Two Medtronic patents. Two different ways to decide what electrical stimulation should do next: question a noisy neural response, or use a different feedback signal.
US11559689B2, US12128235B2 · Independent PatentOracle research · 20 September 2026
One signal reflects the electrical pulse and its associated charge changes. The other is a physiological response travelling along nerve fibres. Confusing the two would obscure the central difference between these inventions.
The first patent describes suspending ECAP-based adjustment after persistent high readings and a minimum-output test. Its illustrated recovery restores stimulation settings, monitors the signal and waits before restarting the loop.
The second patent compares detectable neural responses with the number of pulses. When that ratio is insufficient, it can use the directly sensed stimulation signal instead. That decision does not itself establish that noise caused the low ratio.
The two disclosures address different failure modes in closed-loop electrical stimulation. US 11,559,689 B2 begins with an evoked compound action potential (ECAP), meaning a sensed neural response, and focuses on preventing a corrupted or persistently abnormal ECAP measurement from driving therapy progressively downward. US 12,128,235 B2 instead makes the directly sensed stimulation signal usable as feedback when ECAPs are absent or occur too infrequently for ECAP-based control. These are patent disclosures and inventor-reported technical observations, rather than evidence of a combined product or clinical benefit.
The distinction requested by the reader is fundamental. A stimulation signal, called an artifact in the first disclosure, is electrical potential sensed directly from delivery of the pulse and associated tissue charge changes. It occurs during, or substantially overlaps, pulse delivery. An ECAP is different: it is the physiological response after a pulse excites nerve fibers, with a propagating action potential reaching sensing electrodes later. A large pulse-correlated signal therefore does not by itself establish neural recruitment. Both disclosures also recognize that the direct stimulation signal can be larger than, and obscure, the ECAP.
US11559689B2 · Original FIG. 1. Blue leaders identify parts; they add no new signal paths.
Inspect original page 3 ↗The implant delivers programmed electrical stimulation through electrodes on the leads. This drawing supplies physical context, not the feedback policy.
The illustrated leads place electrodes near the spinal cord. The description allows other electrode configurations and target tissues.
A clinician or patient can use the programmer to send programs, settings or commands, within the described controls.
C1’s ordinary control loop receives a characteristic of an ECAP elicited by stimulation, compares that characteristic with an expected range, and changes a therapy-defining parameter. Disclosed ECAP features include amplitude, area, slope, peak ratios, timing, and latency; an example uses N1-P2 amplitude. The cited feedback examples identify changes to current or voltage amplitude, pulse width, frequency, and pulse shape. Electrode combinations are also programmable parameters, but these examples do not establish that the ECAP loop changes the electrode arrangement. In the illustrated high-ECAP response, processing circuitry reduces the current amplitudes of both control pulses (Ic) and informed pulses (Ii) when the reading exceeds the ECAP high threshold, ETH. When the reading returns within range, it moves them toward defaults. A reading below the low threshold can instead cause increments. These are disclosed examples, not fixed numerical settings mandated for every implementation. Figure 1 in C1 provides the physical context: implantable medical device 110, leads 130A and 130B near spinal cord 120, and external programmer 150. It is an orientation drawing, not a diagram of the feedback decisions.
US11559689B2 · Original FIG. 10. Follow the patent’s YES/NO branches. ERS means ECAP-responsive stimulation. Its restart box is labelled 2028 in the drawing and 1028 in the description.
Inspect original page 12 ↗The NO branch keeps the current settings. An outside-range reading proceeds to the high-threshold test; a low reading has a separate increment branch.
The high-threshold branch reduces the control and informed pulse amplitudes. This is one illustrated response, not a universal setting.
With control-pulse amplitude at the illustrated minimum, zero, a continuing high reading through the detection timer can trigger suspension.
The diagram suspends ERS, then raises the pulse amplitudes toward defaults and holds them there while monitoring continues. Suspending the loop does not mean stopping all stimulation.
The ECAP value must fall below the high threshold and the low-noise timer must expire before the pictured restart. Other recovery embodiments are described separately.
The important limit is that a high ECAP-like measurement may be coupled noise rather than a neural response. C1 gives external noise as an example that can be interpreted as an ECAP. If the controller keeps reducing output until a parameter reaches its minimum—illustrated by control-pulse amplitude Ic reaching approximately zero—and the sensed value remains above ETH through a noise-detection timer, the processor suspends or disables ECAP-responsive stimulation rather than continuing to reduce therapy. It may restore control and informed parameters to defaults, or optionally to a sub-default level, while sensed ECAPs are monitored but no longer used to adjust therapy. The sub-default alternative is described as a way to provide some stimulation while reducing possible overstimulation risk in an awkward posture.
A second C1 recovery route keeps the control-pulse amplitude at minimum while returning informed-pulse amplitude to its default. That matters because a control pulse is the pulse intended to elicit the measured ECAP: with Ic at zero, a detectable ECAP should not be produced by a control pulse. If the sensed value falls below ETL and remains low through a low-noise timer, Ic can be raised toward its default. A separate confirmation timer can then expire before the closed loop restarts. If the reading again exceeds ETH during that attempted recovery, Ic is reduced again. C1 also describes blocking a requested patient increase while the controller is actively changing parameters or values sit outside tolerance around defaults; decreases may remain permissible in examples.
Claim 1 recites receiving information indicative of an ECAP elicited by a stimulation pulse, executing the closed-loop policy, and determining whether the characteristic lies inside or outside an expected range. It does not expressly condition that execution step on an inside-range determination. When outside, the claim requires disabling the policy, setting the parameter to a threshold value, starting a noise-detection timer, and enabling the policy after that timer expires. Dependent claim 4 adds control and informed pulses, sensing the ECAP after a control pulse, ETH and ETL, and incremental changes to control- and informed-pulse parameters for a value above ETH. Claims 5 through 9 add particular suspension, low-noise-timer, restoration, and confirmation-timer conditions. Thus, the source supplies a state-machine-like set of alternatives, but does not specify universal timer durations, fixed thresholds, or one required fallback amplitude.
US12128235B2 · Original FIG. 1. The physical arrangement resembles the first patent’s drawing. The different feedback policy is established by the text and claims.
Inspect original page 5 ↗The disclosed device can deliver control pulses and therapy pulses. The text also describes temporary or external-device examples.
The two pictured leads carry electrodes. The description allows one lead, additional leads or other configurations.
The programmer can send settings and commands. Its presence alone does not tell the reader which feedback signal controls a later pulse.
C2 permits feedback control from the stimulation signal itself: the directly sensed electrical signature of the delivered pulse and related charge changes. The processor compares a measured characteristic, such as full or partial amplitude, slope, or area, with a target value or range and changes a parameter for a later pulse. Below a target minimum, it may increase a parameter; above a target maximum, it may decrease one; inside the range, it may hold the setting. The disclosed change can be a fixed step, proportional to the deviation, or function-based. C2 presents this signal as information that may reflect electrode-to-target distance and help choose stimulation parameters; it does not equate artifact amplitude with a direct measurement of a propagated neural response. Figure 1 in C2 provides the physical context: implantable medical device 110, leads 130A and 130B near spinal cord 120, and external programmer 150. It is an orientation drawing, not a diagram of the feedback decisions.
US12128235B2 · Original FIG. 16. This sheet shows stimulation-signal amplitude feedback. It does not depict the separate ECAP-count ratio test described in the text.
Inspect original page 22 ↗The feedback here is the stimulation signal associated with the pulse. It is distinct from a propagated neural response.
This flow uses amplitude. The description also discusses other measurable characteristics, including slope and area.
The measured amplitude is compared with a target range. The drawing preserves both outcomes rather than assuming that every pulse requires a change.
YES maintains one or more upcoming-pulse parameters; NO changes them. The description gives increase-below and decrease-above examples.
C2 addresses pulses that yield too few detectable ECAPs under its threshold-ratio test, including examples where the neural response is too small or absent. The processor may examine a sequence of pulses, count detectable ECAPs, calculate the ratio of detectable ECAPs to pulses, and compare that result with a threshold ratio. If the ratio is greater than the threshold, later control may use ECAP characteristics, optionally together with stimulation-signal characteristics. If it is not greater, later control may instead use stimulation-signal characteristics and not the corresponding ECAP characteristic. This can cover sub-threshold pulses that do not produce detectable ECAPs but still yield a directly detectable stimulation signal. Control and informed pulses may be interleaved; control pulses may or may not contribute to therapy, and longer informed pulses need not provide a fully measurable signal.
C2 changes the expected stimulation-signal target as well as the stimulation setting. An accelerometer can identify posture, and the processor may select a target range using both posture and the delivered pulse amplitude. Posture-associated transfer functions relate pulse amplitude to expected stimulation-signal amplitude; described examples distinguish supine, seated, and standing relationships. C2 also identifies a third, residual phase after the first and second pulse-related phases. In an example, that phase represents ionic rebalancing from residual charge and can supply feedback, although it may not be detectable when an ECAP is detectable. These are context-sensitive target-selection embodiments, not a declared method for identifying noise.
Independent claim 1 requires sensing a first delivered stimulation pulse, determining that a characteristic exceeds a target stimulation-pulse value, changing a parameter for a later pulse, detecting ECAPs from the pulse sequence, and determining whether the ECAP-to-pulse ratio exceeds a threshold. Claim 2 adds the upper-target case and a decrease. Claim 4 adds the lower-target case and an increase. Claims 6 and 7 distinguish the ECAP-plus-stimulation-signal branch from the stimulation-signal-without-corresponding-ECAP branch; claim 8 adds posture and amplitude selection of the target. The fallback parameter and restoration sequence is not an independent-claim requirement: dependent claim 3 adds setting a fallback parameter and restoring the prior value if a later sensed pulse no longer exceeds target. Claims 19 through 24 separately address residual-phase sensing and corresponding ratio branches.
The publications describe comparable feedback architectures: stimulation generation, electrodes and sensing circuitry, processor-selected pulse parameters, optional control/informed pulse schedules, and possible ECAP sensing. Their central difference is the decision made when feedback becomes unsuitable. C1 treats a persistently out-of-range ECAP measurement after reduction to a limit as a possible noise condition. It suspends ECAP-driven adjustment and returns therapy to default or an alternative level while checking whether the measurement stabilizes. C2 treats an insufficient proportion of detectable ECAPs as a basis to select direct stimulation-signal feedback instead. C2 does not prescribe C1’s timer-based noise-suspension sequence, and C1’s cited claims do not require C2’s ECAP-ratio selection logic.
When an ECAP is detectable, both disclosures can use neural-response feedback to change later stimulation. When the ECAP is too small or appears too infrequently, C2 formalizes a threshold-ratio decision and can use the directly sensed stimulation signal instead; C1 notes that stimulation signals can be useful when too few ECAPs are detectable, although its central claimed regime is ECAP range handling and possible-noise control. When a high reading persists despite output reduction, C1 treats the pattern as possible distortion and suspends the adaptive loop with timed re-entry checks. In C2, independent claim 1 instead requires a target-exceedance determination and parameter change; dependent claim 3, not claim 1, adds a fallback parameter followed by restoration when a later sensed pulse no longer exceeds target. Neither disclosure makes every abnormal signal noise or makes direct artifact sensing proof of efficacy.
The selected Figure 1 sheets are orientation drawings rather than control-flow diagrams. In C1, Figure 1 labels system 100, patient 105, spinal cord 120, implantable medical device 110, leads 130A and 130B, and external programmer 150. C2’s Figure 1 uses the same type of depicted system arrangement and labels. The figures help separate the implanted device-and-lead pathway from the external programmer. They do not depict C1’s ETH, ETL, timers, or suspension state, nor C2’s ECAP-ratio calculation, residual phase, or posture transfer functions. Those control details come from the descriptions and claims, not from the selected sheets.
Start with C1P21S01 for the source’s direct distinction between pulse artifact and ECAP. Then read C1P31S03 for ordinary ETH/ETL adjustment; C1P32S01 through C1P33S03 for minimum-output testing, timer-driven suspension, and restart; and C1P34S01 through C1P34S02 for the alternative that restores informed therapy while holding control pulses at minimum. Finish with claims 1 and 4 through 9 at C1P39S03 through C1P40S02, followed by device claims 13 through 23 at C1P40S02 through C1P41S02. This route separates the broad operational examples from the narrower claim dependencies.
For C2, begin with C2P28S01 through C2P28S03 for why direct pulse sensing is proposed when ECAPs cannot be detected and for the artifact-versus-ECAP definition. Next read C2P35S03 and C2P36S01 through C2P36S03 for the threshold-ratio branch and high, low, and in-range decisions. Read C2P37S02 through C2P37S04 and C2P43S01 through C2P43S03 for posture-specific targets, transfer functions, and residual-phase embodiments. Finally, read claims 1 through 8 at C2P50S03 through C2P51S02, then claims 19 through 27 at C2P52S01 through C2P53S01. In particular, compare independent claim 1 with dependent claim 3 before treating fallback-and-restoration behavior as required.
After reading the logic, return to the admitted original Figure 1 sheets. C1F1 identifies the C1 disclosed physical system: IMD 110 connects to leads 130A and 130B placed along spinal cord 120, with external programmer 150 outside the body. C2F1 identifies the analogous C2 layout. Use the images to anchor which components deliver, sense, and communicate, but use the cited text for the decision conditions and claims because neither sheet visualizes the feedback branches.
Two selected publications, with 39 of 51 identified description and claim pages supplied to the AI comparison. Four original drawings are explained here. All 94 original pages are included in the downloadable PDF and offline edition; this is a focused reading, not an exhaustive review of every page.
The PDF contains the explanation, annotations and every original page, with internal source links. The ZIP includes this web edition, the PDF, source and review records, and checksums. Downloaded files remain readable offline.
The comparison passed a complete source and image review after recorded editorial corrections. Those corrections and their provider usage remain in the original job history. Codex separately inspected the original control-flow sheets, source passages, callout targets and final presentation. The overlays are assisted editorial work, not independent human review.
The AI comparison reviewed FIG. 1 from each patent and selected complete text pages. The additional FIG. 10 (US11559689B2, original page 12) and FIG. 16 (US12128235B2, original page 22) were inspected directly for this edition, alongside their supporting text.
US11559689B2 omits description/claim pages 22, 27, 28 and 38 from the AI packet. US12128235B2 omits pages 34, 38 and 44–49. Every omitted page remains in the originals. The comparison does not establish clinical benefit or describe one combined marketed device.
The original patent is the source of the proposed capabilities. This independent example does not establish deployment, measured performance, a client relationship or endorsement by Medtronic.
This curated example includes additional editorial work and annotations on the original drawings. Comparable bespoke work receives its own scope and quote.
Standard automated explanations and comparisons are separate services. The reading plan and price are confirmed before paid work begins.
Tell us which patents and questions matter to you. We’ll agree the annotation, review and delivery scope before quoting a bespoke edition.
Discuss a similar annotated report ↗