Technology
and medical device manufacture. Teams must develop or acquire new cross-disciplinary skills that bridge software engineering, product development, clinical risk management, cybersecurity, data science, and regulatory compliance. As if this wasn’t complex enough many of the
traditional software and compliance skills of development will not easily apply to SaMDs. For example, it’s typical with a Machine Learning (ML) software product to update the ML model with real-time data once it is out on the market but in the healthcare sector much stricter safety protocols apply to every additional data set.2
Surveillance therefore needs to be built in
prior to launch. Developing AI-enabled medical software
requires a broader and more sophisticated skill set than traditional software development. User experience and UI design skills are also critical and come with their own set of specific compliance requirements as do Verification and Validation (V&V) skills and IEC 62304. Determining whether such a range of skills
is available inhouse is critical to success. Working with partners that can offer software roadmaps, help hire lead engineers and advise on regulatory issues such as Post Market Surveillance processes can help build a strong business in case of investor scrutiny into team composition. More established manufacturers that want to
develop a whole product will typically start with a hardware element and often find that they lack the in-house skills to build a software stack to accompany it. Although these larger players typically use a product design company and a cybersecurity expert for example, they would need to ensure there is internal capacity for maintenance once the solution is up and running. Additionally, in these cases it’s critical to find a
trusted partner that can help train internal teams, so they are up to date on their skills.
AI brings further complexity AI-driven software designed for medical purposes must respond to added expectations around safety, risk, data privacy and clear governance. Generative AI adds yet another layer of complexity. Because these models can create new content on their own, they are more susceptible to bias and “hallucinations”. The foundation models underpinning GenAI are often trained on vast datasets and are not always transparent making their use in regulated healthcare settings a thorny issue.3 MLOps requires systematic review of not only the code but also the data pipelines, as data quality is fundamental for AI. This means sanitising and anonymising patient data so that no individual is identifiable and verifying that every file listed in the manifest is present, correct and appropriately handled. The sensitive
nature of the data handled in healthcare makes anonymisation even more important.
Data governance and developing safe software To safeguard patient data the Association for the Advancement of Medical Instrumentation (AAMI) Technical Information Report (TIR) 45 2023 recommends that organisations designate four separate roles that may give voice to the positions and concerns of four different stakeholders. These stakeholders should be: the product owner, the cybersecurity expert, the usability expert and the risk management lead. In conjunction with the data governance lead and data protection officer, a designated cybersecurity expert could contribute to decision as to where specific data sets need to be stored and whether it should be encrypted. They would be responsible for advising on encryption at rest, justification for storing specific data sets or making a selection to minimise risk, preserve anonymisation and optimise integration with NHS or other third- party patient data management systems. Similarly, the cybersecurity expert will be the one to advise and plan on a course of action in case of a security breach.
It is also worth homing in on the management
of a key asset in software development, Software of Unknown Provenance (SOUP) and its management in the healthcare setting. In most industries, SOUP responds to a practical need offering proven components that can be used to save time and money. In the medical sector, every piece of SOUP must be assessed much more thoroughly and to be included in the Quality Management System (QMS). To ensure SOUP is compliant, teams must maintain a clear record of every external software component they use, why it was chosen and how it links
54
www.clinicalservicesjournal.com I June 2026
ImageFlow -
stock.adobe.com
DC Studio -
stock.adobe.com
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72
