How SEDGE Complies with WHO’s 6 Guiding Principles for AI Use in Healthcare


The World Health Organization (WHO) recently released its guideline on the use of artificial

intelligence (AI) in healthcare, titled Ethics and Governance of Artificial Intelligence for Health.

The guideline identifies potential risks, challenges, and recommendations that stem from an

ethical point of view when it comes to the use of AI in healthcare, and stipulates six principles as

the basis for AI regulation and governance. SEDGE, the AI-powered tool that turns medical data into actionable insight, is happy to show its compliance with the six principles outlined in this guideline.


1. People stay in control


Human autonomy is at the top of the WHO list, and with good reason. SEDGE supports this with its built-in ‘explainability’ function that helps people understand the predictions outcomes generated by the AI. It does this by unpacking what factors resulted in the formation of the predictions, keeping people in the driving- seat. Privacy and confidentiality are protected through SEDGE’s strict adherence to international privacy legislation, such as the GDPR.


2. Public interest, public safety, and well-being are paramount

The WHO guidelines stipulate that the use of AI in healthcare must meet regulatory requirements for safety, accuracy, and efficacy - with measurable quality control - within well-articulated use

cases or indications. For any product the quality of a product, service or process should

always be checked in project management. SEDGE supports project management best

practice and the role and responsibility this places on every AI project lead to ensure that

predictions provided by the AI tool do not compromise these criteria in any way.


3. Transparent, explainable, intelligible


Transparency is only possible when sufficient literature and data is made available to the public for review, debate, and consultation about a design or deployment of an AI. SEDGE aligns itself with this, in that every AI project has a clear vision statement document defining the purpose of the AI project, and how it will bring value to people and the business. A detailed document showing the source of data, data quality checks, missing data and how it needs to be interpreted, data exclusions, imputation of data, and so on, is clearly documented. As for explainability, SEDGE displays the variables that have a strong influencing factor in terms of supporting or contradiction of the target, and the degree of support or contradiction. SEDGE also provides multiple models to extract the model with the best accuracy through the use of its AutoML feature, which can be run by users who have little or no knowledge of Data Science. This supports the intelligibility objective.


4. Responsible and accountable


As per WHO, AI technologies perform specific tasks, it is the responsibility of stakeholders to ensure that they are used under appropriate conditions and by appropriately trained people. To ensure people are appropriately trained to use the AI, SEDGE has various videos and user guides that give users the basic and detailed training they need to use the SEDGE AI model building platform appropriately. Provision is made for the AI project leads to build effective mechanisms

into the AI application that allows users to question and redress for individuals.


5. Inclusive and equitable


The AI must be designed to include the widest equitable use and access. SEDGE by default uses all the variables which are defined, highlighting aspects that may lead to bias in the model. The list of fields which are data sensitive are highlighted by SEDGE, and this advises the users to be aware of the fields which are Sensitive data and should be treated with caution and as per the guidelines provided by GDPR or equivalent personal data protection regulations.


6. Responsive and sustainable


The WHO’s final principle for AI in healthcare stipulates that designers, developers, and end users should be transparent and honest when it comes to using and testing the AI to ensure it meets expectations and requirements, without compromising on the environment. AI systems should also be designed to minimize their environmental consequences and increase energy efficiency. SEDGE provides model management functionality that takes into account the model drift that

usually takes place when data changes over time. As model accuracy drops over time, the models will need to be retrained with fresh, new data, and retested for old and new conditions.


Image prediction models take a long time to train. It is therefore important to use

Transfer Learning features which allow existing learnt models to be used as the starting

point for the new models’ learning. This reduces the environmental impact, as the

process of learning the model from the start can take a long time for learning, impacting

prolonged use of GPU / TPU servers, which affects the environment and consumes a

large amount of energy.


The information or views expressed in this article is authentic to the best of our knowledge, and

as such, it is prone to errors and the absence of some key information, for more information

please review our Terms of Use.