Merge pull request #420 from alan-turing-institute/add-collaborators

Add Case Studies section + Cemrg case study to documentation

site/docs/assets/style.css

@@ -16,3 +16,7 @@ li.md-tabs__item {
 article.md-content__inner {
   padding-bottom: 12rem;
 }
+.youtube-video {
+  aspect-ratio: 16 / 9;
+  width: 100%;
+}

site/docs/community/case-studies/cemrg-app-transcript.md

@@ -0,0 +1,96 @@
+# Transcript from video
+
+Hello and welcome to our presentation showcasing the work of the Cardiac Electromechanics research Group CEMRG at Imperial College London as a case study for the TEA platform.
+
+Today we are excited to delve into one of our flagship projects, the CemrgApp, which is revolutionising the way we analyse cardiovascular data.
+
+Jose and I are members of CEMRG at Imperial College.
+
+Our research group applies statistical, machine learning and simulation approaches to study the physiology, pathophysiology, diagnosis, and treatment of the heart.
+
+We are an inherently interdisciplinary research group, bringing together mathematicians, engineers, statisticians, experimental researchers and clinicians.
+
+The combination of technical expertise and clinical insight makes CEMRG uniquely positioned to drive translational research and make a real impact in the field of cardiovascular medicine.
+
+Now we turn our attention to the innovative tool that is at the forefront of our efforts, the scar quantification tool or SQT.
+
+Within the CemrgApp platform, expect to learn the importance of scar identification and how it is performed, the role of the SQT in analysing scar tissue in the heart and how it works.
+
+The fairness assurance analysis made for the SQT and how fairness claims were chosen and backed.
+
+Scar tissue in the heart can lead to serious complications such as such as abnormal heart rhythms, making accurate identification paramount for effective diagnosis and treatment.
+
+Identifying scar tissue accurately can help clinicians make informed decisions and improve patient outcomes.
+
+The only non invasive tool to assess scar tissue is through an advanced 3D imaging technology known as Late Gadolinium Enhanced Cardiac Magnetic Resonance or LGE-CMR to identify scar tissue in the heart in which scar tissue appears brighter than normal healthy tissue.
+
+Manually assessing these scans can vary depending on the operator.
+
+The main focus of this case study is the Scar Quantification tool, or SQT, within the similar job platform.
+
+This tool is a crucial component of our efforts to analyse the scar tissue in the heart.
+
+It enables us to visualise and validate the steps required for accurate quantification of scar tissue.
+
+Now let's delve into how the SQT works.
+
+Taking an LGE-CMR scan as input, the tool goes through three main phases.
+
+One, segmentation to isolate the left atrium within the heart, 2 generating a 3D representation of its geometry, and then three, creating a scar map based on the signal intensity from the LGE scan.
+
+The user design presents several push buttons, sequentially numbered to present steps in the workflow, which are then visualised within the same user interface.
+
+In the video, we can see the later stages of the process where a 3D model is created with different colors to indicate the amount of scar tissue per region.
+
+Fairness assurance is crucial for ensuring that the SQT operates ethically and accurately.
+
+In this case study, we have considered factors such as bias, mitigation, diversity, and inclusivity.
+
+CemrgApp in general is tailored to patient specific workflows where the physiological data corresponds to an individual.
+
+While this brings many opportunities for personalised healthcare, it also carries risks, most notably the possible risk of unequal performance for individuals or subgroups of the population.
+
+Therefore, it is worth considering how such risk can be identified, evaluated and mitigated.
+
+Alongside the Turing team, we approach the development of an insurance case for the SQT using the following general approach.
+
+First we explored a general understanding of a concept of fairness.
+
+Then we consider different examples of practical fairness issues that are present and use court attributes of fairness as strategies for identifying exemplary claims and evidence for a draft assurance case.
+
+This became an introspection exercise of our software which provided us with a snapshot of our project with regards to fairness, its strengths, shortcomings and possibilities for future development.
+
+We settled on the following core attributes and the impact in the project bias mitigation across the project life cycle, diversity and inclusivity for project governance, non discrimination in model outcomes and equitable impact of the system.
+
+Our bias mitigation claim is that the SQT reduces undesirable operator variability to limit the impact of cognitive biases when using the image processing pipeline and chosen thresholds.
+
+Standardising analysis methods and providing specific threshold backed by the relevant literature are essential steps in ensuring that clinical decisions are driven by objective data rather than individual interpretations.
+
+It's essential for the SQT to engage patients and healthcare professionals for more backgrounds.
+
+The tool promotes special engagement and interactive decision making through accessible visualisations, making healthcare more inclusive for everyone.
+
+Our claim about this is that the tool supports special engagement and interactive decision making through accessible and informative visualisations and that the user interface and dashboard are intuitive and follow the best practices for presentation of information.
+
+Regards with adequatable impact, the tool is open source and easily accessible to allow clinicians to run software and the tool is efficient in terms of computational resources allowing for widespread use.
+
+Ensuring equitable impact is another key goal of the SQT.
+
+By being open source and efficient in terms of resources, the tool can reach more healthcare providers and ultimately benefit more patients.
+
+Non discrimination is paramount in the development of the SQT.
+
+With deep learning techniques and careful validation, we are working to minimise biases and ensure fairness in the analysis process.
+
+For this, our claim claim is that the tool does not discriminate across different patient groups and ensures quality and consistency of the roles across users with different level of training.
+
+Finally, let's address the challenges and limitations we face in ensuring fairness assurance for the SQT.
+
+While we've made significant progress, there are still hurdles to overcome such as user engagement and validation across different demographics.
+But rest assured, we're committed to continuous improvement and ensuring that the SQT operates ethnically and fairly for all patients.
+
+As we conclude today's presentation, it is crucial to recognise that advancement in tools like the Scar Quantification Tool are pivotal for improving patient outcomes and ensuring fairness and equity in healthcare.
+
+Indeed, by prioritising accuracy, inclusivity and fairness in our analysis processes, we are not only advancing cardiovascular medicine, but also fostering a more equitable future in healthcare.
+
+Thank you for listening.

site/docs/community/case-studies/cemrg-app.md

@@ -0,0 +1,49 @@
+# Case Study: Advancing Cardiovascular Medicine with CemrgApp
+
+<!-- Embed video and transcript here -->
+<iframe class="youtube-video" src="https://www.youtube.com/embed/2V8FAu2U4MM" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture" allowfullscreen></iframe>
+
+## Introduction to CemrgApp
+
+The CemrgApp, developed by the Cardiac Electromechanics Research Group (CEMRG) at Imperial College London, is a transformative tool designed to enhance the analysis of cardiovascular data through advanced imaging and computational techniques. This case study focuses on how CemrgApp uses the Trustworthy and Ethical Assurance (TEA) Platform to enhance its Scar Quantification Tool (SQT), specifically through the integration of fairness assurance analyses.
+
+!!! warning
+
+    This case study is a work-in-progress and represents a snapshot of ongoing developments. The set of claims and analyses should not be considered a complete assurance case.
+
+## Overview of the Scar Quantification Tool
+
+The SQT employs Late Gadolinium Enhanced Cardiac Magnetic Resonance (LGE-CMR) to non-invasively identify scar tissue in the heart’s left atrium. This tool is essential for clinicians to detect potential cardiac issues and strategise effective treatments. The SQT follows a sophisticated process:
+
+1. **Segmentation**: Delineates the left atrium from the rest of the heart.
+2. **3D Modeling**: Generates a three-dimensional representation of the atrium.
+3. **Scar Mapping**: Applies color-coded mapping to visualise scar tissue based on signal intensity.
+
+These stages are managed through an intuitive interface designed to streamline the clinical workflow.
+
+## Applying TEA Platform for Fairness Assurance
+
+A crucial aspect of this case study is the fairness assurance analysis conducted for the SQT. We structured an assurance case focusing on mitigating bias, promoting diversity and inclusivity, and ensuring equitable impact and non-discrimination in model outcomes. The process included:
+
+- **Identifying Fairness Attributes**: Establishing core attributes of fairness including bias mitigation, inclusivity, equitable impact, and non-discrimination that align with the TEA Platform’s methodology.
+- **Developing Claims**: Each attribute of fairness was addressed through specific, actionable claims backed by empirical evidence and theoretical research.
+- **Operationalisation**: Implementing these claims through software features and clinical practices, such as standardised analysis protocols to reduce operator variability and bias in diagnostic outcomes.
+
+## Key Fairness Claims and Their Justification
+
+- **Bias Mitigation**: The SQT standardises scar tissue analysis, which limits the variability that can arise from manual interpretations influenced by cognitive biases.
+- **Inclusivity**: The tool’s interface is designed to facilitate comprehensive patient and clinician engagement, supporting informed decision-making through clear, interactive visualisations.
+- **Equitable Impact**: Open-source access and minimal resource requirements make the SQT broadly available, ensuring diverse clinical settings can benefit from advanced diagnostic tools.
+- **Non-Discrimination**: Advanced machine learning models are employed and regularly validated to minimise biases and maintain consistent performance across diverse patient groups.
+
+## Challenges in Assurance
+
+While the TEA Platform provided a robust framework for developing the fairness assurance case, challenges such as demographic validation, user engagement, and iterative validation of the tool's fairness have been acknowledged. These issues underscore the necessity for continuous improvement and adaptation in response to evolving clinical needs and technological advancements.
+
+## Conclusion
+
+The CemrgApp case study exemplifies the effective application of the TEA Platform’s fairness assurance methodology in a high-stakes clinical environment. By prioritising ethical standards and actionable fairness claims, the project not only improves cardiovascular healthcare outcomes but also advances the field towards more equitable and transparent medical technology practices.
+
+!!! info "Ongoing Developments and Future Directions"
+
+    CEMRG continues to refine the SQT’s functionalities, with the TEA Platform playing a crucial role in ensuring these enhancements adhere to the highest standards of fairness and ethical responsibility. Future updates will focus on expanding validation studies to encompass broader demographics, enhancing user training, and integrating user feedback more seamlessly into development cycles.

site/docs/community/case-studies/index.md

@@ -0,0 +1,11 @@
+# Case Studies
+
+Welcome to the Case Studies section of the TEA Platform documentation. Here, you will find a curated collection of detailed accounts illustrating how various research and software development teams across industries have successfully used the TEA Platform to implement ethics into their data science and AI projects. These stories provide insights into the practical application of the platform in diverse scenarios, showcasing its flexibility and the wide range of problems it can solve.
+
+Each case study is designed to give you a comprehensive view of the challenges faced by teams, the specific features of the TEA Platform they used, the solutions they implemented, and the outcomes they achieved. Whether you are a software engineer, a project manager, or a compliance officer, these case studies will provide valuable lessons and inspiration for leveraging the TEA Platform in your own work.
+
+!!! example "Want to explore what the TEA Platform can do for your project?"
+
+    - **Contact Us**: If you have questions or need advice on how to apply the TEA Platform to your project, our team is here to help.
+    - **Sign Up**: Get started with the TEA Platform today and transform the way you approach AI assurance.
+    - **Share Your Experience**: If you have a story to tell about how you've used the TEA Platform, we would love to hear from you! Your insights could help others succeed.

site/mkdocs.yml

@@ -191,6 +191,9 @@ nav:
           - Local Deployments: platform-details/reset-database/local.md
   - Community of Practice:
       - Community of Practice: community/index.md
+      - Case Studies:
+          - Case Studies: community/case-studies/index.md
+          - Advancing Cardiovascular Medicine with CemrgApp: community/case-studies/cemrg-app.md
       - Community Support: community/community-support.md
       # - Why We Ask for Access to Your GitHub: community/github-access.md  # TODO: No GitHub access currently
       - Upcoming community events: blog/index.md