AI Skin Cancer Innovation Challenge
Use of Artificial Intelligence to accelerate the detection of skin cancer in referrals from primary to secondary care
This challenge will provide an opportunity for companies to enhance or refine existing AI algorithms developed using healthcare data, for testing in redesigned community to hospital referral pathways. This will lead to accelerated detection and diagnosis of skin cancer and better use of constrained healthcare resources.
All applications for this challenge will be managed through the Innovate UK website CLICK HERE
DO NOT SUBMIT YOUR SOLUTION TO SCOTLAND INNOVATES WEBSITE. THIS CHALLENGE IS BEING ADMINISTERED VIA IUK.
Improving outcomes and experience for people accessing Health and Social Care and delivering a “Once-for-Scotland’’ approach to support a rapid diagnosis of skin cancer is a key priority for NHS Scotland, supported by the Centre for Sustainable Delivery. Skin cancers are the most common group of cancers diagnosed worldwide. Melanoma Skin Cancer Statistics show Melanoma is one of the most common cancers in young adults. Link to Statistics It is estimated that 325 000 new melanoma cases and 57 000 deaths due to melanoma occurred in 2020, with large geographic variations in incidence across countries and world regions. If 2020 rates remain stable, the global burden from melanoma is estimated to increase to 510 000 new cases and 96 000 deaths by 2040.” Global Burden of Cutaneous Melanoma in 2020 and Projections to 2040
“Basal cell carcinoma (BCC) is the most common form of skin cancer. An estimated 3.6 million cases of BCC are diagnosed in the U.S. each year. Squamous cell carcinoma (SCC) is the second most common form of skin cancer. An estimated 1.8 million cases of SCC are diagnosed in the U.S. each year.8,1. More than 5,400 people worldwide die of nonmelanoma skin cancer every month.” Skin Cancer Facts & Statistics - The Skin Cancer Foundation
A paper published in 2013 estimated that by 2020 the NHS cost of skin cancer in England would be between £170-£190 million. Measuring current and future cost of skin cancer in England. An analysis of Medicare payments for non-melanoma cancer published in 2016 noted that “Five million cases cost $8.1 billion in 2011.” The Economics of Skin Cancer: An Analysis of Medicare Payment Data
Diagnosis and management of suspected skin cancer represents over 50% of a specialist dermatology service workload. Demand for dermatological care already outstrips capacity, exacerbated by vacancies across Primary and Secondary Care. Using ISD data and local NHS Board data it was estimated that:
• Of the 146,694 dermatology referrals resulting in booked appointments in Scotland in 2019-2020, 50% would be estimated to be skin lesions (73,347).
• Of those, 70% would be non-malignant (51,343), which if detected earlier in the pathway would free up clinical time and reduce anxiety for people worrying if they have skin cancer
• and 30% would be malignant (22,004), allowing earlier treatment and cure.
• Potential for releasing even 30% (15,402) of the non-malignant appointments would equate to 2567 hours, or £2,310,435 (@£150/10min outpatient appointment).
To meet demand, a new pathway of care, integrating existing expert healthcare professionals with validated AI solutions, is being developed using test beds within NHS Scotland. The COVID pandemic has seen the introduction of several innovative methods of delivering care for those with suspected skin cancer, using digital referral and consultation methods. Image-based, diagnostic AI has huge potential to support diagnostic triage by GPs, reducing referrals to secondary care, accelerating diagnosis and treatment whilst delivering safe care nearer to home and alleviating patient anxiety.
The challenge solutions will, within redesigned patient pathways:
• Increase the speed of diagnosing skin cancer from images with referral data received by specialist dermatology services from GP Practices and community centres;
• Allow improved and accelerated diagnosis of urgent potential cancer cases;
• Contribute to finding missed skin cancer diagnoses linked with the pandemic;
• Reduce the number of people now presenting with later stage skin cancer;
• Increase secondary care productivity by at least 10%;
• Reduce backlog of appointments and waiting lists across Scotland and elsewhere in the world.
The Dermatology AI Consortium is working in collaboration with the Centre for Sustainable Delivery to develop a “ Once for Scotland” Digital Dermatology Pathway which will enable the Consortium to achieve an ambitious target of diagnosing skin cancer in 25 minutes by 2025 utilising machine learning . This LINK describes work that has been undertaken by the Consortium and CfSD, and the joint milestones .
To improve existing AI algorithms for skin cancer to accelerate detection and diagnosis.
To achieve this applicants will be provided with real world images and metadata in a standardised DICOM format in a trusted research environment. The data pipeline that has been created for this challenge contains standardised image and metadata, captured as part of everyday patient pathways from specifically designed community locality image centres, labelled by consultant dermatologists and wrapped in DICOM format. The specific use case for these algorithms is the stream of potential skin cancer referrals from primary to secondary care.
In the first phase of the challenge you will use pre-existing, pre-procurement algorithms to test small volumes of patient data, using all images and metadata to assess performance of your algorithms with the datasets. You will be expected to understand the performance of your algorithms prior to the challenge, and in phase 1, success will be determined not by the absolute performance, but by the ability to improve performance using new DICOM data fields. The end of phase 1 assessment process will include a written report and interview, involving specific reflections on how you have been able to incorporate and use new data fields, and the steps you would perceive that are required to improve algorithm performance in phase 2. In addition, we may also seek preliminary performance features such as the area under the cost-coverage curve set at different rejections rates and/or expected calibration errors.
The assessment panel will comprise of computer scientists, clinician scientists with an interest in AI, and innovation professionals. Successful AI algorithms from this challenge will have the potential to accelerate detection and diagnosis of skin cancer, optimise use of clinical time and increase available outpatient appointments.
Successful companies going through to Phase 2 will improve and refine the capability of the algorithm to diagnose skin cancer using simulated, or actual healthcare data and present data back to the clinician in a way that fits with the clinical diagnostic process and pathway.
Standardised clinical DICOM standards and metadata have been agreed with the British Association of Dermatologists AI group.
A report describing clinical, operational and information workflows and technical architecture for the four NHS Boards has been completed informing minimum standards and specifications required for testing and introducing use of artificial intelligence into patient pathways.
Algorithms should have already been developed for this specific use case in this challenge. This challenge should permit algorithms to be developed further through
using DICOM standardised images and metadata captured through a uniform clinical pathway.
Development of completely new algorithms and commercially available algorithms developed without use of real-world healthcare data are out with the scope of this challenge. We are therefore seeking partners who have developed skin cancer algorithms, for the specific use case outlined above, preferably tested on real world data sets, who wish to improve the diagnostic ability of their AI before commercialisation.
Commercially available digital equipment and technology for data capture, storage and transmission are out of scope of this challenge, However, future procurement of this, or related innovation to digitally enable dermatology pathways is being synchronised with this work.
Long term vision
The NHS Scotland AI Skin Cancer Consortium was created in 2021 to address healthcare challenges and accelerate the introduction of AI into skin cancer detection and diagnosis. The consortium has set a challenge to diagnose skin cancer within 25 minutes by 2025 across Scotland. The goal of innovation and development of initiatives is to improve outcomes and experiences for all who access the NHS and will:
• Permit rapid diagnosis of skin cancer
• Improve quality of care for patients with an earlier diagnosis and treatment of skin cancer
• Reduce mortality from skin cancer through access to early diagnosis
• Streamline clinical pathway and national database of images linked to diagnosis and outcome
• Form part of an integrated pathway of care, where AI complements the skills of health care professionals in delivering safe and timely care
• Improved patient journey and create operational efficiencies
• Adhere to the core principles of realistic medicine
Such initiatives must work with local, regional, and national NHS Boards, Regional Test beds, Primary Care groups where appropriate to provide solutions which are readily tested, evaluated and reported in 12-18 month timescales to affect wider national strategies and NHS adoption.
Funding Type: Pre-Commercial Procurement
The Scottish Health Industry Partnership (SHIP) is funding this Dermatology AI challenge. Organisations can apply for a share of £500,000, inclusive of VAT over a 2 phase competition.
The programme will be delivered in up to 2 phases. This is phase 1. A decision to proceed with phase 2 will depend on the outcomes from phase1. Only successful candidates from phase 1will be able to take part in phase 2.
Phase 1 is allocated £90,000 for a maximum of 3 projects, Phase 1 projects can range in size up to total costs of £30,000 inclusive of VAT.
Phase 2 is allocated £300,000 for an anticipated 2 projects. Phase 2 projects can range in size up to total costs of £150,000 inclusive of VAT.
Your project’s total eligible costs must include all costs associated with any subcontractors and VAT. Successful applicants receive 100% funding and access to advice from the NHS Scotland innovation hubs.
Innovators must work with the NHS Test Bed and partners to provide solutions within a 12-18 month timescale.
The challenge ambition is to support pre-commercial experimental development, feasibility testing and evidence gathering to enable future commercial procurement across NHS Scotland.
There is a preference for projects at an advanced stage of development, near ready to be deployed in a real-world situation. For projects with advanced prototypes, we would like to see evidence of a roadmap to achieve regulatory compliance.
Projects showing high potential but at an earlier stage of development may also be considered.
To lead a project, you: • Can be any type of organisation of any size, registered in the UK, European Union (EU) and the European Economic Area (EEA) that can demonstrate a credible and practical route to market
• Can work alone or with other organisations as subcontractors
• must work in conjunction with one of the NHS Scotland Innovation test beds Must provide details of certification and compliance with relevant standards, accreditation and regulatory approval for well-developed prototypes.
Contracts will be awarded to a single legal entity only.
• Competition opens : Monday 27th March
• Online briefing event : Monday 17th April 1.00pm - 2:30pm
• Competition closes : Friday 19th May 11:00 am