Dual Hormone Artificial Pancreas Algorithms 2026 — PatSnap Eureka
Dual Hormone Artificial Pancreas Control Algorithms
Dual-hormone artificial pancreas systems deliver both insulin and glucagon to automate bidirectional glucose regulation in type 1 diabetes. This landscape maps algorithm architectures, switching logic patents, and key assignees from 2010 to 2025.
Bidirectional Glucose Control: From MPC to Deep Reinforcement Learning
The dual-hormone artificial pancreas (DH-AP) extends the standard CGM-algorithm-pump framework by adding a second delivery channel for glucagon, enabling bidirectional glucose control. The core algorithmic challenge is mutual exclusivity: insulin and glucagon act antagonistically, so controllers must prevent simultaneous co-administration at any time step while maintaining tight glycemic regulation.
Model Predictive Control (MPC) is the dominant algorithmic family in retrieved records, applied in linear, nonlinear (NMPC), and extended-state-observer-augmented forms. Switching logic — hysteresis-based, threshold-based, or three-way decision — determines which hormone is delivered at each control interval, making it the central architectural differentiator across DH-AP patents and literature.
A 2022 meta-analysis synthesizing 17 randomized crossover trials (438 participants) found a mean difference in time-in-range (TIR) of +2.69% for DH-AP versus single-hormone AP — non-significant overall — but confirmed superior performance over sensor-augmented pumps. The narrowing TIR gap narrows the general-population case for DH systems while strengthening the rationale for hypoglycemia-prone subpopulations.
In this dataset, records span 2010 to 2025, with a foundational literature cluster (2011–2016), an engineering development phase (2017–2022), and a commercialization-oriented patent wave (2022–2025). Medtrum Technologies Inc. accounts for at least 15 distinct patent records in retrieved records, far exceeding any other single assignee. Chinese-origin entities collectively represent the largest share of patent filings in this dataset.
Control Architecture Distribution and Filing Activity Over Time
Retrieved records reveal four dominant algorithmic clusters — switched MPC with hysteresis, ESO-augmented MPC, hybrid cPID/cMPC, and reinforcement learning — with patent filing activity concentrated in two bursts: 2019–2022 (academic CN filings) and 2022–2025 (commercial WO/US/EP filings from Medtrum Technologies).
Algorithm Cluster Distribution — Dual-Hormone AP Patents and Literature (Dataset Snapshot)
Switched MPC with hysteresis logic and hybrid cPID/cMPC architectures represent the two most patent-active clusters in this dataset, with MPC-based approaches collectively covering the majority of retrieved records.
↗ Click bars to explorePatent Filing Activity by Phase — Dual-Hormone AP in Retrieved Records
Filing activity in this dataset clusters in three distinct phases: a foundational literature phase (2010–2016), an engineering development phase (2017–2022), and a commercialization patent wave (2022–2025) driven predominantly by Medtrum Technologies.
↗ Click bars to exploreKey Clinical and Research Applications of Dual-Hormone Closed-Loop Control
Retrieved records document DH-AP algorithm development and validation across four primary application contexts, from T1D home use to preclinical animal models, each presenting distinct algorithmic and regulatory requirements.
Type 1 Diabetes Home Use
All retrieved dual-hormone patents and the majority of literature records target T1D management. A 2022 meta-analysis of 17 randomized crossover trials (438 participants) found DH-AP achieved a mean TIR difference of +2.69% versus single-hormone AP. A 2016 study reported the first short-term home-use results of an integrated wearable bihormonal device.
Primary Clinical TargetPramlintide Dual-Hormone Variant
A 2022 systematic review of four crossover studies using pramlintide (amylin analog) paired with insulin in a DH-AP found all four studies demonstrated improved postprandial control. Pramlintide suppresses postprandial glucagon and slows gastric emptying, addressing a specific limitation of insulin-only systems in meal scenarios. This vertical has minimal patent coverage in retrieved records.
Postprandial ManagementPediatric Glucose Regulation
A 2020 study addressed pediatric closed-loop control, where glycemic variability is highest among T1D populations. Switched control with time-varying insulin-on-board (IOB) constraints was validated specifically for the pediatric population without requiring pre-meal insulin boluses. This application demands tighter safety constraints than adult systems due to heightened hypoglycemia risk.
Pediatric PopulationPreclinical Animal Model Validation
A 2022 study developed a low-order nonlinear animal model of glucose dynamics for bihormonal intraperitoneal insulin and glucagon delivery in animal subjects. This reduced-parameter model serves as a stepping stone toward implantable DH devices. Porcine and rodent models are documented as AP validation platforms in multiple retrieved records.
Preclinical ResearchLeading Assignees in Dual-Hormone Artificial Pancreas — Dataset Snapshot
In retrieved records, Medtrum Technologies Inc. accounts for at least 15 distinct patent filings across WO, EP, and US jurisdictions, far exceeding any other assignee in this dataset. Academic institutions including Harvard College and Beijing Institute of Technology hold smaller but technically significant portfolios in algorithm-level and embedded systems IP.
Top Assignees by Filing Count — Dual-Hormone AP Patents in Retrieved Records (Dataset Snapshot)
↗ Click bars to exploreMedtrum Technologies Inc.
Medtrum Technologies holds at least 15 distinct patent records in this dataset across WO (2023), EP (2024), and US (2024–2025 pending) jurisdictions, representing the single largest assignee by filing count in retrieved records. Their portfolio covers compound cPID/cMPC architectures, risk-space transformation algorithms (rMPC, rPID), multi-drug dual-hormone infusion systems, and modular fault-tolerant control units with priority-based automatic switching. WO filings cluster around May 2023, indicating a coordinated international filing round; US filings remain pending as of 2025.
China / International (WO, EP, US)President and Fellows of Harvard College
Harvard College holds three active patents in this dataset spanning WO (2021, 2022) and US (2022, 2024) jurisdictions. Key filings include event-triggered MPC for embedded artificial pancreas systems — which reduces energy consumption by triggering MPC updates only on glucose events rather than fixed intervals — and iterative learning control with sparse measurements for insulin injection in T1D. Both the 2022 WO and 2024 US filings for event-triggered MPC are listed as active.
United StatesNext-Generation Approaches in Dual-Hormone AP Algorithm Design
Records dated 2021–2025 in this dataset identify five forward-looking technology directions, ranging from risk-space algorithm reformulation to deep reinforcement learning for bihormonal personalization.
Risk-Space Transformation for Symmetric Glycemic Optimization
Multiple 2023–2025 Medtrum patents across WO, EP, and US describe converting asymmetric blood glucose from physical measurement space to a symmetric blood glucose risk space before applying MPC or PID optimization. This architectural innovation directly addresses the fundamental asymmetry of glycemic control — where hypoglycemia is acutely more dangerous than mild hyperglycemia — by reframing it as a symmetric optimization problem. The approach is filed under both rMPC and rPID algorithm variants.
Event-Triggered MPC for Embedded Low-Power Wearables
Harvard’s 2024 US patent (active) and its 2022 WO predecessor introduce event-triggered MPC that updates only when glucose events occur, rather than at fixed computational intervals. This energy-aware approach is identified as critical for embedded low-power wearable deployment as DH-AP moves toward fully implantable form factors. The computational trigger reduction addresses a key hardware constraint for next-generation devices.
Switched MPC with Hysteresis vs. Hybrid cPID/cMPC: Key Dimensions
Click any row to explore further.
| Dimension | Switched MPC / Hysteresis | Hybrid cPID/cMPC |
|---|---|---|
| Primary Source | Academic literature (2018, 2022) | Commercial patents — Medtrum Technologies (2023–2025) |
| Hormone Selection Logic | Hysteresis or threshold switching between insulin and glucagon MPC sub-controllers | Priority-based and failure-state switching among cPID, cMPC, and compound algorithm modules |
| Controller Coupling | Two independent MPC sub-controllers; mutually exclusive activation | Bidirectional coupling: cPID input is intermediate output of cMPC and vice versa; iterative convergence to I3 |
| Key Performance Result | 89.3% TIR, zero hypoglycemic events in 50 virtual patients (NMPC, 2022) | Risk-space symmetric optimization across hypo- and hyperglycemia; commercial regulatory readiness |
| Disturbance Handling | ESO variant estimates total disturbances of simplified glucose metabolic model in real time | Redundant detection modules; fault-tolerant multi-module architecture with automatic switching under failure |
| Jurisdictional Coverage | Academic publications; CN patents (Beijing Institute of Technology, 2022, active) | WO (2023), EP (2024), US (2024–2025 pending) — Medtrum Technologies |
| Maturity Stage | Research and in silico validation; some clinical pilot studies | Commercial-scale systems engineering; pre-commercial regulatory readiness features present |
| Second Hormone Handling | Explicit glucagon sub-controller; switched based on glucose zone or hysteresis boundary | Multi-drug infusion system supporting dual-hormone delivery; modular pump channels |
Frequently Asked Questions: Dual-Hormone Artificial Pancreas Algorithms
The core challenge is bidirectional control: the algorithm must deliver insulin to prevent hyperglycemia and deploy glucagon to rescue from hypoglycemia, while ensuring mutual exclusivity — insulin and glucagon are antagonistic, so both must not be co-administered at the same time step. Switching logic (hysteresis, threshold-based, or three-way decision) is the central architectural differentiator.
The 2022 meta-analysis synthesized 17 randomized crossover trials (438 participants) comparing DH-AP to single-hormone AP, CSII, and PLGS systems. It found a mean difference in time-in-range (TIR) of +2.69% for DH-AP versus single-hormone AP, which was non-significant. DH-AP did demonstrate superior performance over sensor-augmented pumps.
In Medtrum’s compound artificial pancreas algorithm, the cPID and cMPC are bidirectionally coupled: the input to the cPID is the intermediate output of the cMPC, and vice versa. This iterative process converges to an optimized infusion dose (I3) when the outputs of both algorithms equalize (I1 = I2). The system also converts blood glucose from physical measurement space to a risk space for symmetric optimization.
The Extended State Observer (ESO) is embedded alongside MPC to estimate unknown disturbances in the glucose-insulin-glucagon metabolic model in real time. It corrects for model-plant mismatch caused by inter- and intra-patient variability, enabling more robust switching decisions between the insulin and glucagon MPC sub-controllers. Beijing Institute of Technology holds two active CN patents (2022) implementing this architecture.
Pramlintide is a synthetic amylin analog already in clinical use. A 2022 systematic review of four crossover studies found that insulin+pramlintide DH-AP improved postprandial control in all four studies, as pramlintide suppresses postprandial glucagon and slows gastric emptying. Unlike glucagon, pramlintide requires no special formulation or separate pump configuration, and the combination has minimal patent coverage in retrieved records.
Retrieved patent records span WO (predominantly Medtrum Technologies, 2023), US (Harvard College active; Medtrum 2024–2025 pending), CN (Beijing Institute of Technology active 2022; Yanshan University inactive 2019), EP (Medtrum 2024 pending), and IN (Dr. Anchana P Belmon inactive 2022). Chinese-origin assignees collectively account for the largest share of patent records in this dataset.
Data and insights on this page are based on a limited patent and literature dataset and are for reference only. Figures may not represent the complete technology landscape.