HEALWELL's DARWEN(TM) AI Outperforms Published Foundation Model Benchmarks and Demonstrates Real-World Patient Impact in New Peer-Reviewed Research
DARWEN AI shows higher clinical summarization and treatment-plan accuracy than benchmark general models and supports better heart failure therapy optimization.
Rhea-AI Summary
HEALWELL AI (HWAIF) reported new peer-reviewed research showing its DARWEN™ AI platform outperforming published benchmarks for leading general-purpose AI models and delivering real-world patient impact in 2026 studies.
At AMIA 2026, DARWEN™-powered SMARTSuite clinical summaries exceeded OpenAI GPT, DeepSeek and Llama benchmarks with average improvements of 43% in ROUGE-1, 21% in ROUGE-2 and 43% in ROUGE-L. Separate AI-MEDX 2026 validation showed DARWEN™ treatment-plan accuracy of 54.7% versus 37.3% for Meta-Llama-3-70B-Instruct in blinded clinician review. A Canadian Journal of Cardiology study of SMART Patient ID identified over 4,100 undertreated heart failure patients and helped drive a 70% improvement in treatment optimization within six months.
Research using DARWEN™ has been accepted across 12 fall 2026 forums and includes collaborations with AstraZeneca, Bayer, Galderma, UCB, Takeda, Arcutis and Merck.
Positive
- ROUGE-1 59.8 vs GPT 41.8, Llama 43.5, DeepSeek 39.9
- ROUGE-2 37.0 vs GPT 34.3, Llama 28.5, DeepSeek 29.8
- ROUGE-L 48.8 vs GPT 36.8, Llama 29.8, DeepSeek 36.6
- Treatment-plan accuracy 54.7% vs Meta-Llama-3-70B-Instruct 37.3%
- SMART Patient ID found 4,100+ undertreated heart failure patients
- Heart failure treatment optimization improved by 70% within six months
- DARWEN research accepted at 12 global forums in fall 2026
- Studies include collaborations with AstraZeneca, Bayer, Galderma, UCB, Takeda, Arcutis and Merck
Negative
- None.
AI-generated analysis. How Rhea-AI works. Not financial advice.
New research accepted for presentation at AMIA 2026 demonstrates the performance of HEALWELL's DARWEN™ AI-powered SMARTSuite, including clinical summarization results that outperformed published benchmarks for OpenAI's GPT, DeepSeek and Llama across all three measures used to evaluate clinical summarization accuracy.
Separate DARWEN™ AI validation research accepted for oral presentation at AI-MEDX 2026 demonstrated nearly
50% higher treatment-plan accuracy than a leading off-the-shelf general-purpose AI model in blinded clinician review, producing more clinically specific outputs, including drugs and doses.New DARWEN™ AI research in the Canadian Journal of Cardiology demonstrates the impact of SMART Patient ID, identifying more than 4,100 heart failure patients not receiving guideline-directed medical therapy, helping drive a
70% improvement in treatment optimization within six months.HEALWELL and its pharmaceutical company partners have secured multiple fall 2026 research acceptances at leading scientific forums, including EADV 2026, ACG 2026, The Menopause Society 2026 and ISPOR Europe 2026, spanning women's health, dermatology, gastroenterology and oncology.
Toronto, Ontario--(Newsfile Corp. - September 15, 2026) - HEALWELL AI Inc. (TSX: AIDX) (OTCQX: HWAIF) ("HEALWELL" or the "Company"), a healthcare artificial intelligence company focused on preventative care, is pleased to announce new research supporting the clinical performance and real-world impact of its DARWEN™ AI platform, with studies accepted for presentation or publication across 12 medical, health informatics, artificial intelligence and health economics forums during the fall of 2026. The research demonstrates DARWEN™'s ability to translate healthcare-specific AI into clinically relevant outcomes, including performance advantages over leading general-purpose AI models in clinical summarization and treatment-plan generation, as well as measurable improvements in the identification and treatment optimization of patients with unmet clinical needs.
DARWEN™ AI is HEALWELL's proprietary, patented and clinically validated AI platform that transforms complex structured and unstructured healthcare data into traceable, actionable knowledge. DARWEN™ powers SMART Search, SMART Summary and SMART Patient ID, helping clinicians find critical information, understand complex patient histories and identify patients with unmet clinical needs earlier.
New research being presented at AMIA 2026 demonstrates the performance of HEALWELL's DARWEN™ AI-powered SMARTSuite, including clinical summarization results that outperformed published benchmarks for OpenAI's GPT, DeepSeek and Llama across all three measures used to evaluate clinical summarization accuracy. The three measures, ROUGE-1, ROUGE-2 and ROUGE-L, are used to evaluate the quality and accuracy of generated clinical summaries. DARWEN™ demonstrated an average
Separate research accepted for oral presentation at AI-MEDX 2026 demonstrated nearly
New research published in the Canadian Journal of Cardiology demonstrates the real-world impact of HEALWELL's DARWEN™ AI-powered SMART Patient ID in identifying patients with unmet clinical needs. The study identified more than 4,100 patients with heart failure who were not receiving guideline-directed medical therapy, enabling targeted clinical intervention and helping drive a
Aaron Leibtag, Head of Global Health Systems AI Strategy at HEALWELL and Co-Founder of Pentavere, said, "DARWEN™'s continued clinical and scientific validation comes as the platform gains commercial traction across health systems and life sciences. Outperforming published benchmarks for leading foundation models while demonstrating real-world patient impact validates our strategy. HEALWELL recently announced a multi-year deployment of DARWEN™-powered SMART Search and SMART Summary across Iowa's statewide Health Information Exchange, which receives data from 115 hospitals, as well as DARWEN™-powered clinical trial matching through WELLTRUST™'s network of more than 100,000 consented patients. Together, these deployments demonstrate HEALWELL's strategy of translating DARWEN™'s scientific differentiation into commercial scale."
Dr. Christopher Pettengell, Chief Medical Officer of HEALWELL, commented, "The important question for clinical AI is not whether a model can produce an answer that sounds medically plausible. It is whether it can consistently identify the right information, preserve the clinical context and provide the specificity required for clinicians to act. Outperforming published benchmarks for leading foundation models while demonstrating real-world patient impact validates our strategy, build the evidence first, then scale what works across health systems and life sciences."
Dr. Pettengell added, "Our upcoming peer-reviewed presentations and publications provide further evidence that purpose-built, clinically validated AI can meaningfully outperform general-purpose approaches on these dimensions. Combined with the breadth of therapeutic-area research being presented this fall, we believe this reinforces DARWEN™ as an increasingly differentiated clinical intelligence AI platform."
Fall 2026 Program: Broad Scientific Validation Across Healthcare and Life Sciences
HEALWELL's fall 2026 clinical and AI scientific program includes research accepted at 12 global programs, including AMIA 2026, AI-MEDX 2026, EADV 2026, ACG 2026, The Menopause Society 2026, ISPOR Europe 2026 and the Canadian Cardiovascular Congress.
These DARWEN™ AI-powered real-world evidence & AI demonstration studies span women's health, dermatology, inflammatory bowel disease, oncology, heart failure and rare disease, demonstrating the platform's breadth across patient identification, clinical decision support, research and evidence generation. The studies include collaborations with leading global pharmaceutical companies including AstraZeneca, Bayer, Galderma, UCB, Takeda, Arcutis and Merck.
Together, the fall scientific program and recent commercial deployments demonstrate HEALWELL's progression from scientific validation to enterprise deployment and commercial scale, while expanding DARWEN™ across health systems and life sciences. HEALWELL expects to provide additional details on individual studies and results as they are presented and published.
About HEALWELL AI
HEALWELL is a healthcare artificial intelligence company focused on preventative care. Its mission is to improve healthcare and save lives through early identification and detection of disease. Using its own proprietary technology, the Company is developing and commercializing advanced clinical decision support systems that can help healthcare providers detect rare and chronic diseases, improve efficiency of their practice and ultimately help improve patient health outcomes. HEALWELL is executing a strategy centered around developing and acquiring technology and clinical sciences capabilities that complement the Company's road map. HEALWELL is publicly traded on the Toronto Stock Exchange under the symbol "AIDX" and on the OTC Exchange under the symbol "HWAIF". To learn more about HEALWELL, please visit https://healwell.ai/.
Forward-Looking Statements
Certain statements in this press release constitute "forward-looking information" and "forward-looking statements" (collectively, "forward-looking statements") within the meaning of applicable Canadian securities laws, including statements regarding the anticipated presentation and publication of research findings at AMIA 2026, AI-MEDX 2026, EADV 2026, ACG 2026, The Menopause Society 2026, ISPOR Europe 2026 and the Canadian Cardiovascular Congress; the Company's belief that DARWEN™ is an increasingly differentiated clinical intelligence AI platform; the Company's expectation to provide additional details on individual studies and results as they are presented and published; and the Company's strategy of translating scientific differentiation into commercial scale. Forward-looking statements are often, but not always, identified by words or phrases such as "expects", "believe", "anticipated", "intends", "plans", "will", "strategy" or variations of such words and phrases, or statements that certain future conditions, actions, events or results "will", "may", "could", "would", "should" or "can" be taken, occur or be achieved, or the negative of any of these terms. Forward-looking statements are necessarily based upon management's perceptions of historical trends, current conditions and expected future developments, as well as a number of specific factors and assumptions that, while considered reasonable by HEALWELL as of the date of such statements, are outside of HEALWELL's control and are inherently subject to significant business, economic and competitive uncertainties and contingencies which could result in the forward-looking statements ultimately being entirely or partially incorrect or untrue. Forward-looking statements contained in this press release are based on various assumptions, including, but not limited to, the following: (i) changes to the schedules and agendas for upcoming clinical and AI scientific programs; (ii) HEALWELL's ability to maintain and leverage its relationships with its commercial partners; (iii) the continued adoption of the software, tools and solutions created by HEALWELL; (iv) the stability of general economic and market conditions; (v) sufficiency of working capital and access to financing to support HEALWELL's existing operations and strategic plan; (vi) HEALWELL's ability to comply with applicable laws and regulations; (vii) HEALWELL's continued compliance with third party intellectual property rights; (viii) the effects of competition in the industry; (ix) the requirement for increasingly innovative product solutions and service offerings and HEALWELL's ability to continue to develop and commercialize them; (x) technologies working as intended or at all; (xi) trends in customer growth and the adoption of new technologies in the industry; and (xii) that the risk factors noted below, collectively, do not have a material impact on HEALWELL's business, operations, revenues and/or results. By their nature, forward-looking statements are subject to inherent risks and uncertainties that may be general or specific and which give rise to the possibility that expectations, forecasts, predictions, projections, or conclusions will not prove to be accurate, that assumptions may not be correct, and that objectives, strategic goals and priorities will not be achieved.
Known and unknown risk factors, many of which are beyond the control of HEALWELL, could cause the actual results of HEALWELL to differ materially from the results, performance, achievements, or developments expressed or implied by such forward-looking statements. Such risk factors include but are not limited to those factors which are discussed under the section entitled "Risk Factors" in HEALWELL's most recent annual information form dated March 18, 2026, which is available under HEALWELL's SEDAR+ profile at www.sedarplus.ca. There can be no assurance that forward-looking statements will prove to be accurate, as actual results and future events could differ materially from those anticipated in such statements. HEALWELL disclaims any intention or obligation to update or revise any forward-looking statements whether as a result of new information, future events or otherwise, except to the extent required by applicable law. All of the forward-looking statements contained in this press release are qualified by these cautionary statements.
For more information:
Pardeep S. Sangha
Investor Relations, HEALWELL AI Inc.
Phone: 604-572-6392
ir@healwell.ai
Footnotes:
(1) DARWEN™ achieved a ROUGE-1 score of 59.8, compared with published benchmarks of 41.8 for OpenAI's GPT, 43.5 for Llama and 39.9 for DeepSeek; DARWEN™ achieved a ROUGE-2 score of 37.0, compared with 34.3 for GPT, 28.5 for Llama and 29.8 for DeepSeek; and DARWEN™ achieved a ROUGE-L score of 48.8, compared with 36.8 for GPT, 29.8 for Llama and 36.6 for DeepSeek.
(2) DARWEN™ achieved an accuracy score of
(3) O. Sergerie, S. Goodman, S. Zieroth, R. Choi, A. Gupta, Y. Khaykin, A.K. Hhameed, K. Leblanc, M. McDonald, C. Pettengell, R. Roy, M. Tan, A.M. Yip, J.E. Ezekowitz, A. Ducharme, A Natural Language Processing-Driven Electronic Medical Record Program to Optimize Medical Therapy for Heart Failure in Community Practices, CJC Open (2026), doi: https://doi.org/10.1016/j.cjco.2026.07.003.
To view the source version of this press release, please visit https://www.newsfilecorp.com/release/314364
FAQ
AI-generated questions and answers. How Rhea-AI works. Not financial advice.
What is DARWEN AI and which products does it power?
DARWEN™ AI is HEALWELL's proprietary, patented and clinically validated healthcare AI platform that converts complex structured and unstructured data into traceable, actionable knowledge. It powers SMART Search, SMART Summary and SMART Patient ID, which are designed to help clinicians find critical information, understand complex patient histories and identify patients with unmet clinical needs earlier.
Which clinical areas are covered by HEALWELL's fall 2026 DARWEN research program?
The fall 2026 program spans women's health, dermatology, inflammatory bowel disease, oncology, heart failure and rare disease. These DARWEN™ AI-powered studies focus on patient identification, clinical decision support, research and evidence generation across healthcare and life sciences.
How does HEALWELL link DARWEN's scientific results to commercial deployment?
HEALWELL cites a multi-year deployment of DARWEN™-powered SMART Search and SMART Summary across Iowa's statewide Health Information Exchange, which receives data from 115 hospitals, and DARWEN™-powered clinical trial matching via WELLTRUST™'s network of more than 100,000 consented patients, as examples of translating DARWEN™'s scientific differentiation into enterprise deployment and commercial scale.