Skin Cancer Diagnosis with Artificial Intelligence (AI)
by Tomasz Puk | Jan 11, 2019
The following case study of Pro4People IEC 62304 compliant software development service describes a project we delivered for an innovative UK medical startup – Skin Analytics. Their mission is to help more people survive skin cancer. The client wanted to develop a unique and innovative skin cancer diagnosis system.
The client had started in 2012, surely from the right angle, with the design and training of an AI component, and then proving it with clinical evidence. The developed solution uses Artificial Intelligence to process skin lesions images in order to provide early detection of skin cancer conditions. An image can be taken from home or at health care providers such as pharmacies, health centers or hospitals. This way, the preventive diagnosis is made available in the blink of an eye.
Starting the Project
SkinAnalytics contacted Pro4People, as they were looking for a software solution provider and AWS Consulting Partner that would develop a backend part of the overall solution. The backend was supposed to be highly reliable and scalable, and ready for global deployments based primarily on Amazon Web Services cloud infrastructure. The solution had to comply with medical device regulations applicable to health software such as IEC 62304, ISO 14971, and parts of ISO 13485 standards. The developed system needed to fit in, and interface with the already existing components of Skin Analytics ecosystem, such as mobile apps or health care providers’ applications.
The initial 4 week-long phase, called Pre-Development, started in July. This is a typical way in which we at Pro4People, start health software projects: It gives the initial project team time to work together and agree on all deliverables and processes so specific to Software Development Life Cycle. The major deliverables from that phase were:
- software requirements
- software development plan
- initial architecture of the system
- initial detailed designs required to understand the complexity of the overall solution
- agreed processes to be applied
- agreed list of deliverables expected at the major project milestones.
The biggest advantage, in addition to hard deliverables, was, undoubtedly, an opportunity to get to know each other. Setting up effective communication patterns with such tools as JIRA, Confluence, Slack, GoToMeeting helped us to make sure we would be working as one, though geographically distributed team.
IEC 62304 Training
In the meantime, both Product and Project teams took part in a common IEC 62304 training at Pro4People office in Wrocław. The training was delivered by our experienced project manager based on the IEC 62304 Know-how set from ins2outs quality management system. The goal of the training was to make sure we all worked on the same processes and deliverables required from Software Development Life Cycle. It helped us also to introduce a common naming convention and align on the deliverables. All the team members received training completion certificates helping them to meet the competencies requirement from ISO 13485 standard. Equipped with the knowledge about tools, processes, development, CI, CD, we were ready to start the implementation.
The actual software development phase started at the end of August. For that phase we used Agile / Maturity approach. The basic development follows IEC 62304 standard, with all the required deliverables and activities in agreement with the SDLC. Nevertheless, that software development effort was additionally divided in 2 week-long sprints.
Pro4People provided a fully managed and independent Project Team. It included the roles such as: Project Manager, Business Analyst, Software Developers, Quality Assurance Engineer and DevOps/Configuration Manger. The team was working closely with the Skin Analytics Product Team with their CTO and Product Owner being the main contact points.
Technically, the overall solution was designed for Amazon Web Services cloud environment. The overall architecture was designed with trouble free future deployments to local data centers and private cloud deployments in mind. The architecture of the overall system followed a loosely-coupled pattern. This way, the delivered solution is highly scalable and most of its components are capable of horizontal scaling. Meeting a global market demand will not be a problem when the interest in the service develops.
Since the system was planned to fit into the existing Skin Analytics ecosystem, the external interfaces were agreed, designed and documented in minute detail. Thus the future integration with any third-party health care provider would be simplified. All the required, meticulous designs were created iteratively, making room for Agile / Maturity approach.
The regulatory part of the project involved executing all the IEC 62304 required processes and delivering documents and records. It is quite a lengthy list so in case you are looking for details, please refer to IEC 62304 know-how set.
Start your Medical Device Project
Verification and Validation
Throughout the whole software development life cycle, testing activities were executed. The approach to verification and validation was first described and agreed with the client in theTest Strategy. It introduced the required tests levels, approach to the problem resolution process and the way the deliverables should be verified at each project milestone.
The testing, with respect to Agile part of the project, was covered by the so called “progressive tests”. Each issue in the spring backlog was documented in JIRA. Once they were resolved, the Quality Assurance Engineer verified them independently before closing.
In addition, within Test Design and Implementation Process the tests were gradually documented (or implemented in case of automatic ones). This way the appropriate test levels were prepared. When a milestone or a release came, the agreed test levels were executed to verify the quality of the project and its deliverables.
With the final release, all the test levels were executed. Used test specifications and test execution reports were delivered to the client as part of a Release Package.
The first MVP version of the project was delivered in December after 5 months of development. Quite an impressive result IMHO, for a Class IIa Medical Device operating in AWS environments as a global solution. Skin Analytics team was very satisfied with our collaboration and continued the activities in the Product Life Cycle by setting up staging environments, validation and commercializing the new service. We finished the project with the lessons learned, meeting, and switching to, other challenges.
From our perspective, it was a great experience to work with the Skin Analytics team on such an innovative and much needed solution in the domain of skin cancer diagnosis. The project was surely not an easy one, but a mutual understanding of regulated project requirements and excellent communication were the key to success in that assignment.
Neil Daly - CEO of Skin Analytics
“We selected Pro4People as a strategic development partner based on their deep understanding of delivering Software as a Medical (SaMD) products. They’ve been engaged, collaborative and an extremely valuable partner. “
CEO of Skin Analytics
Were you interested in reading this case study?
The next successful project can be yours.
Please contact us
Above we have discussed the Skin Cancer diagnosis with Artificial Intelligence. Should you wish to talk to us about supporting you in such a project, please check out one of the below links.