Investigating ECG Diagnosis with Ambiguous Labels using Partial Label Learning

2025-12-11

AI Summary

ECG diagnosis suffers from pervasive label ambiguity due to disease comorbidity and inter-observer disagreement, yet most existing models assume clean, deterministic labels—limiting clinical applicability. This work systematically investigates the suitability and robustness of Partial Label Learning (PLL) for multi-label ECG classification—the first such study in this domain. We adapt nine state-of-the-art PLL algorithms to ECG tasks and propose a clinically grounded ambiguity modeling framework integrating expert consensus, therapeutic relevance, and diagnostic hierarchy. Extensive evaluation on PTB-XL and Chapman datasets demonstrates strong generalization across both structured and unstructured ambiguity patterns. Empirical analysis uncovers critical limitations of current PLL methods in temporal representation learning and clinical semantic integration. Our work establishes foundational theoretical insights and an empirical benchmark for developing ambiguity-aware, clinically trustworthy ECG diagnostic systems.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Reasoning under Uncertainty: Relational Probabilistic Models
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Accountability, Transparency, and Ethics for personalization

Abstract

Label ambiguity is an inherent problem in real-world electrocardiogram (ECG) diagnosis, arising from overlapping conditions and diagnostic disagreement. However, current ECG models are trained under the assumption of clean and non-ambiguous annotations, which limits both the development and the meaningful evaluation of models under real-world conditions. Although Partial Label Learning (PLL) frameworks are designed to learn from ambiguous labels, their effectiveness in medical time-series domains, ECG in particular, remains largely unexplored. In this work, we present the first systematic study of PLL methods for ECG diagnosis. We adapt nine PLL algorithms to multi-label ECG diagnosis and evaluate them using a diverse set of clinically motivated ambiguity generation strategies, capturing both unstructured (e.g., random) and structured ambiguities (e.g., cardiologist-derived similarities, treatment relationships, and diagnostic taxonomies). Our experiments on the PTB-XL and Chapman datasets demonstrate that PLL methods vary substantially in their robustness to different types and degrees of ambiguity. Through extensive analysis, we identify key limitations of current PLL approaches in clinical settings and outline future directions for developing robust and clinically aligned ambiguity-aware learning frameworks for ECG diagnosis.

Problem

• Addresses label ambiguity in real-world ECG diagnosis using Partial Label Learning.
• Evaluates PLL methods' robustness to structured and unstructured ECG label ambiguities.
• Identifies limitations of current PLL approaches for clinical ECG applications.

Innovation

• Adapted nine PLL algorithms for ECG diagnosis
• Evaluated methods using structured and unstructured ambiguity strategies
• Identified limitations and future directions for clinical PLL frameworks