A method of scalp area of action dose mapping and prevention of duplication control

Abstract

The application discloses a regional dose mapping and anti-repetition control method executed by a processor of a scalp pulse microdrop ejection device. The processor combines registration, pose, calibration and motion covariance into a predicted covariance to generate a conservative predicted footprint, F_pc=F_p⊕B(r_p); reads an official occupation D, an evidence uncertainty quarantine occupation Q and an unfinished reservation occupation P in a liquid mass dose field, calculates a candidate post-accumulation and an allowable candidate mass according to D_j^+=D_j+Q_j+P_j+w_p,j·q_p, and reserves the post-accumulation and the mass through an event identifier and a map version atom. When executed, the processor collects mass flow, valve state, working distance and pose evidence, forms effective, invalid or uncertain states according to definite conditions, and respectively idempotent submits, rolls back or quarantines according to a mass upper limit.

Description

Technical Field

[0001] This invention relates to the fields of scalp surface positioning, pulsed droplet injection control, liquid mass-dose mapping, and device output interlocking, and particularly to a method and system for forming a closed control chain from spatial positioning uncertainty, liquid mass-dose ledger, pre-output spatial reservation, physical output evidence, and anomaly recovery. Background Technology

[0002] Scalp pulsed microdroplet jetting devices need to sequentially deliver liquid to multiple discrete locations on the scalp surface. Known devices can use cameras, inertial sensors, or reference markers to determine the nozzle position and can determine the output volume for the next position based on previous output positions or delivered volumes. Therefore, simply creating a skin location map, recording the applied locations, or starting and stopping the nozzle according to historical doses are all existing technological approaches.

[0003] For example, US11524114B2 discloses determining the delivery amount at different locations based on the position of the needleless delivery device and the previous delivery amount; US11151781B2 discloses constructing a scalp spatial model using multiple reference signals and scalp images; US20180028755A1 discloses associating actual output or device status information with delivery events; and US6758845B1 discloses recording successfully treated locations to control subsequent outputs. These documents address location, dose, evidence, or treated area recording, but do not address how to avoid duplicate outputs when the electronic recording is inconsistent with the actual output state of a single droplet pulse.

[0004] If the controller only records a candidate position as completed after issuing a drive command, omissions will occur when the valve is not open, the fluid path is blocked, or the actuator head leaves. If the controller restores a candidate position to incomplete when evidence is missing, events that have already been output but for which evidence is incomplete may be executed again. Multiple candidate threads reading the same dose map simultaneously, or repeated submissions of the same event during communication retries, will also cause the recorded dose to be lower than the actual output dose.

[0005] Furthermore, droplet footprints have both area and direction, and errors exist in scalp registration, nozzle pose, nozzle calibration, and scalp movement. Comparing only the nozzle center position cannot determine whether the edge of the candidate droplet footprint overlaps with the historical footprint and causes the local liquid mass to exceed the upper limit; using only outcome-based conditions such as "confidence level qualified" or "conservative envelope" cannot provide a reproducible blocking boundary.

[0006] Therefore, a technical solution for controller execution is needed: to expand candidate footprints with explicit covariance relationships, record formal occupancy, evidence-uncertain isolation occupancy and incomplete reserved occupancy separately with the same liquid mass unit, make atomic reservations before output, and select idempotent commit, rollback or conservative isolation based on mass flow rate, valve, working distance and pose evidence. Summary of the Invention

[0007] I. Technical problems to be solved

[0008] The technical problem to be solved by this invention is: in the presence of positioning error, concurrent candidates, output failure, lack of evidence and power failure replay, to ensure that the electronic dose ledger in the scalp space unit does not underestimate the liquid mass occupation that may be caused by droplet pulses, and to block physical output when a candidate event may cause any space unit to exceed the hard limit.

[0009] II. Technical Solution

[0010] To solve the above-mentioned technical problems, the present invention provides a control system 100. For example... Figure 1 As shown, system 100 includes a scalp coordinate map module 110, a positioning and registration module 120, an effective event interface 130, an action footprint model 140, a multi-layer dose ledger 150, a candidate action prediction module 160, an anti-repetition judgment module 170, a transaction control module 180, and a session persistence module 190; the system is also connected to a mass flow sensor, a valve status sensor, a working distance sensor, a nozzle pose sensor, a droplet pulse action mechanism, and a droplet output enable gate.

[0011] The scalp coordinate mapping module 110 establishes a 3D triangular mesh S and a spatial cell set G bound to the session_ID. The positioning and registration module 120 outputs the transformation T_SN from the nozzle coordinate system N to the map coordinate system S, the root mean square error of registration e_reg, and the registration covariance Σ_reg. The candidate action prediction module 160 generates a nominal predicted footprint F_p from the candidate pose, the nozzle calibration model M, and the candidate liquid mass q_p, and sets Σ_p=Σ_reg+Σ_pose+Σ_cal+Σ_motion, r_p=s·√λ_max(Σ_p)+m, and F_pc=F_p⊕B(r_p) to obtain the conservative predicted footprint F_pc and its cross-spatial cell weights w_p,j.

[0012] The dose ledger 150 uses only liquid mass as the control dose field in this application and maintains formal occupancy D_j, evidence-uncertain isolation occupancy Q_j, and incomplete reserved occupancy P_j separately. For candidate event p and spatial unit g_j, the expected cumulative mass is D_j^+=D_j+Q_j+P_j+w_p,j·q_p. In this formula, "+" is ordinary mathematical addition, D_j^+ is the new expected quantity, and no compound assignment is performed on D_j.

[0013] The anti-duplication determination module 170 determines the candidate mass that will not cause any intersecting spatial cell to exceed the hard upper limit of liquid mass U_j according to q_allow=min_{j:w_p,j>0}[(U_j-D_j-Q_j-P_j) / w_p,j]. When q_p≤q_allow, it passes; when q_min≤q_allow<q_p, q_p is downquantized to the controllable mass step size of the device and recalculated; when q_allow<q_min, the registration error or covariance exceeds the limit, F_pc enters the forbidden area, or a reserved conflict occurs, it blocks.

[0014] Transaction control module 180 only reserves F_pc intersection space units atomically with event_ID and map_version in the pass state, and then re-opens the droplet output enable gate after verifying the version and pose drift. After execution, valid event interface 130 forms VALID, INVALID, or UNCERTAIN states based on mass flow rate, valve status, working distance, and pose evidence; VALID events update D_j, INVALID events release P_j without incrementing D_j, and UNCERTAIN events update Q_j according to the conservative actual footprint F_c and the actual mass upper bound q_i^U. The same event_ID generates only one occupancy update in the transaction log.

[0015] III. Beneficial Effects

[0016] The predicted outer radius is calculated using the maximum eigenvalue of the covariance, the confidence factor, and the upper bound of the model residuals, allowing the influence of registration error and nozzle calibration error on candidate footprints to be verified through matrices, calibration records, and dimensional gauges.

[0017] D_j, Q_j, P_j and the candidate increment are all in units of liquid mass. It is expected that there will be no conversion or mutual cancellation between different physical quantities in the cumulative formula. Whether the candidate exceeds U_j can be directly verified by the addition and overlapping weights of the same unit.

[0018] q_allow converts the hard upper limit per spatial cell into the maximum liquid mass that the candidate can output; when the maximum allowable amount is lower than the device's minimum stable output mass q_min, the droplet output enable gate remains inactive.

[0019] The event_ID, map_version, and F_pc are written to the same atomic reserved transaction before physical output, which avoids two concurrent candidates obtaining permission based on the old dose ledger at the same time.

[0020] Mass flow rate, valve status, working distance, and pose evidence collectively define the event state, preventing uncoupled droplet pulses from being written into formal occupancy based solely on drive commands or single contact signals.

[0021] Uncertain events are not cleared, but are written to the isolated occupant Q_j according to F_c and q_i^U; subsequent candidates are calculated based on D_j+Q_j+P_j, which can avoid the repeated execution of outputs that have occurred but have incomplete evidence.

[0022] Append-based event logging and the unique event_ID key ensure that network retries, power outage recovery, and transaction replay remain idempotent; these effects can be verified through concurrent reservation tests, repeated commit tests, sensor missing tests, and state-by-state power outage tests, rather than relying on unverified physiological or therapeutic effects. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the overall architecture of the scalp pulse microdroplet dose mapping and anti-repetition control system of the present invention;

[0024] Figure 2 A schematic diagram of the scalp surface coordinates, reference markers, and spatial units at the session level;

[0025] Figure 3 A schematic diagram showing the relationship between the nominal predicted footprint, the conservative predicted footprint after covariance expansion, and the actual footprint after execution.

[0026] Figure 4 A schematic diagram illustrating the generation relationship between immutable event records, overlapping weights, and the liquid quality D / Q / P dose ledger;

[0027] Figure 5 This diagram illustrates the expected cumulative liquid mass, the permissible candidate mass, and the criteria for passing, reducing, or blocking.

[0028] Figure 6 A schematic diagram of the state machine for reserving, executing, committing, rolling back, and isolating transactions;

[0029] Figure 7 The control flow diagram consists of spectrum establishment, prediction expansion, atomic reservation, evidence classification, occupation update and interruption recovery.

[0030] Explanation of reference numerals in the attached figures

[0031] 100—Control System; 110—Scalp Coordinate Map Module; 111—Map Establishment Unit; 112—Spatial Unit Division Unit; 113—Map Version Unit; 120—Positioning and Registration Module; 121—Pose Acquisition Unit; 122—Coordinate Transformation Unit; 123—Registration Uncertainty Unit; 130—Valid Event Interface; 131—Mass Flow Rate and Valve Evidence Unit; 132—Working Distance and Pose Evidence Unit; 133—Event State Determination Unit; 140—Action Footprint Model; 141—Nominal Predicted Footprint Unit; 142—Actual Footprint Unit; 143—Conservative Footprint Unit; 15 0—Dose Ledger; 151—Formal Occupation D; 152—Isolation Occupation Q; 153—Reserved Occupation P; 154—Immutable Event Record; 160—Candidate Action Prediction Module; 161—Candidate Parameter Unit; 162—Overlapping Weight Unit; 170—Anti-Duplicate Judgment Module; 171—Expected Accumulation Unit; 172—Allowable Candidate Quality Unit; 173—Pass, Decrease, or Block Output Unit; 180—Transaction Control Module; 181—Atomic Reservation Unit; 182—Microdroplet Output Enable Unit; 183—Idempotent Commit Unit; 184—Rollback or Isolation Unit; 190—Session Persistence Module. Detailed Implementation

[0032] The following describes exemplary embodiments of the present invention in conjunction with the accompanying drawings. All thresholds represent device parameters that can be obtained through sensors, calibration records, or geometric calculations; numerical examples are used to illustrate the calculation process and are not intended as scalp safety limits or prototype test results.

[0033] Example 1: System Connections, Terminology, and Data Objects

[0034] like Figure 1 As shown, the scalp coordinate mapping module 110, the positioning and registration module 120, and the candidate action prediction module 160 form a pre-output spatial prediction chain; the mass flow rate and valve evidence unit 131, the working distance and pose evidence unit 132, the event state determination unit 133, and the action footprint model 140 form a post-output evidence chain; the multi-layer dose ledger 150, the anti-duplication determination module 170, and the transaction control module 180 close the two chains. The physical output of this embodiment is only scalp pulsed droplet injection.

[0035] Candidate events C_p are represented at least as {candidate_ID, session_ID, map_version, pose_p, F_p, F_pc, q_p, w_p, j, Σ_p, calibration_version}. Actual event records R_i are represented at least as {event_ID, session_ID, map_version, transform_version, calibration_version, E_i, pose_i, F_a, F_c, q_i, q_i^U, time_range, evidence_hash}. The event_ID is unique within the session and serves as the common association key for space reservation, sensor evidence, and occupancy updates.

[0036] Event status E_i is VALID, INVALID, or UNCERTAIN. VALID requires the valve to be open, the mass flow integral to reach q_det, the working distance to be within the allowable range throughout, and the pose drift to be acceptable; INVALID requires the valve to be closed at all times or the upper limit of the mass measurement to remain below q_det; all other events with detectable outputs but insufficient evidence of coupling, synchronization, or range are classified as UNCERTAIN. The three statuses respectively lead to formal commit, rollback, and isolation paths.

[0037] Example 2: Session-level scalp coordinate map and spatial unit

[0038] like Figure 2 As shown, the depth camera begins mapping after acquiring a continuous scalp surface and at least three non-collinear reference markers at the nozzle's position. The mapping unit 111 reconstructs the depth point cloud into a three-dimensional triangular mesh S; if the number of continuous surfaces or reference markers is insufficient, an executable session is not established and the droplet output enable gate remains closed.

[0039] The positioning and registration module 120 first uses the reference marker to obtain the initial rigid body transformation, and then calculates T_SN by minimizing the distance from the scalp sampling point to the triangular mesh surface. e_reg is the root mean square distance from the points participating in the registration to the corresponding mesh surface, and Σ_reg is estimated by the registration Jacobian matrix and the residual. When e_reg > e_max or λ_max(Σ_reg) > σ_reg,max^2, the anti-duplication determination module 170 outputs the blocking status.

[0040] Spatial cell partitioning unit 112 divides the triangular mesh into fine spatial cells g_j with a feature size no larger than half the minor axis of the nominal droplet footprint. Map versioning unit 113 stores the reference marker coordinates, mesh topology, T_SN, e_reg, Σ_reg, map_version, and checksum.

[0041] Example 3: Conservatively predicted footprint before output and actual footprint after output

[0042] like Figure 3 As shown, the nozzle calibration model 140 takes the working distance d, the angle α between the nozzle axis and the local scalp normal, the driving pressure p, and the pulse width t_p as inputs, and outputs the center, major axis, minor axis, and direction of the elliptical nominal predicted footprint F_p, as well as the candidate liquid mass q_p, from the lookup table corresponding to calibration_version.

[0043] Nozzle calibration can be performed on a simulated scalp device with a weighing membrane and a two-dimensional aspiration array. For each group of d, α, p, and t_p, microdroplet pulses are repeatedly output, the mass difference before and after weighing is given as q_p, and the aspiration array or tracer image is given as the footprint boundary; the 95th percentile of the boundary error of the independent verification sample is taken as m, and the covariance of repeated measurements of the footprint center is taken as Σ_cal.

[0044] Before outputting, let Σ_p = Σ_reg + Σ_pose + Σ_cal + Σ_motion, and calculate r_p = s·√λ_max(Σ_p) + m and F_pc = F_p⊕B(r_p). Σ_pose comes from the nozzle pose calculation, and Σ_motion comes from the displacement of consecutive scalp surface frames within the candidate time window; s is fixed and written to calibration_version. The disabled zone, hard upper bound, and space reservation all use F_pc instead of the nozzle center or F_p.

[0045] After execution, F_a is generated based on the actual nozzle pose and the calibration lookup table corresponding to the actual drive parameters. The registration, pose, scalp motion, and quality evidence errors during execution are synthesized into Σ_i, and F_c is generated according to r_i=s·√λ_max(Σ_i)+m and F_c=F_a⊕B(r_i). The VALID event uses F_a to update D_j, and the UNCERTAIN event uses F_c to update Q_j.

[0046] Example 4: Dosage Field, Compatible Units, and Spatial Accumulation

[0047] like Figure 4 As shown, the dose ledger 150 sets a liquid mass dose field, with units of milligrams or grams, but fixed to one unit within the same session. Formal occupancy D_j represents the liquid mass confirmed for delivery by the VALID event; Isolation occupancy Q_j represents the upper bound of the liquid mass that may be delivered by the UNCERTAIN event; Reserved occupancy P_j represents the candidate liquid mass temporarily occupied by the RESERVED or EXECUTING event before completion.

[0048] For footprint F_i and spatial unit g_j, the overlap weight is calculated according to w_i,j=∫_(g_j)K_i(x)dx / ∫_(F_i)K_i(x)dx, where K_i(x) is the calibrated liquid mass surface density kernel, and is normalized to make Σ_jw_i,j=1. The increments of D_j, Q_j, and P_j are all multiplied by the liquid mass using the overlap weight of the corresponding footprint.

[0049] For VALID events, formal occupancy is updated according to D_j(n) = D_j(n-1) + w_i,j·q_i; for UNCERTAIN events, isolation occupancy is updated according to Q_j(n) = Q_j(n-1) + w_c,j·q_i^U; for candidate events that have not yet been completed, P_j is equal to the sum of all intersecting valid reservations w_p,j·q'_p. All three layers are stored in the same liquid mass unit.

[0050] The mass flow sensor outputs m_dot(t) and an upper bound of error ε_m(t). The actual liquid mass is taken as q_i = ∫m_dot(t)dt, and the upper bound of the mass error is taken as ε_q = ∫ε_m(t)dt. When the signal is not saturated, q_i^U = q_i + ε_q; when the signal is saturated, the upper limit of the liquid mass under the current pressure and pulse width is obtained from the lookup table corresponding to calibration_version as q_i^U.

[0051] Immutable event log 154 is appended to without in-place overwriting. Both the formal dose map D and the isolation map Q are materialized views reconstructed based on event status, footprint, overlap weights, and fluid mass; the original record and checksum are preserved when updating map_version or calibration_version.

[0052] Example 5: Pre-execution estimated dosage and three-level judgment

[0053] like Figure 5 As shown, for candidate event p and spatial unit g_j, the expected cumulative amount is D_j^+=D_j+Q_j+P_j+w_p,j·q_p. This formula does not rewrite D_j into itself, nor does it use compound assignment in programming language; each term represents the liquid mass allocated to g_j.

[0054] The allowable candidate quality is q_allow=min_{j:w_p,j>0}[(U_j-D_j-Q_j-P_j) / w_p,j]. If q_p≤q_allow, the candidate quality passes; if q_min≤q_allow<q_p, then it is downquantized by a controllable step size Δq to q'_p=floor(q_allow / Δq)·Δq and recalculated; if q_allow<q_min, the device cannot stably output a droplet pulse without exceeding the limit, so it outputs a blocked state.

[0055] When F_pc enters the restricted area, e_reg or Σ_reg exceeds the limit, map_version mismatch occurs, or F_pc conflicts with other uncompleted reservations, a blocking status is output regardless of the size of q_allow. Geometric overlap itself does not necessarily block; as long as the expected cumulative amount does not exceed U_j and there are no reservation conflicts, controlled partial overlap can be performed.

[0056] Example 6: Atomic Reservation, Concurrency Protection, and Idempotent Commit

[0057] like Figure 6 As shown, the states include IDLE, RESERVED, EXECUTING, COMMIT, ROLLBACK, and QUARANTINE. After a successful determination, the transaction control module 180 compares map_version, checks for space conflicts, writes event_ID, F_pc, w_p,j, q'_p, and P_j within the same transaction, and increments the transaction version; if any step fails, the droplet output enable gate is not opened.

[0058] When the RESERVED record has not timed out, the map_version is consistent, and the drift of the current pose relative to the reserved pose is not greater than δ_pose,max, both the reserved valid latch and the processor permission terminal are valid, and only then does the droplet output enable gate transmit the driving pulse to the droplet pulse action mechanism. During execution, each sensor sample and timestamp are written to the cache corresponding to event_ID.

[0059] During VALID, R_i is appended, P_j is released, and D_j is updated; during INVALID, P_j is released and D_j is not incremented; during UNCERTAIN, the execution lock and P_j are released, and Q_j is incremented by F_c and q_i^U. Network retransmissions with the same event_ID only read the original transaction result; when migrating from UNCERTAIN to VALID or INVALID, a separate state migration transaction is used, and the original isolation level is deducted from Q_j before committing D_j or completing a rollback.

[0060] Example 7: Interface for Valid Event Evidence

[0061] The sampling period of the mass flow signal is no greater than one-tenth of the pulse width t_p. A valve status sensor detects the liquid valve core position; the measuring axis of the working distance sensor is parallel to the nozzle axis; and the nozzle pose sensor outputs six-degree-of-freedom pose and covariance. All four types of evidence include event_ID and a timestamp.

[0062] A signal is classified as VALID only when the valve is open, q_i ≥ q_det, the mass flow rate sample is complete, all d(t) are within [d_min, d_max], and the pose drift is no greater than δ_pose,max. A signal is classified as INVALID if the valve is always closed or q_i + ε_q < q_det. A signal is classified as UNCERTAIN if there is mass signal saturation, missing valve status, missing or out-of-bounds working distance samples, out-of-limit pose drift, or the time difference between any two types of evidence exceeds τ_max.

[0063] Example 8: Online Processes, Interruption Recovery, and Local Drift

[0064] like Figure 7 As shown, step S101 involves acquiring continuous scalp surface data and reference markers to establish a session map; S102 involves calculating T_SN, e_reg, and Σ_reg and performing a registration threshold check; S103 involves generating F_p and expanding it to F_pc according to the covariance formula; S104 involves querying D_j, Q_j, and P_j, and calculating D_j^+ and q_allow; S105 involves reserving atoms after passing the check; S106 involves verifying map_version and pose drift and then enabling droplet output; S107 involves executing and acquiring four types of evidence; S108 involves forming VALID, INVALID, or UNCERTAIN states; and S109 involves submitting D_j, rolling back P_j, or isolating Q_j, respectively.

[0065] A session packet includes at least the session_ID, triangular mesh, reference flags, event log, D_j, Q_j, P_j, U_j, map_version, transform_version, calibration_version, checksum, and sensor cache index. After a power outage, incomplete reservations and caches are read based on the event_ID: if it can be proven that no output was generated, a rollback is performed; if all valid conditions are met, an idempotent commit is performed; otherwise, all are isolated.

[0066] After the device is repositioned, the common reference markers and scalp surface are reacquired. When the root mean square error of the registration of the old and new common reference markers is not greater than e_recover,max, the new transform_version is saved and D_j and Q_j are reconstructed according to the event footprint; when it is greater than e_recover,max, the original session and its occupancy are frozen, the historical dose is not shifted to the new spectrum, nor is the same session continued after being cleared.

[0067] Session data uses a randomized session_ID and does not store direct identity identifiers; scalp surface data, event logs, and sensor caches are preferably processed on the device side or controlled edge nodes and managed according to access control and retention periods. The control determination of this invention is not based on disease diagnosis or treatment effect.

[0068] Example 9: Droplet mass calibration, evidence threshold, and output gate

[0069] q_min and Δq are determined by repeated weighing with the driving pressure p and pulse width t_p gradually changed under fixed d and α: the minimum average mass with a coefficient of variation not exceeding the product's set upper limit and an output higher than q_det in each step is taken as q_min, and the minimum mass difference between adjacent stable driving gears is taken as Δq. q_det is taken as the upper limit of the no-load noise mass integral plus a predetermined measurement margin.

[0070] d_min, d_max, and δ_pose,max are determined by the effective input range of the nozzle calibration model; e_max and σ_reg,max are determined by the correspondence between the registration error and the footprint position error on the simulated scalp. All thresholds, along with the sensor model, sampling rate, calibration date, and firmware version, are written to calibration_version. Version changes invalidate unexecuted reservations.

[0071] Example 10: Complete Computation and Transaction State Example

[0072] A square spatial unit with a side length of 2 mm is established on a 20 mm × 20 mm local scalp patch. The candidate event F_p is an ellipse with a major axis of 6 mm and a minor axis of 4 mm; r_p = 1 mm is calculated according to Σ_p, s, and m, resulting in F_pc with a major axis of 8 mm and a minor axis of 6 mm. The normalized overlap weights of F_pc with g1 to g4 are 0.40, 0.30, 0.20, and 0.10, respectively.

[0073] Let the dosage unit be mg, and let D_j for g1 to g4 be 6.0, 8.0, 8.7 and 4.0 mg, Q_j be 0, 0.4, 0 and 0 mg, P_j be 1.0, 0, 0 and 0 mg, U_j be 10.0 mg, q_min be 0.5 mg and Δq be 0.1 mg.

[0074] If the initial q_p is 8.0 mg, then the candidate mass margins calculated by hard upper bound for the four space units are 7.50, 5.33, 6.50, and 60.00 mg, respectively, so q_allow is 5.33 mg. The processor downquantizes by Δq to obtain q'_p = 5.3 mg, and the recalculated D_j^+ are 9.12, 9.99, 9.76, and 4.53 mg, respectively, all of which do not exceed U_j, thus allowing entry into the atom reservation.

[0075] The transaction control module establishes a RESERVED record using event_ID=E1007, the current map_version, F_pc, w_p,j, and q'_p. If, after execution, q_i is measured to be 4.8 mg, the weights of F_a and g1 to g4 are 0.35, 0.35, 0.20, and 0.10, respectively, and the four types of evidence satisfy the VALID condition, then D_j is updated to 7.68, 9.68, 9.66, and 4.48 mg, respectively; D_j remains unchanged when E1007 is resubmitted.

[0076] If the valve status of E1007 remains closed and q_i + ε_q < q_det, then it is classified as INVALID and P_j is released; if the mass flow signal indicates output but the working distance sample is missing, then it is classified as UNCERTAIN, and Q_j is updated with the overlap weight of F_c and q_i^U. The above values ​​only verify the formula and transaction path and do not represent product safety limits.

[0077] Example 11: Prototype Verification and Application for Supporting Documents

[0078] The mapping and expansion verification used a simulated scalp with known 3D coordinates, a repeatable curved surface, and fixed reference markers. The values ​​of e_reg, λ_max(Σ_reg), F_p boundary error, F_pc coverage of the actual footprint, and e_recover after repositioning were reported. The nozzle calibration was performed by repeatedly weighing and recording the 2D footprint at the combination points of d, α, p, and t_p.

[0079] Anti-duplicate verification includes at least the following scenarios: complete center duplication, overlapping edges, non-duplicate center but F_pc tail exceeding the limit, q_allow between q_min and q_p, q_allow below q_min, entry into the forbidden zone, and reserved conflict scenarios. In the concurrency test, two intersecting candidates are submitted simultaneously, verifying that only one transaction writes to P_j; in the idempotency test, the same event_ID is sent repeatedly, verifying that D_j or Q_j only increments once.

[0080] Evidence tests simulated scenarios including valve not open, missing or saturated mass flow samples, working distance exceeding limits, pose drift exceeding limits, and evidence time difference exceeding limits. Power-off tests simulated reserving, executing, writing but not submitting, and submitting but not responding states, and verifying the restored COMMIT, ROLLBACK, or QUARANTINE. Currently, prototype test results are not provided. During formal implementation, original sensor data, event logs, calibration tables, threshold sources, and firmware hashes should be preserved.

[0081] Other implementation methods

[0082] The scalp surface coordinate map can be generated using different depth cameras without changing the 3D triangular mesh, reference markers, e_reg and Σ_reg outputs; the nozzle calibration model can use a piecewise lookup table or a regression table that interpolates d, α, p and t_p, but must output F_p, q_p, Σ_cal and m, and be subject to calibration_version constraints.

[0083] Atomic reservations can be implemented using an embedded database or an append-only transaction log that supports comparison and exchange; space conflicts can be determined using a raster index or an R-tree index. Regardless of the storage implementation used, physical output is only allowed if the reservation transaction succeeds and both inputs to the droplet output enable gate are valid, and the event_ID unique key must override the updates to D_j and Q_j.

[0084] This application does not use gas, acoustic, optical, radio frequency, or mechanical action as the output mode covered by the claims, nor does it use diagnostic results, physiological indicators, or therapeutic effects as control conditions. The scope of protection of this invention is defined by the control chain of the scalp pulsed microdroplet jetting device as specified in the claims.

Claims

1. A method for controlling dose mapping and preventing duplicate application in the scalp application area of ​​a scalp pulsed microdroplet jetting device, characterized in that, Executed by the processor of the scalp pulsed droplet ejection device, the process includes: acquiring scalp surface data and at least three non-collinear reference markers from a positioning sensor; establishing a scalp surface coordinate map bound to a session identifier and divided into multiple spatial units; calculating the coordinate transformation from the nozzle coordinate system to the scalp surface coordinate map; and outputting the registration root mean square error e_reg and the registration covariance matrix Σ_reg; before generating the drive signal for the next droplet pulse, generating a nominal predicted footprint F_p based on the candidate nozzle pose, nozzle calibration model, and candidate liquid mass q_p; and combining Σ_reg, the candidate pose covariance Σ_pose, the nozzle calibration covariance Σ_cal, and the scalp motion covariance within the candidate execution time window. The predicted covariance Σ_p is obtained by summing Σ_motion. The predicted expansion radius is determined according to r_p=s·√λ_max(Σ_p)+m, and a conservative predicted footprint is generated according to F_pc=F_p⊕B(r_p), where s is the confidence factor bound to the calibration version, λ_max is the largest eigenvalue of the matrix, m is the upper bound of the residual of the nozzle calibration model, and B(r_p) is a disk with radius r_p. The overlap weight w_p,j of F_pc and the j-th intersecting spatial unit g_j is calculated, and the sum of w_p,j of all intersecting spatial units is 1. In the dose ledger with liquid mass as the uniform unit, the formal occupancy D_j formed by the submitted valid events in the j-th spatial unit is read. The isolation occupancy Q_j formed by events with uncertain evidence and the reserved occupancy P_j formed by events that have been reserved but not yet completed are calculated according to D_j^+=D_j+Q_j+P_j+w_p,j·q_p, and the allowable candidate mass is calculated according to q_allow=min_{j:w_p,j>0}[(U_j-D_j-Q_j-P_j) / w_p,j], where U_j is the hard upper limit of the liquid mass of the j-th spatial unit; when e_reg≤e_max, λ_max(Σ_reg)≤σ_reg,max^2, F_pc does not enter the forbidden area, and F_pc does not conflict with other uncompleted reservations. When q_p ≤ q_allow, the output is in the pass state; when the above space and registration conditions are met and q_min ≤ q_allow < q_p, the candidate liquid mass is reduced to a controllable mass q'_p not greater than q_allow and F_p, F_pc, w_p,j and D_j^+ are recalculated; when q_allow < q_min, e_reg > e_max, λ_max(Σ_reg) > σ_reg,max^2, F_pc enters the forbidden zone or a reserved conflict occurs, the output is in the blocking state and the microdroplet output enable terminal is kept inactive, where q_min is the minimum liquid mass that the scalp pulse microdroplet jetting device can stably output after calibration;After the output passes, a space reservation is created using the session-unique event identifier event_ID, the current map version map_version, F_pc, w_p,j, and the final candidate liquid mass atoms. The droplet output enable terminal is activated only when the space reservation is valid, the current map version is still map_version, and the drift of the current nozzle pose relative to the reserved pose is no greater than δ_pose,max. During the droplet pulse execution, mass flow rate signals, valve status signals, nozzle-to-scalp working distance signals, and nozzle pose signals are acquired, according to q_i=∫_( The actual output liquid mass q_i is obtained from t_0)^(t_1)m_dot(t)dt. When the valve status indicates open, q_i≥q_det, the mass flow signal is complete, the working distance d(t) is always within [d_min,d_max], and the pose drift is no greater than δ_pose,max, the event is classified as a valid state. When the valve status indicates closed throughout the droplet pulse and q_i+ε_q<q_det, the event is classified as an invalid state. All other events besides the valid and invalid states are classified as states of uncertain evidence. The uncertain states of the evidence include at least the following: mass flow rate signal saturation, inconsistency between valve status and mass flow rate signal, missing working distance samples, and time difference between mass flow rate signal and pose signal greater than τ_max, where ε_q is the upper bound of mass measurement error and q_det is the lower bound of detectable output mass. The actual footprint F_a is generated based on the actual nozzle pose and nozzle calibration model during execution, and a conservative actual footprint F_c is generated using the covariance and model residuals during execution, following the same expansion relationship as F_pc. In the valid state, event records are appended using event_ID and... The formal occupancy of the F_a intersecting space unit is updated idempotently according to D_j←D_j+w_i,j·q_i; in the invalid state, the space reservation is released without increasing D_j; in the state of uncertain evidence, the execution lock is released, and the isolation occupancy of the F_c intersecting space unit is updated to the upper bound of the liquid mass according to Q_j←Q_j+w_c,j·q_i^U, where q_i^U=q_i+ε_q, and when the mass flow rate signal is saturated, q_i^U takes the upper limit of the liquid mass of the current driving parameter in the calibration table; repeated submissions with the same event_ID do not increase D_j or Q_j again.

2. The method according to claim 1, characterized in that, The scalp surface coordinate map is a three-dimensional triangular mesh acquired by a depth camera; the at least three non-collinear reference markers are fixed to a positioning frame that does not move relative to the scalp; the processor obtains the coordinate transformation through initial registration of the reference markers and iterative registration of scalp surface points to the mesh surface; e_reg is the root mean square distance from the scalp sampling point participating in the registration to the corresponding mesh surface.

3. The method according to claim 1, characterized in that, The nozzle calibration model is a lookup table indexed by the working distance d, the angle α between the nozzle axis and the local scalp normal, the driving pressure p, and the pulse width t_p. The lookup table outputs the major axis, minor axis, direction, and candidate liquid mass q_p of the ellipse F_p. The upper limit m of the residual is taken as the 95th percentile of the footprint boundary error in the independent calibration samples.

4. The method according to claim 1, characterized in that, Σ_pose is formed by the pose calculation covariance of the nozzle pose sensor, Σ_cal is formed by the covariance of the footprint center in the repeated nozzle calibration test, and Σ_motion is formed by the displacement covariance of continuous scalp surface frames within the candidate execution time window; s and m are written into the calibration_version record bound to the nozzle calibration model, and the space reservation that has not yet been executed is invalidated after the calibration_version changes.

5. The method according to claim 1, characterized in that, For the overlap between the actual footprint F_a or the conservative actual footprint F_c and the spatial unit g_j, the overlap weight is calculated according to w_i,j=∫_(g_j)K_i(x)dx / ∫_(F_i)K_i(x)dx, where K_i(x) is the liquid mass surface density kernel given by the nozzle calibration model, F_i is F_a or F_c, and the sum of w_i,j of each intersecting spatial unit is 1.

6. The method according to claim 1, characterized in that, The controllable quality step size of the scalp pulse microdroplet injection device is Δq; when q_min≤q_allow<q_p, the reduced candidate liquid quality is determined according to q'_p=floor(q_allow / Δq)·Δq, where floor is a floor function, and the pass status is output only when q'_p≥q_min and each D_j^+ recalculated using q'_p is not greater than the corresponding U_j.

7. The method according to claim 1, characterized in that, The atomic creation space reservation includes comparing the current map_version in the same database transaction, checking that there is no conflict reservation in the intersecting space cell of F_pc, writing event_ID and reserving P_j and incrementing the map transaction version; if any comparison fails, P_j is not written, and the calculation described in claim 1 is re-executed based on the updated D_j, Q_j and P_j.

8. The method according to claim 1, characterized in that, The sampling period of the mass flow rate signal is no more than one-tenth of the pulse width; both the working distance signal and the nozzle pose signal have timestamps associated with the event_ID; the event is only allowed to be classified as a valid state when the mass flow rate, working distance, and nozzle pose signals all have complete samples within [t_0-τ_max,t_1+τ_max].

9. The method according to claim 8, characterized in that, The uncertain state of evidence also includes situations where the mass flow signal indicates q_i≥q_det while the valve status signal is missing, the working distance d(t) exceeds [d_min,d_max] by at least one sample, or the pose drift during execution is greater than δ_pose,max; after the isolation occupancy Q_j is written, the event prevents any subsequent candidate event that causes D_j+Q_j+P_j to exceed U_j.

10. The method according to claim 1, characterized in that, The event record is an append-only record, which includes at least event_ID, session identifier, map_version, coordinate transformation version, calibration_version, F_a, F_c, q_i or q_i^U, event status, evidence time range, and evidence hash; the commit transaction is executed with event_ID as the unique key, and the original transaction result is returned if the same event_ID already exists.

11. The method according to claim 10, characterized in that, During interrupt recovery, read the incomplete space reservation and sensor cache bound to event_ID: when the cache proves that the valve is always closed and q_i+ε_q<q_det, perform invalid state rollback; when the cache meets all the conditions of the valid state, commit idempotently according to the original event_ID; in other cases, write to Q_j according to the evidence uncertain state; when the root mean square error of the registration of the common reference flag before and after recovery is greater than e_recover,max, freeze the original map and its D_j, Q_j and P_j and establish a new session.

12. A scalp-area dose mapping and anti-repetition control system, characterized in that, The system includes a processor, a depth camera, a nozzle pose sensor, a mass flow sensor, a valve status sensor, a working distance sensor, a microdroplet pulse action mechanism, a dose ledger memory, a transaction log memory, and a microdroplet output enable gate. The processor is configured to: generate a session-level scalp surface coordinate map with spatial cells, coordinate transformation, registration root mean square error e_reg, and registration covariance matrix Σ_reg using the depth camera and nozzle pose sensor; and, before microdroplet pulse output, based on the nozzle calibration model... A nominal predicted footprint F_p is generated by adding Σ_reg to the pose, calibration, and scalp motion covariance, and forming a conservative predicted footprint F_pc according to r_p=s·√λ_max(Σ_p)+m and F_pc=F_p⊕B(r_p); the formal occupancy D_j, isolation occupancy Q_j, and reserved occupancy P_j in units of liquid mass are read from the dose ledger memory, according to D_j^+=D_j+Q_j+P_j+w_p,j·q_p and q_allow=m. The `in_{j:w_p,j>0}[(U_j-D_j-Q_j-P_j) / w_p,j]` method is used to determine whether candidate droplet pulses pass, have their quality reduced, or are blocked. Upon passing, the `event_ID` and `map_version` are atomically written to the reserved occupancy `P_j`. When the reservation, map version, and pose drift all pass verification, the droplet pulse action mechanism is driven by the droplet output enable gate. During execution, evidence bound to `event_ID` is obtained from the mass flow sensor, valve status sensor, working distance sensor, and nozzle pose sensor. A valid state is formed when the valve is open, the actual output liquid quality is not lower than `q_det`, the evidence sample is complete, the working distance is within the allowable range, and the pose drift is qualified. An invalid state is formed when the valve is always closed and the upper bound of the liquid quality measurement is lower than `q_det`. In other cases, an uncertain evidence state is formed, and idempotent updates to the formal occupancy `D_j`, rollbacks to the reserved occupancy `P_j`, or quality upper bound updates to the isolated occupancy `Q_j` are performed respectively.

13. The system according to claim 12, characterized in that, The field of view of the depth camera covers the nozzle's position and at least three non-collinear reference markers; the mass flow sensor is located upstream of the liquid outlet of the microdroplet pulse action mechanism; the valve status sensor detects the position of the liquid valve core; the measuring axis of the working distance sensor is parallel to the nozzle axis; and the nozzle pose sensor outputs a six-degree-of-freedom pose and pose covariance.

14. The system according to claim 12, characterized in that, The transaction log storage stores the IDLE, RESERVED, EXECUTING, COMMIT, ROLLBACK, and QUARANTINE states. The transition conditions from RESERVED to EXECUTING include no timeout, consistent map_version, and pose drift not exceeding δ_pose,max. The transition conditions from EXECUTING to COMMIT, ROLLBACK, or QUARANTINE correspond to valid, invalid, or uncertain evidence states, respectively.

15. The system according to claim 12, characterized in that, The droplet output enable gate has a processor permission input and a reserved valid latch input. The drive pulse is transmitted to the droplet pulse action mechanism only when both inputs are at a valid level. In the blocking state, when the reserved latch fails, when map_version changes, or when the processor is reset, at least one input is pulled to an invalid level.

Images