Pavlo Radiuk

Research

Explainable artificial intelligence that domain experts can check, applied to medical imaging, energy infrastructure and urban mobility.

Approach

Most explanations of deep networks are heatmaps: they show where a model looked, not what it concluded in terms an expert would use. My work instead constructs a transition matrix — an explicit, linear mapping from the features a network has learned to the features a domain expert already reasons with — so that an explanation becomes a change of basis that can be inspected, tested and audited.

Recent work extends the approach with Lie-group equivariance, rough-set production rules that turn mapped features into readable decision rules, and conflict-aware auditing that shows where model and expert disagree. Explanations are evaluated against expert judgement; agreement reached Cohen’s κ of 0.89 for ECG and 0.80 for cardiac MRI.

Key publications

  1. CARDIA-X: Global Semantic Transition and Rough-Set Rules for Auditable Post Hoc Electrocardiographic Explainability

    P. Radiuk, O. Barmak, L. Klymenko, I. Krak

    AI 7(9):376, 2026.

    DOI PDF Code

  2. Equivariant Transition Matrices for Explainable Deep Learning: A Lie Group Linearization Approach

    P. Radiuk, O. Barmak, L. Bedratyuk, I. Krak

    Machine Learning and Knowledge Extraction 8(4):92, 2026.

    DOI PDF Code

  3. Explainable Deep Learning for Cardiac MRI: Multi-Stage Segmentation, Cascade Classification, and Visual Interpretation

    V. Slobodzian† (mentored researcher), O. Barmak, P. Radiuk, L. Klymenko, I. Krak

    Springer Science, Business Media LLC (preprint), 2026.

    DOI

All 19 publications on explainable AI

Medical AI

Explainable deep learning for ECG arrhythmia detection, cardiac MRI segmentation and classification, and pneumonia diagnosis from chest radiographs and CT — the line of work that began with my Ph.D. Explanations are expressed in clinical features and checked with clinicians.

Recent publications

  1. CARDIA-X: Global Semantic Transition and Rough-Set Rules for Auditable Post Hoc Electrocardiographic Explainability

    P. Radiuk, O. Barmak, L. Klymenko, I. Krak

    AI 7(9):376, 2026.

    DOI PDF Code

  2. Explainable Deep Learning for Cardiac MRI: Multi-Stage Segmentation, Cascade Classification, and Visual Interpretation

    V. Slobodzian† (mentored researcher), O. Barmak, P. Radiuk, L. Klymenko, I. Krak

    Springer Science, Business Media LLC (preprint), 2026.

    DOI

  3. EMTKD at the edge: An adaptive multi-teacher knowledge distillation for robust cardiac MRI classification

    O. Chaban† (mentored researcher), E. Manziuk, O. Markevych, S. Petrovskyi, P. Radiuk

    5th Edge Computing Workshop, CEUR Workshop Proceedings 3943, pp. 42–57, 2025.

    PDF

All 24 publications in this area

Energy and UAV inspection

UAV-based inspection of wind turbines and photovoltaic plants: defect detection from RGB, thermal and multispectral imagery, criticality assessment with fuzzy logic and adaptive inspection trajectories, developed with partners in Ukraine and Poland. Related work covers UAV localisation and precision agriculture.

Recent publications

  1. A Geo-consistent Weather Augmentation Method for Improving UAV Localization Robustness under Adverse Meteorological Conditions

    V. Vozniak† (mentored researcher), O. Barmak, P. Radiuk, I. Krak

    Journal of Intelligent & Robotic Systems, 2026.

    DOI

  2. Architecture of cyberphysical systems for UAV-based late-fusion defect detection in photovoltaic modules

    A. Sachenko, P. Radiuk, A. Lysyi, O. Melnychenko, O. Zastavnyy

    INNOVATIVE TECHNOLOGIES AND SCIENTIFIC SOLUTIONS FOR INDUSTRIES No. 1(35), pp. 79–99, 2026.

    DOI PDF

  3. Automated Synthesis of Hierarchical Deep Learning Cascades for Identifying Visually Similar Objects in UAV Imagery

    D. Borovyk† (mentored researcher), O. Barmak, P. Radiuk, I. Krak

    Technologies 14(6):360, 2026.

    DOI PDF

All 19 publications in this area

Smart mobility

Reinforcement-learning traffic-signal control and traffic-pattern recognition for the Khmelnytskyi pilot of the Horizon Europe project U_CAN, which supports Ukrainian cities on their path to climate neutrality.

Recent publications

  1. Topography-Aware Deep Reinforcement Learning with Contextual Reward Engineering for Sustainable and Efficient Urban Traffic Control

    O. Ryzhanskyi† (mentored researcher), O. Barmak, E. Manziuk, P. Radiuk, I. Krak

    Future Transportation 6(2):82, 2026.

    DOI PDF

  2. AI-Driven Traffic Signal Control System to Reduce CO2 Emissions

    O. Ryzhanskyi† (mentored researcher), V. Pavlyshyn† (mentored researcher), P. Radiuk, E. Manziuk, O. Barmak, I. Krak

    YAISD Workshops 2025, CEUR Workshop Proceedings 3974, pp. 18–27, 2025.

    PDF

  3. An Adaptive Machine Learning Approach to Sustainable Traffic Planning: High-Fidelity Pattern Recognition in Smart Transportation Systems

    V. Pavlyshyn† (mentored researcher), E. Manziuk, O. Barmak, P. Radiuk, I. Krak

    Future Transportation 5(4):152, 2025.

    DOI PDF

All 6 publications in this area

Data and software

  • Code · 2026

    transition-matrix-cardia-x

    CARDIA-X: Global Semantic Transition and Rough-Set Rules for Auditable Post Hoc Electrocardiographic Explainability

  • Code · 2026

    transition-matrix-etm-xai

    Equivariant Transition Matrices for Explainable Deep Learning: A Lie Group Linearization Approach

  • Code · 2026

    transition-matrix-dss

    SEMTRA: Global Semantic Transition and Rough-Set Rules for Auditable Post-Hoc Explainability

  • Code · 2024

    transition-matrix-dl

    Explainable Deep Learning: A Visual Analytics Approach with Transition Matrices

  • Code · 2021

    An-Early-Diagnosis-of-Pneumonia-on-Individual-Radiographs

    An Approach to Early Diagnosis of Pneumonia on Individual Radiographs based on the CNN Information Technology

  • Dataset · 2025

    Custom ACDC test set

    Prepared cardiac MRI images and segmentation masks for testing cardiac-MRI analysis models (Zenodo, with V. Slobodzian, O. Barmak, L. Klymenko and I. Krak).

† mentored junior researcher