Abstract
<title>Abstract</title> <p>Motion can distort cardiac electrical impedance tomography (EIT) and reduce the reliability of short-window stroke-volume (SV) estimation. We evaluated whether a nearby motion-free EIT segment can provide a local clean reference for suppressing a motion-contaminated cardiac EIT. Six clinical recordings with synchronized EIT, electrocardiography (ECG), and clinical-reference SV were analyzed. For each motion-contaminated target window, a principal component analysis (PCA) spatial basis was estimated from the preceding clean segment and the target was projected onto this basis before standard ECG-guided cardiac extraction. In 2,963 complete-case clinical-reference SV observations from the clean10-motion20 endpoint, clean-reference PCA reduced the median SV error from 8.28 mL to 7.83 mL compared with window-wise principal component analysis-independent component analysis (PCA-ICA), with subject-level median improvement in all six recordings. The benefit was concentrated on the high-baseline-error windows, whereas some already-low-error windows worsened. A 10 s clean reference did not show a clear loss relative to 20 s on matched targets and increased detector-level clean-reference availability. Local clean-reference PCA may therefore be useful as a reference-guided recovery strategy for motion-contaminated cardiac EIT, but the validation of new recordings and the reference-independent quality gating are required before routine use.</p>