional fine-resolution input features, but the relationship between these features and SIF is unknown. To address this, we propose Coarsely-Supervised Smooth U-Net (CS-SUNet), a novel method for this coarse supervision setting. CS-SUNet combines the expressive power of deep convolutional networks with novel regularization methods based on prior knowledge (such as a smoothness loss) that are crucial for preventing overfitting. Experiments show that CS-SUNet resolves fine-grained variations in SIF more accurately than existing methods.

Infectious Hospital Area, the First Affiliated Hospital of University of Science and Technology of China, Hefei, China

“…We omit the evaluation of the 4SDrug, DrugRec, and REFINE methods from our study due to their reliance on additional symptom information [34,35] or unavailability of the source codes [3]. Implementation Details.…”

“…patients' longitudinal medical history, such as the approach taken by Choi et al [6], who employ a two-level temporal attention mechanism. Similarly, Shang et al [29] pre-train their model on single-visit data 3 and fine-tune it using multi-visit data 4 . Additionally, Shang et al [30] incorporate a graph-augmented memory module and DDI graph to reduce adverse drug-drug interactions, while Yang et al [42] consider drug molecule structures for medication security.…”

“…Sun et al [34] present a causal model to address recommendation bias, and Yang et al [44] leverage a molecular substructure-aware attentive method. More recently, Bhoi et al [3] propose a fine-grained medication recommendation approach. However, it is important to note that many of these methods heavily rely on historical records, making them less suitable for single-visit patients [3,6,29,34], or may exhibit reduced accuracy in such cases [30,[40][41][42]44].…”

“…More recently, Bhoi et al [3] propose a fine-grained medication recommendation approach. However, it is important to note that many of these methods heavily rely on historical records, making them less suitable for single-visit patients [3,6,29,34], or may exhibit reduced accuracy in such cases [30,[40][41][42]44]. Instance-based methods focus on the current medical condition of patients.…”

Predicting Physiological Response in Heart Failure Management: A Graph Representation Learning Approach using Electronic Health Records

“…The utility of graph representation learning in treatment outcome prediction is currently in its infancy. A recent work 27 performed lab test response prediction by first using Transformers to encode the longitudinal diagnosis and medication information in the patient's EHR. It then uses Graph Attention Networks (GAT) to encode the similarity among the patients and the lab interaction-based external knowledge.…”