Unified Screening for Multiple Diseases
Authors: Yiğit Narter, Alihan Hüyük, Mihaela Van Der Schaar, Cem Tekin
ICML 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experimental results demonstrate that the mathematical characterizations of the optimal policies align closely with the numerical solution of the convex program (see Figure 1). Importantly, we compare the performance of our unified screening model against that of independent screening programs. In silico experiments for the unified screening of two diseases demonstrate that survival times improve compared to the optimal individual screening programs, validating the benefit of unified screening (see Figure 2). 5. In-Silico Experiments |
| Researcher Affiliation | Academia | 1Department of Electrical and Electronics Engineering, Bilkent University, Ankara, Turkey 2Department of Computer Science, Harvard University, Cambridge, MA, US 3University of Cambridge, Cambridge, UK. |
| Pseudocode | No | The paper describes mathematical formulations and propositions but does not include any clearly labeled pseudocode or algorithm blocks. The screening and diagnostic policies are described in paragraph text and equations, not in a structured code-like format. |
| Open Source Code | Yes | Code is available at https://github.com/ynarter/UniScreen. |
| Open Datasets | No | The paper describes generating its own synthetic data for in-silico experiments: "Adverse event times Tn, n [N] (in years) are generated from N(µn, σ2)." and "We set p X(x) f Beta(x1; α, β)f Beta(x2; α, β) where f Beta(x; α, β) is the pdf of the Beta distribution." It does not provide access information for a publicly available or open dataset. |
| Dataset Splits | No | The paper describes a simulation-based approach where it runs "M = 200 independent simulations for each of the N = 10,000 patient feature vectors x = (x1, x2) to estimate the expected outcomes under various screening actions." It does not describe explicit training/test/validation splits for a dataset, as it's a simulation study. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used to run the experiments, such as GPU or CPU models, or memory amounts. |
| Software Dependencies | No | The paper mentions using "CVXPY" for solving the linear program but does not specify a version number for CVXPY or any other software dependencies. |
| Experiment Setup | Yes | We set the default budget B = 10, α = β = 5 and the individual screening costs as c1 = c2 = 1. At each Monte Carlo iteration, we simultaneously run M = 200 independent simulations for each of the N = 10,000 patient feature vectors x = (x1, x2)... We fix the diagnosis threshold to γn = 0.95 throughout. |