Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

From Signals to Causes: A Causal Signal Processing Framework for Robust and Interpretable Clinical Risk Prediction

Surajit Das, Maxine Tan

TL;DR

This article presents schematic causal structures and a comparative analysis of correlation-based, causal, and neuro-symbolic approaches to guide the design of robust and interpretable medical decision-support systems.

Abstract

Learning-based signal processing systems increasingly support high-stakes medical decisions using heterogeneous biomedical signals, including medical images, physiological time series, and clinical records. Despite strong predictive performance, many models rely on statistical correlations that are unstable across acquisition settings, patient populations, and institutional practices, limiting robustness, interpretability, and clinical trust. We advocate a causal signal processing perspective in which biomedical signals are treated as effects of latent generative mechanisms rather than as isolated predictive inputs. Using clinical risk prediction as a motivating example, we show how disease-related factors generate observable biomarkers, while acquisition processes act as confounders influencing signal appearance. In clinical disease risk prediction from chest CT scans and patient risk factors, correlational models may fail under scanner changes, whereas causal abstractions remain invariant. Building on this view, we propose a unifying conceptual framework integrating causal modeling with learning-based signal processing and neuro-symbolic reasoning. Statistical models extract multimodal representations that are mapped to interpretable causal abstractions and combined with symbolic knowledge encoding clinical risk factors and guidelines. This structure enables clinically grounded explanations, counterfactual reasoning about hypothetical interventions, and improved robustness to distribution shifts arising from changes in acquisition conditions or screening policies. Rather than introducing a specific algorithm, this article presents schematic causal structures and a comparative analysis of correlation-based, causal, and neuro-symbolic approaches to guide the design of robust and interpretable medical decision-support systems.

From Signals to Causes: A Causal Signal Processing Framework for Robust and Interpretable Clinical Risk Prediction

TL;DR

This article presents schematic causal structures and a comparative analysis of correlation-based, causal, and neuro-symbolic approaches to guide the design of robust and interpretable medical decision-support systems.

Abstract

Learning-based signal processing systems increasingly support high-stakes medical decisions using heterogeneous biomedical signals, including medical images, physiological time series, and clinical records. Despite strong predictive performance, many models rely on statistical correlations that are unstable across acquisition settings, patient populations, and institutional practices, limiting robustness, interpretability, and clinical trust. We advocate a causal signal processing perspective in which biomedical signals are treated as effects of latent generative mechanisms rather than as isolated predictive inputs. Using clinical risk prediction as a motivating example, we show how disease-related factors generate observable biomarkers, while acquisition processes act as confounders influencing signal appearance. In clinical disease risk prediction from chest CT scans and patient risk factors, correlational models may fail under scanner changes, whereas causal abstractions remain invariant. Building on this view, we propose a unifying conceptual framework integrating causal modeling with learning-based signal processing and neuro-symbolic reasoning. Statistical models extract multimodal representations that are mapped to interpretable causal abstractions and combined with symbolic knowledge encoding clinical risk factors and guidelines. This structure enables clinically grounded explanations, counterfactual reasoning about hypothetical interventions, and improved robustness to distribution shifts arising from changes in acquisition conditions or screening policies. Rather than introducing a specific algorithm, this article presents schematic causal structures and a comparative analysis of correlation-based, causal, and neuro-symbolic approaches to guide the design of robust and interpretable medical decision-support systems.
Paper Structure (64 sections, 3 theorems, 25 equations, 3 figures, 1 table)

This paper contains 64 sections, 3 theorems, 25 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Background: Clinical Risk Prediction and Signal Processing
  3. Conventional Learning-Based Formulation
  4. Sources of Variability in Medical Signals
  5. Limitations of Correlation-Based Risk Models
  6. Motivation for a Causal Signal Processing Perspective
  7. A Causal View of Multimodal Medical Signals
  8. Causal Roles in Medical Data
  9. Minimal Structural Causal Model.
  10. Causal Structure of Multimodal Risk Prediction
  11. Signal Processing View.
  12. Distribution Shift as Intervention
  13. Beyond Correlational Interpretability
  14. Counterfactual Interpretation.
  15. Implications for Signal Processing Methodology
  16. ...and 49 more sections

Key Result

Theorem 1

Under assumptions eq:factorization--eq:predictive_invariance, if $Z = \phi(X)$ satisfies predictive invariance across environments, then there exists a measurable function $\psi$ such that up to an invertible transformation.

Figures (3)

  • Figure 1: Conceptual comparison between correlational and causal signal processing paradigms. Conventional learning-based systems exploit statistical associations in observed signals, including spurious correlations induced by acquisition and institutional factors. Causal signal processing models signals as outcomes of latent generative mechanisms, seeking representations that remain invariant under interventions on non-causal variables.
  • Figure 2: Causal DAG and neuro-symbolic pipeline for clinical risk prediction. (a) Patient risk factors influence latent disease state, which generates observable biomarkers. (b) Acquisition factors affect signal appearance without altering disease. (c) Learned representations map signals to causal abstractions that support symbolic reasoning and intervention-aware decisions.
  • Figure 3: Proposed causal signal processing framework for clinical risk prediction. Observed multimodal signals are mapped to latent causal abstractions that distinguish disease mechanisms from acquisition-related variability. This separation enables invariant representation learning, counterfactual reasoning, and intervention-aware Scholkopf2021 decision support across deployment environments.

Theorems & Definitions (4)

  • Theorem 1
  • proof
  • Corollary 1: Robustness Without Causal Graphs
  • Corollary 2: Implicit Counterfactual Interpretation