Update README.md

README.md

@@ -7,147 +7,36 @@ sdk: static
 pinned: false
 ---
 
-# AI-POCUS Evidence Infrastructure Hub for LMICs
+# AI-POCUS Evidence Hub
 
-The **AI-POCUS Evidence Hub** is an LMIC-anchored, globally connected infrastructure initiative advancing the responsible development, validation, synthesis, and policy translation of **AI-enabled point-of-care ultrasound** for low-resource and humanitarian settings.
+The **AI-POCUS Evidence Hub** is an LMIC-anchored infrastructure initiative for the responsible development, validation, synthesis, and translation of **AI-enabled point-of-care ultrasound** in low-resource and humanitarian health systems.
 
-Point-of-care ultrasound can bring real-time, radiation-free imaging to the bedside, primary care clinic, mobile outreach programme, and crisis-affected field setting. When combined with artificial intelligence, POCUS has the potential to support non-specialist health workers with image acquisition, scan quality assessment, interpretation, triage, referral, treatment guidance, and monitoring across high-burden clinical areas.
+The Hub supports a multi-country Community of Practice working to make AI-POCUS evidence more standardised, interoperable, reusable, and policy-ready.
 
-Yet the AI-POCUS evidence base remains fragmented. Studies are difficult to compare across devices, scanning protocols, patient populations, reference standards, validation methods, metadata structures, and implementation contexts. The field lacks the shared evidence infrastructure needed to determine which tools are safe, effective, generalisable, equitable, and implementation-ready for LMIC and humanitarian health systems.
+## Scope
 
-The AI-POCUS Evidence Hub addresses this gap by building the common standards, registries, living evidence systems, federated data resources, and collaborative infrastructure needed for AI-POCUS evidence to accumulate into policy-ready knowledge.
+AI-POCUS can extend diagnostic capacity beyond hospitals and specialist services by combining portable ultrasound with automated acquisition support, scan quality assessment, image interpretation, and clinical decision guidance.
 
----
-
-## Mission
-
-Our mission is to make AI-POCUS evidence **standardised, interoperable, reusable, and decision-ready** for the settings where diagnostic access gaps are greatest.
-
-We aim to support equitable AI-POCUS adoption by helping researchers, implementers, ministries of health, global health agencies, humanitarian actors, and clinical partners generate and use evidence that is transparent, comparable, locally relevant, and aligned with real-world decision needs.
-
----
-
-## Why AI-POCUS?
-
-In low-resource and humanitarian settings, limited access to diagnostics contributes to delayed and missed diagnosis across infectious disease, maternal health, emergency care, chronic disease, trauma, and acute care. Diagnostic systems built around fixed imaging facilities, specialist personnel, and equipment-intensive workflows often fail to reach patients where they first present.
-
-POCUS offers a practical alternative. It is portable, reusable, radiation-free, and increasingly deployable through handheld probes connected to tablets or smartphones. It can support frontline decisions across lung, cardiac, abdominal, pelvic, obstetric, emergency, and primary care applications.
-
-AI-enabled POCUS addresses one of the major barriers to scale: operator dependence. AI systems can support acquisition guidance, anatomical landmark detection, pathology classification, quality control, decision support, and workflow integration.
-
-However, responsible adoption requires more than promising models. It requires shared infrastructure for evidence generation, validation, synthesis, governance, and implementation.
+This organization hosts shared resources for AI-POCUS research, evidence synthesis, validation, and implementation across priority use cases including infectious disease, maternal health, emergency care, and primary care.
 
----
-
-## What We Are Building
-
-The Hub is structured around five integrated evidence infrastructure components.
+## What We Host
 
-### 1. Shared AI-POCUS Taxonomy and Standards
-
-We are developing a common taxonomy and reporting framework for AI-POCUS research, validation, and implementation. This includes standard definitions for:
-
-- Clinical use cases
-- Scanning protocols
-- Devices and probes
-- AI model classes
-- Annotation schemas
-- Metadata structures
-- Validation designs
-- Reference standards
-- Equity dimensions
-- Implementation settings
-- Dataset, model, and protocol cards
+This Hugging Face organization will host and maintain:
 
-This taxonomy acts as the definitional layer that makes future evidence easier to compare, synthesise, and reuse.
-
-### 2. Living Evidence Maps and AI-Enabled Living Evidence Synthesis
-
-The Hub will maintain continuously updated evidence maps across priority AI-POCUS use cases, including tuberculosis triage, paediatric pneumonia, maternal health, emergency care, and multi-disease deployment.
-
-These evidence maps will feed into **AI-enabled Living Evidence Synthesis**, allowing new studies, validation outputs, and implementation signals to be incorporated as the field evolves.
-
-This approach is designed to move beyond static reviews toward continuously updated evidence products aligned with policy, procurement, guideline, and implementation timelines.
-
-### 3. Federated Data, Model, and Protocol Registry
-
-The Hub will support an open, federated registry of AI-POCUS datasets, validated models, and research protocols. The registry will be aligned with FAIR principles, DICOM-compatible data structures, STARD-AI, TRIPOD-AI, and responsible AI reporting practices.
-
-The registry will include:
-
-- Open ultrasound dataset cards
-- Dataset licensing and access information
-- Model cards
-- Protocol registrations
-- Validation plans
-- Metadata templates
-- Persistent identifiers
-- Federated benchmarking pathways
-
-Where appropriate, the Hub will support federated validation and model evaluation without requiring raw patient-level data to leave originating institutions.
-
-### 4. Dynamic Policy Question Bank
-
-The Hub will maintain a structured, living repository of priority policy questions co-produced with LMIC policymakers, ministries of health, implementation partners, and global health stakeholders.
-
-Questions will span:
-
-- Diagnostic accuracy
-- Clinical utility
-- Safety and interpretability
-- Equity of access
-- Health system integration
-- Implementation feasibility
-- Training and workflow requirements
-- Cost-effectiveness
-- Budget impact
-- Procurement readiness
-- Humanitarian deployment suitability
-
-The goal is to ensure that evidence synthesis outputs are linked to real decision windows, including national strategies, donor investments, guideline cycles, and procurement processes.
-
-### 5. Capacity Sharing and Ecosystem Strengthening
-
-The Hub is built around an existing multi-country Community of Practice, bringing together researchers, clinicians, implementers, policymakers, data scientists, and evidence synthesis experts.
-
-We support:
-
-- LMIC-led working groups
-- Open science toolkits
-- Training resources
-- Protocol templates
-- Evidence synthesis methods
-- Dataset and model documentation
-- Responsible AI validation practices
-- Peer exchange across institutions
-- Implementation guidance for frontline settings
-
-The Hub prioritises LMIC leadership, equitable governance, and practical capacity building across the AI-POCUS evidence lifecycle.
-
----
-
-## Core Outputs
-
-The AI-POCUS Evidence Hub will release and maintain:
-
-- A standardised AI-POCUS taxonomy
-- AI-POCUS reporting and metadata templates
-- Living evidence maps
-- AI-enabled living evidence syntheses
 - Open ultrasound dataset registries
 - Dataset cards and model cards
-- Federated protocol and validation registries
+- AI-POCUS taxonomies and metadata standards
+- Protocol templates and reporting frameworks
+- Validation and benchmarking resources
+- Federated evaluation tools
+- Living evidence maps
+- AI-enabled living evidence synthesis outputs
 - Policy question banks
-- Evidence-to-decision briefs
-- Implementation guidance
-- Training modules
-- Open-source tools and reusable infrastructure
-
----
+- Implementation and training resources
 
 ## Priority Use Cases
 
-The Hub initially focuses on AI-POCUS applications with high relevance for LMIC and humanitarian settings, including:
+Initial priority areas include:
 
 - Tuberculosis triage and referral
 - Paediatric pneumonia diagnosis
@@ -156,130 +45,66 @@ The Hub initially focuses on AI-POCUS applications with high relevance for LMIC
 - Lung ultrasound for infectious disease
 - Multi-disease frontline diagnostic workflows
 - AI-supported acquisition and scan quality assessment
-- Task-shifted imaging by non-specialist health workers
-
-Additional use cases will be incorporated as the Community of Practice expands.
-
----
-
-## Governance and Community
 
-The Hub is governed through an LMIC-majority, rotating Steering Committee and thematic working groups. Governance is designed to ensure that priorities are shaped by the institutions and health systems generating the evidence and serving the populations most affected by diagnostic gaps.
+## Evidence Infrastructure
 
-Working groups include:
+The Hub is designed to address fragmentation in the AI-POCUS evidence base by supporting shared standards for:
 
-- AI Validation and Safety
-- Clinical Workflows and Training
-- Implementation and Scale-Up
-- Ethics, Equity, and Data Governance
-- Evidence Synthesis and Policy Translation
-- Data Standards and Interoperability
-
-The Hub operates through principles of collective impact, responsible AI, open science, FAIR data, equitable partnership, and policy-oriented evidence generation.
+- Clinical use cases
+- Devices and probes
+- Scanning protocols
+- Annotation schemas
+- Reference standards
+- Metadata structures
+- Model validation
+- External benchmarking
+- Equity analysis
+- Implementation context
+- Data and model documentation
 
----
+These resources are intended to make AI-POCUS studies easier to compare, validate, synthesize, and translate into policy and implementation guidance.
 
-## Data and Model Sharing
+## Data and Model Access
 
 The Hub supports responsible, tiered access to AI-POCUS research outputs.
 
-Open resources may include:
-
-- De-identified metadata
-- Dataset documentation
-- Model cards
-- Protocols
-- Evidence maps
-- Synthesis datasets
-- Code and analysis pipelines
-- Training materials
-
-Restricted resources may include:
+Open resources may include dataset documentation, metadata templates, protocols, evidence maps, model cards, code, and training materials.
 
-- Patient-level ultrasound imaging
-- Sensitive clinical metadata
-- Site-specific validation datasets
+Access to patient-level ultrasound imaging or sensitive clinical metadata may be restricted and governed by local ethics approvals, data use agreements, institutional review, and partner requirements. Where possible, the Hub supports federated validation and benchmarking without requiring raw clinical data to leave originating institutions.
 
-Access to restricted data will depend on local ethics approvals, data use agreements, institutional sign-off, and governance requirements from contributing partners.
+## Community and Governance
 
-Where possible, the Hub promotes federated approaches that allow validation and benchmarking without transferring raw clinical data.
+The Hub is connected to a multi-country AI-POCUS Community of Practice involving researchers, clinicians, implementers, data scientists, policymakers, evidence synthesis experts, and global health partners.
 
----
-
-## Responsible AI Commitments
-
-The AI-POCUS Evidence Hub is committed to responsible, transparent, and context-aware AI development.
+Working areas include:
 
-Our work emphasises:
-
-- External validation across sites and populations
-- Equity analysis across sex, age, geography, pregnancy status, displacement status, disease severity, and other relevant dimensions
-- Transparent reporting using STARD-AI and TRIPOD-AI
-- Clear model and dataset documentation
-- Human-centred implementation
-- Interpretable outputs for frontline users
-- Governance for data protection and consent
-- Avoidance of extractive data practices
-- LMIC-led priority setting and authorship
-- Alignment with local regulatory and ethical frameworks
-
----
+- AI validation and safety
+- Clinical workflows and training
+- Implementation and scale-up
+- Ethics, equity, and data governance
+- Evidence synthesis and policy translation
+- Data standards and interoperability
 
-## Relationship to Global Health Priorities
+The Hub prioritises LMIC leadership, equitable governance, open science, responsible AI, FAIR data principles, and policy-oriented evidence generation.
 
-The Hub contributes to **SDG 3** by strengthening diagnostic capacity for conditions that drive preventable mortality in low-resource and crisis-affected settings.
-
-It also contributes to **SDG 16** by supporting evidence-informed institutions, transparent governance, and resilient decision-making infrastructure in fragile and resource-constrained health systems.
-
-The Hub aligns with global priorities for:
-
-- Universal health coverage
-- Digital health equity
-- Responsible AI
-- Diagnostic access
-- Humanitarian health response
-- TB elimination
-- Maternal and child health
-- Evidence-informed policymaking
-- Open science and FAIR data infrastructure
-
----
+## Reuse and Citation
 
-## Current Development Areas
+Each repository will include its own documentation, citation guidance, license terms, and access conditions.
 
-This organization space will host resources related to:
-
-- AI-POCUS dataset registries
-- Open ultrasound dataset cards
-- Model cards
-- Protocol templates
-- Evidence maps
-- Living evidence synthesis outputs
-- Metadata standards
-- Validation frameworks
-- Implementation toolkits
-- Training resources
-- Community of Practice outputs
-
-Repositories will be added as Hub infrastructure is developed and released.
-
----
+If you use Hub resources, please cite the relevant dataset, model, protocol, evidence map, or synthesis product using the citation information provided in the corresponding repository.
 
 ## Contact
 
-For collaboration, governance, dataset contribution, or Community of Practice enquiries:
+For collaboration, dataset contribution, governance, or Community of Practice enquiries:
 
+ 
 EPFL LiGHT Laboratory  
 Email: mary-anne.hartley@epfl.ch
 
----
-
-## Citation
-
-If you use Hub resources, please cite the relevant dataset, model, protocol, or evidence product using the citation information provided in each repository.
-
----
+Website: [AI-POCUS Community of Practice](https://www.ai-pocus-lmics.org/)
 
 ## License
 
-Unless otherwise specified, Hub code and documentation will be released under permissive open-source licenses such as MIT or Apache 2.0. Dataset access and reuse terms will be specified individually according to consent, ethics approvals, data use agreements, and partner governance requirements.
+Unless otherwise specified, code and documentation will be released under permissive open-source licenses such as MIT or Apache 2.0.
+
+Dataset access and reuse terms will be specified individually according to consent, ethics approvals, data use agreements, and partner governance requirements.