{"paper":{"title":"Neural Surrogate Forward Modelling For Electrocardiology Without Explicit Intracellular Conductivity Tensor","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A neural network maps left atrial intracellular potentials to ECGs without needing explicit conductivity tensor inputs at inference time.","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Cesare Magnetti, Jakub Grzelak, Oleg Aslanidi, Shaheim Ogbomo-Harmitt","submitted_at":"2026-05-13T11:26:28Z","abstract_excerpt":"Accurate forward modelling is essential for non-invasive cardiac electrophysiology, particularly in atrial fibrillation, where electrical activation is highly disorganised. Conventional physics-based forward models require explicit specification of intracellular conductivity tensors, which are not directly measurable in clinical practice and introduce structural modelling errors. This proof-of-concept study presents a deep learning approach that learns a direct mapping from left atrial intracellular electrical potentials to far-field ECGs without requiring explicit intracellular conductivity i"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Despite training only on 74 subjects, the model achieved an R2 of 0.949 ± 0.037, highlighting potential to reduce structural uncertainty and improve non-invasive AF assessment.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The learned mapping from intracellular potentials to ECGs generalizes beyond the training set and captures the underlying physics sufficiently without explicit conductivity tensors.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"A deep learning surrogate learns to predict ECGs from atrial potentials with R²=0.949 without requiring conductivity tensor inputs.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A neural network maps left atrial intracellular potentials to ECGs without needing explicit conductivity tensor inputs at inference time.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"4cfe45bc991b9ba1c7744ffb41cb482568b4eaeb18243f843cac90b407d92ffe"},"source":{"id":"2605.13366","kind":"arxiv","version":1},"verdict":{"id":"f79750d8-4c95-428d-8297-056f4bd752f1","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T19:16:11.441927Z","strongest_claim":"Despite training only on 74 subjects, the model achieved an R2 of 0.949 ± 0.037, highlighting potential to reduce structural uncertainty and improve non-invasive AF assessment.","one_line_summary":"A deep learning surrogate learns to predict ECGs from atrial potentials with R²=0.949 without requiring conductivity tensor inputs.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The learned mapping from intracellular potentials to ECGs generalizes beyond the training set and captures the underlying physics sufficiently without explicit conductivity tensors.","pith_extraction_headline":"A neural network maps left atrial intracellular potentials to ECGs without needing explicit conductivity tensor inputs at inference time."},"references":{"count":6,"sample":[{"doi":"","year":2015,"title":"Forward Problem of Electrocardiography: Is It Solved? Circulation: Arrhythmia and Electrophysiology","work_id":"84aa4436-2cc3-4289-bb1a-32c0a61f16be","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"Predicting Atrial Fibrillation Recurrence by Combining Population Data and Virtual Cohorts of Patient -Specific Left Atrial Models","work_id":"a2b2c32d-1e17-4614-b186-854cba87b63d","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"In-Silico Investigation of the Right and Left Atrial Contributions to the P-Wave Morphology in ECG of Healthy and Atrial Fibrillation Patients","work_id":"d884527d-12fa-459d-873c-44e96639c5b7","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"DiffusionNet: Discretization Agnostic Learning on Surfaces","work_id":"04d701e7-c34e-4dac-aca2-b02810ad98de","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2019,"title":"Universal atrial coordinates applied to visualisation, registration and construction of patient specific meshes","work_id":"088ac29a-6763-4b0d-9fdc-dcb1b414d88c","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":6,"snapshot_sha256":"f84d0bd48d9c30379a4cd1d359b3164f4a17fbfee7398eadf129503d37f060f0","internal_anchors":0},"formal_canon":{"evidence_count":1,"snapshot_sha256":"1dacd6c0776e4e7e4fe3737ad0db72685b682db979bd5d7b538ce252f7d4928f"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}