How to Learn Electrical Impedance Tomography (EIT): From Boundary Voltages and Conductivity Inversion to Bedside Lung Ventilation, Perfusion and Physics-Aware Reconstruction

## Wait, What? EIT Tries to Image the Inside by Measuring Only the Outside Place a ring of electrodes around the chest. Inject a tiny alternating current through one pair of electrodes. Measure voltages on the others. Change the current-injection pair and repeat. The lungs are not wired directly. No electrode sits inside each voxel. Instead, every boundary measurement contains information from a broad volume of tissue. The reconstruction problem asks: > **which internal conductivity distribution could have produced this pattern of boundary voltages?** That inverse problem is difficult because many internal distributions can produce very similar surface measurements. ## The One-Sentence Answer **Learn EIT by tracing injected AC current → conductivity-dependent electric field → boundary voltages → forward finite-element model → regularized inverse reconstruction, then add electrode contact, body-shape error, difference imaging, temporal filtering, ventilation/perfusion physiology and uncertainty before turning an impedance image into a claim about regional lung recruitment, perfusion or pathology.** > **Educational biomedical engineering only. EIT findings require trained clinical interpretation and do not replace CT, clinical examination or emergency care.** # Beginner Layer — Conductivity ## Stage 1: Biological Tissues Conduct Electricity Differently Conductivity depends on water, ions, cell membranes, tissue structure and frequency. ## Stage 2: Air Is Much Less Conductive Than Blood or Saline The lung’s electrical properties change strongly as it fills with air. ## Stage 3: This Makes Breathing an EIT-Friendly Signal Inspiration generally increases thoracic impedance in ventilated lung regions. # Electrode Belt ## Stage 4: Place Electrodes Around the Chest Common systems use a belt with many evenly spaced electrodes. ## Stage 5: Electrode Position Defines the Measurement Plane Moving the belt changes what slice-like region is most sensitive. ## Stage 6: Skin Contact Matters Poor contact introduces noisy voltages, saturation and current-path distortion. # Current Injection ## Stage 7: Inject a Small Alternating Current AC avoids some polarization effects associated with DC. ## Stage 8: Current Amplitude and Frequency Must Remain Within Safety Limits The measurement is deliberately weak. ## Stage 9: Different Electrode Pairs Create Different Sensitivity Fields There is no single straight ray as in CT. Current spreads through the volume. # Boundary Voltages ## Stage 10: Measure Potentials at the Remaining Electrodes ## Stage 11: Each Measurement Depends on the Whole Conductivity Field A local conductivity change perturbs many electrode voltages. ## Stage 12: EIT Is a Diffuse-Field Imaging Method That is why spatial resolution is much lower than CT or MRI. # Forward Problem ## Stage 13: If Conductivity Were Known, Predicting Voltages Is the Forward Problem The electric potential satisfies a conductivity-weighted form of Laplace’s equation. Conceptually: **∇·(σ∇V) = 0** ## Stage 14: The Body Shape Must Be Modelled The boundary influences current flow. ## Stage 15: Electrodes Must Be Included A complete electrode model can include finite electrode size and contact impedance. # Finite-Element Model ## Stage 16: Divide the Body Cross-Section or Volume Into Elements Assign conductivity to each element. ## Stage 17: Solve for Voltage ## Stage 18: Compare Predicted and Measured Boundary Voltages This forward model is the engine inside iterative EIT reconstruction. # The Inverse Problem ## Stage 19: Reverse the Direction Measured voltages are known. Conductivity is unknown. ## Stage 20: The Problem Is Ill-Posed Small measurement errors can produce large conductivity changes if inversion is unconstrained. ## Stage 21: Regularization Is Required The algorithm adds assumptions such as smoothness, small changes, anatomical priors or sparsity. > **Every EIT image contains both measured voltage information and reconstruction prior.** # Sensitivity Matrix ## Stage 22: Linearized EIT Uses a Jacobian The Jacobian says how each voltage measurement changes when conductivity in each element changes. ## Stage 23: Sensitivity Is Not Uniform Boundary regions are generally easier to sense than the centre. ## Stage 24: Reconstructed Amplitude Depends on Position A same-sized conductivity change can appear differently at different locations. # Difference EIT ## Stage 25: Subtract a Reference Measurement Instead of recovering absolute conductivity: **ΔV → Δσ** ## Stage 26: Difference Imaging Cancels Many Fixed Errors Examples include imperfect electrode positions, uncertain body shape and fixed contact impedance. ## Stage 27: Most Lung EIT Uses Difference Imaging It is especially good for dynamic physiology. # Absolute EIT ## Stage 28: Absolute Imaging Attempts to Recover Conductivity Itself ## Stage 29: This Is Much Harder Because model mismatch is no longer cancelled. ## Stage 30: Accurate Geometry and Electrode Impedance Become Critical # GREIT ## Stage 31: GREIT Provides a Consensus Linear Reconstruction Framework for Lung EIT Its design goals include uniform amplitude response, small position error, limited ringing, controlled noise and relatively uniform resolution. ## Stage 32: GREIT Is Not the Only EIT Algorithm But it gives the field a common reconstruction benchmark. # Temporal Resolution ## Stage 33: EIT Can Reconstruct Many Frames per Second This is one of its greatest advantages. ## Stage 34: High Temporal Resolution Comes With Low Spatial Resolution The method is excellent for breath-to-breath change, regional dynamics and bedside trends. It is weak for detailed anatomy. # Ventilation Imaging ## Stage 35: Inspiration Changes Regional Lung Impedance The tidal impedance difference can approximate regional ventilation change. ## Stage 36: EIT Does Not Directly Measure Air Volume in Litres Per Voxel The relationship is a calibrated surrogate. ## Stage 37: Regional Distribution Is Often More Reliable Than Absolute Local Volume # Ventral–Dorsal Distribution ## Stage 38: Gravity and Position Change Where the Lung Ventilates EIT can track shifts between ventral, dorsal, dependent and non-dependent regions. ## Stage 39: Repositioning Also Moves Electrodes and Anatomy Physiological change and geometric change must be separated. # PEEP Titration ## Stage 40: Raising PEEP Can Recruit Previously Collapsed Lung It can also overdistend already open regions. ## Stage 41: EIT Can Estimate Regional Recruitment and Overdistension Trends ## Stage 42: The “Best PEEP” Is Not Defined by EIT Alone Clinical outcomes, pressures, oxygenation and patient condition matter. # 2025 ARDS Evidence ## Stage 43: A 2025 Systematic Review and Meta-Analysis Evaluated EIT-Guided PEEP Titration in ARDS It found promising mortality results but emphasized that the evidence base remained small and largely single-centre. ## Stage 44: Bedside Physiological Information Is Not Automatically Outcome-Proven Therapy # Pendelluft ## Stage 45: Gas Can Shift Between Lung Regions During One Breath A region may inflate while another deflates even with little net global volume change. ## Stage 46: EIT Can Detect Regional Timing Mismatches This is difficult to see in a global ventilator trace. # Regional Ventilation Delay ## Stage 47: Different Lung Regions Can Fill at Different Times Time-based EIT metrics can map delayed recruitment. # Pneumothorax ## Stage 48: Air Outside the Ventilated Lung Changes Impedance and Regional Ventilation ## Stage 49: EIT Can Detect Rapid Regional Changes But confirmation and emergency management depend on the clinical context and reference imaging. # Pleural Effusion ## Stage 50: Conductive Fluid Changes Thoracic Impedance Differently From Air Again, geometry and baseline matter. # Perfusion Imaging ## Stage 51: Blood Is Relatively Conductive Perfusion therefore contributes small impedance changes. ## Stage 52: Cardiac-Related Impedance Signals Can Be Extracted But they are weaker and mixed with other motion. ## Stage 53: Non-Contrast Cardiac-Signal Perfusion Is Difficult Absolute agreement with reference perfusion can be limited. # Saline-Contrast EIT ## Stage 54: Inject a Small Hypertonic Saline Bolus During a Controlled Measurement Saline has high conductivity. ## Stage 55: As the Bolus Passes Through the Pulmonary Circulation, Local Impedance Falls ## Stage 56: The Time Curve Acts Like an Indicator-Dilution Experiment This can produce regional perfusion maps. > **Contrast-enhanced EIT converts a deliberate conductivity tracer into a perfusion receiver.** # Breath-Hold Problem ## Stage 57: Traditional Saline-Contrast EIT Often Uses an End-Expiratory Pause This removes ventilation signal while the bolus passes. ## Stage 58: Some Patients Cannot Tolerate Apnea # 2026 Non-Apnea Frontier ## Stage 59: A January 2026 Randomized Crossover Study Tested Saline-Contrast EIT During Ongoing Ventilation Low-pass filtering removed much of the respiratory component. ## Stage 60: Non-Apnea Maps Agreed Strongly With Conventional Apnea Maps in the Study This expands the measurement envelope. ## Stage 61: Signal Processing Creates a New Assumption The filter must separate perfusion and ventilation without erasing true perfusion dynamics. # Ventilation–Perfusion Maps ## Stage 62: Combine Ventilation and Perfusion Images Classify regions as matched, ventilated but poorly perfused, or perfused but poorly ventilated. ## Stage 63: V/Q EIT Is Model Based It is not identical to nuclear-medicine or CT perfusion. # 2026 Pulmonary Embolism Frontier ## Stage 64: A 2026 Critical Care Study Compared EIT Perfusion With Dynamic Contrast CT and CTPA The investigators developed a wasted-ventilation index for pulmonary arterial occlusion. ## Stage 65: EIT Showed Strong Potential as an Adjunct When Conventional Imaging Is Difficult ## Stage 66: “Adjunct” Is the Correct Scientific Word A promising bedside receiver is not automatically a replacement for the reference diagnostic pathway. # 2026 V/Q Perspective ## Stage 67: A 2026 Intensive Care Medicine Review Describes EIT V/Q Mismatch as an Evolving Bedside Tool The emphasis is increasingly **monitor response and regional physiology**, not simply “make a pretty lung image.” # Neonatal EIT ## Stage 68: Radiation-Free Repeat Imaging Is Particularly Attractive in Neonates ## Stage 69: EIT Can Study Transition at Birth, Surfactant, Ventilation Mode and Position ## Stage 70: Small Body Size Makes Electrode Geometry Even More Important # Contact Impedance ## Stage 71: Electrode–Skin Interfaces Are Not Ideal Conductors ## Stage 72: Contact Impedance Can Change With Sweat, pressure, belt motion and gel drying ## Stage 73: A Noisy Electrode Can Corrupt Many Voltage Measurements Because reconstruction uses a coupled system. # Electrode Movement ## Stage 74: A Moved Electrode Changes the Forward Model ## Stage 75: Movement Can Look Like an Internal Conductivity Change This is one of the central model-mismatch failures. # Body Shape ## Stage 76: A Circular Reconstruction Model Is Only an Approximation to a Human Thorax ## Stage 77: CT-Derived or 3D Anatomical Models Can Improve Forward Prediction ## Stage 78: Anatomical Priors Can Also Bias the Solution If the anatomy model is wrong, a “more realistic” reconstruction can be confidently wrong. # Multifrequency EIT ## Stage 79: Tissue Conductivity Changes With Frequency Cell membranes act as frequency-dependent barriers. ## Stage 80: Measure at Several Frequencies This can add tissue-composition information. ## Stage 81: Multifrequency EIT Moves Toward Electrical Spectroscopy Plus Imaging # Brain EIT ## Stage 82: Stroke and Hemorrhage Can Change Tissue Impedance ## Stage 83: The Skull Severely Reduces and Redistributes Current This makes brain EIT much harder than lung EIT. ## Stage 84: Clinical Brain EIT Remains a Research Frontier High-value claims require strong reference validation. # Fast Neural EIT ## Stage 85: Neuronal Depolarization Produces Very Small Rapid Impedance Changes ## Stage 86: The Signal Is Orders of Magnitude Smaller Than Ventilation Changes This demands extreme instrumentation and averaging. # Machine Learning ## Stage 87: Neural Networks Can Learn Inverse Mappings From Boundary Voltages to Images ## Stage 88: The Ill-Posed Problem Does Not Disappear The prior has moved into the training distribution. ## Stage 89: A Model Can Learn Body Shape, Electrode Layout or Simulator Bias Instead of tissue conductivity. # Physics-Informed Learning ## Stage 90: Include the EIT Forward Equation in Training or Reconstruction This constrains the learned image to reproduce measured voltages. ## Stage 91: Data Consistency Is Stronger Than Visual Realism A smooth lung image that cannot forward-predict the electrodes is weak evidence. # Uncertainty ## Stage 92: EIT Should Ideally Report Where the Solution Is Poorly Constrained ## Stage 93: Centre Regions, Low-Contrast Changes and Electrode Failures Deserve Higher Uncertainty # Professional Layer ## Stage 94: Separate Seven Objects 1. true conductivity distribution; 2. injected current pattern; 3. electrode/contact state; 4. boundary-voltage data; 5. body/finite-element forward model; 6. regularized reconstructed conductivity change; 7. physiological interpretation. ## Stage 95: Professional EIT Is a Boundary-Field Inverse Problem > **Which ventilation or perfusion change remains identifiable after electrode movement, contact impedance, body-shape mismatch, anisotropic tissue conductivity, noise, regularization and learned priors are all allowed to explain the same boundary voltages?** # Evidence: What Makes an EIT Claim Strong? Stronger evidence combines electrode-quality checks, belt-position documentation, known current pattern and frequency, raw voltage quality metrics, body-shape model, reconstruction algorithm/version, reference-frame definition, repeated breaths, physiological perturbation with predicted direction, comparison to CT, spirometry or reference perfusion where appropriate, saline-contrast protocol when perfusion is claimed, forward-model residuals and uncertainty maps. # Misconceptions Worth Hunting – EIT sends one current ray straight through each voxel. – Boundary voltage directly equals local conductivity. – EIT has CT-like spatial resolution. – A brighter EIT region is an anatomical structure. – Difference EIT measures absolute conductivity. – More regularization always improves the image. – A circular body model is harmless. – Ventilation EIT directly measures litres of air in every voxel. – Cardiac-related EIT is automatically a quantitative perfusion map. – Saline-contrast EIT has no tracer-model assumptions. – EIT-guided PEEP has already proved universal outcome benefit. – Deep learning solves the uniqueness problem. – A plausible image can be accepted without forward-data consistency. # Transfer Check One electrode partially detaches and a new bright region appears nearby. Did lung ventilation necessarily increase? **No. Contact or movement artifact is a strong first explanation.** A patient is repositioned from supine to prone and dorsal ventilation rises. Is that automatically biological improvement? **Not automatically. Electrode geometry and body-boundary changes also need checking.** A saline-bolus perfusion map changes when the respiratory filter cutoff changes. Is perfusion uniquely measured? **No. Signal separation is model sensitive.** A neural reconstruction looks anatomically realistic but fails to reproduce measured boundary voltages. Is it scientifically acceptable? **No. It fails data consistency.** EIT and CT disagree on a small peripheral perfusion defect, but agree on the global trend. Must one system be broken? **No. Their spatial resolution, contrast mechanism and reconstruction assumptions differ.** # How We Know the Learning Has Held A learner should be able to explain conductivity and boundary-voltage measurement; explain current injection and electrode patterns; distinguish forward and inverse problems; explain why regularization is required; distinguish difference and absolute EIT; explain GREIT conceptually; interpret ventilation as an impedance-change surrogate; explain PEEP/recruitment and pendelluft applications; explain saline-contrast perfusion; discuss V/Q mapping; identify contact, movement and body-shape artifacts; and audit machine-learning EIT with forward consistency. # Model Limits EIT excels at continuous bedside monitoring with high temporal resolution and no ionizing radiation. It is limited by low spatial resolution, diffuse sensitivity, electrode dependence, forward-model mismatch, ill-posed inversion and difficult absolute quantification. Professional EIT keeps **electrode geometry + contact impedance + current frequency/amplitude + raw boundary voltages + body model + reconstruction prior + reference frame + physiology + comparison modality + uncertainty** visible together. # Teaching Guide Teach in this order: **conductivity → electrode belt → AC injection → boundary voltage → forward model → finite elements → inverse problem → regularization → difference EIT → GREIT → ventilation → PEEP/recruitment → pendelluft → saline perfusion → V/Q → electrode/body artifacts → brain/multifrequency EIT → physics-informed AI → uncertainty → validation.** # Connect This to the eduKate Learning Estate – https://edukatesengkang.com/2026/08/29/how-to-learn-bioelectricity-membrane-potentials-ion-channels/ – https://edukatesengkang.com/2026/08/28/how-to-learn-respiration-gas-exchange-cellular-bioenergetics/ – https://edukatesengkang.com/2026/08/28/how-to-learn-blood-circulation-oxygen-transport-hemodynamics/ – https://edukatesengkang.com/2026/08/29/how-to-learn-ultrasound-acoustic-imaging/ – https://edukatesengkang.com/2026/08/30/how-to-learn-electrochemical-impedance-spectroscopy-eis/ # Research Foundations and Further Learning – GREIT consensus reconstruction framework for 2D lung EIT. – Reviews of EIT in critical-care lung monitoring and neonatal lung imaging. – 2025 meta-analysis of EIT-guided PEEP titration in ARDS. – 2025 IEEE TMI comparison of contrast-enhanced EIT with pulmonary CT. – 5 January 2026 Critical Care: non-apnea saline-contrast lung-perfusion EIT. – 2026 Critical Care: accuracy of EIT ventilation/perfusion maps for pulmonary arterial occlusion. – 2026 Intensive Care Medicine: V/Q mismatch measured by bedside EIT—potentialities and challenges. # The Quiet Ending The beginner asks: “How can electrodes on the skin tell us anything about the lung inside?” The developing biomedical engineer asks: “Which conductivity change best explains these boundary voltages?” The advanced learner asks: “How much of the image belongs to physiology, and how much to electrode geometry and regularization?” And the professional asks: > **Which bedside physiological claim survives after the body model, electrode interface and ill-posed inverse problem are all treated as part of the measurement?**