GI Genius: The Legal regulations for introducing Artificial Intelligence ‘machine-learning' onto the UK market

Currently, all artificial intelligence machines utilizing 'machine learning' fall under the definition of “narrow AI” as opposed to “general AI”. Meaning that machines focus on specific tasks, [1] rather than a broad rage of tasks designed to match human-level intelligence in regards to its broadness and adaptability. [2]

A colonoscopy is performed by passing an endoscope, a thin flexible tube with a small camera, [3] through the anus and into the patients rectum and large intestine. The large intestine is inflated using a scope in order for better visualisation of the lining. An gastroenterologist will visualise abnormalities on the lining of the rectum and large intestine during the procedure for signs of cancer.

Colorectal cancer is the fourth most diagnosed cancer in the UK [4] and the second leading cause cancer death within the UK, however detection of suspicious lesions within the bowel earlier significantly improves patient outcomes [5] with more than 90% of people with bowel cancer surviving for five years or more if diagnosed at the earliest stage, falling to 10% when bowel cancer is diagnosed at the most advanced stage. [6] Implementing artificial intelligence in order to enhance the success rate of detecting lesions during colonoscopies is imperative in order to improve the chances of recovery from colorectal cancer.

The Artificial Intelligence in Health and Care Award is an NHS AI Lab programme run by the Accelerated Access Collaborative (AAC) in partnership with the National Institute for Health Research (NIHR). It aims to support AI technologies in order to aid the NHS in its long term plans in regards to AI. Three rounds have been completed from September 2020 to March 2023. [7] Medtronic Limited was one of the round 3 winners of the AI in Health and Care Award in March 2023. [8], [9] As a result, Medtronic was awarded £2.5 million of government funding in order to conduct a trial of GI Genius within NHS hospitals, [10] an intelligent endoscopy module from Cosmo Pharmaceuticals and distributed by Medtronic. The trial will run from summer 2023 until late 2024

GI Genius uses computer-aided detection (CADe) to process colonoscopy images, identifying subtle lesions, potentially missed by human error, and highlighting them on the endoscopy display. Furthermore, when a potential lesion is detected by GI Genius, CADx is implanted to predict the diagnosis of the identified polyps, helping in preventing unnecessary polyp removals in order to conduct biopsies.

The intelligence GI Genius possesses was trained and validated using a randomised controlled study which used white-light endoscopy videos of 2,684 historically-confirmed polyps from 840 patients who underwent colonoscopies. It was found that overall sensitivity per lesion was 99.7% and false positives occurred at a rate less than 1% [11], [12]

86% of cases of post-colonoscopy colorectal cancer (PCCRC) are thought to have been preventable. 57.8% of cases of post-colonoscopy colorectal cancer (PCCRC), colorectal cancer diagnosed after a colonoscopy where no cancer was found, are due to missed lesions, potentially due to 45% of polyps having a “ flat macroscopic appearance”, [13] as well as the size of polyps, with 22% of polyps under 10mm being missed during regular colonoscopies, [14] these factors cause polyps to be difficult to detect with the human eye.

GI Genius is already available within the UK’s private medical sector [15] however is not yet available on the NHS outside of this trial.

GI Genius falls into the artificial intelligence category of Machine Learning and more specifically the sub-category of Machine Learning known as Deep Learning. Machine learning was defined by Arthur Samuel in 1959 [16] as when AI is applied using algorithms that learn from an apply data in order to make informed decisions. Deep learning is a sub-category of machine learning defined by Rina Dechter in 1986, it functions by “structuring algorithms in layers in order to create an “artificial neural network” [17] that can learn and make intelligent decisions on its own.

There are legal regulations companies have to follow in order to sell their medical device on the UK market. They must achieve a UKCA Marking.​

In order to medical devices such as GI Genius to be available through the NHS they must have a UKCA Marking, a marking required for a product to be sold on the UK market.

There are legal requirements which need to be met in order to acquire a UKCA marking, it must be proven that the device falls into one of the three categories within the MDR 2002, in the case of GI Genius this would be general medical devices or “Part II of the MDR 2002”. [18] Under MDR 2002 7(1) the classification criteria for classes I, IIa, IIb and III are defined in Annex IX of Directive 93/42, taking into account Directive 2003/12 and Directive 2005/50. Annex IX of Directive 93/42 sets out 18 rules in order to narrow the classification of medical devices. An endoscope falls into class IIa. [19], [20] However the GI Genius itself is not an endoscope, it is an “endoscopy module” meaning that it is an “add on” to a traditional endoscope. GI Genius therefore falls under III(1.2) of Annex IX as it is “non-invasive” but “connected to an active medical device in class IIa” as it connects to a traditional endoscope, therefore it is also class IIa. One could argue that Annex IX does not have a large enough scope since it does not explicitly take into account AI medical devices, a continuously growing category of medical devices.

Since GI Genius is class IIa it would have been subject to a conformity assessment in order for the declaration to be approved. Once a certificate has been received from the approved body who conducted the conformity assessment, a UKCA mark can be placed on the product and it can be sold on the UK market.

Furthermore, in order to sell products on the UK market, an application must be made through the Medicines and Healthcare products Regulatory Agency (MHRA), the agency responsible for the regulation of the UK medical devices market. Manufacturers who want to enter the Great Britain Market in order to sell their medical device are required to register with the MHRA. Registering with the MHRA is subject to provisions, since GI Genius is not based within the UK, they would have had to appoint a Responsible Person who was based within the UK in order to submit an application, this person would take on the responsibility for the medical device, assuming the responsibilities of the manufacturer in regards to registering the device. [21]

Following the coordinated assessment pathway, once the MHRA Devices application is confirmed as valid, ethical approval under the Medical Devices Regulations 2002 must be obtained from a Research Ethics Committee (RAC) within the UK Health Departments’ Research Ethics Service. [22] RACs review research proposals and decide if the research is ethical, it is an independent body enabling them to put participants first when reviewing a research proposal. [23]

The MHRA and REC reviews are completed simultaneously as the organisations share information throughout the assessment process in order to reach an informed decision. [24]

GI Genius is innovative and has the potential to save lives by catching cancer early, however, there is a need for explicit legal regulation regarding medical devices which utilise and rely on artificial intelligence.

References:

[1] Dennis Schlegel, Yasin Uenal, 'A Perceived Risk Perspective on Narrow Artificial Intelligence' [2021] Conference: 25th Pacific Asia Conference on Information Systems (PACIS)

[2] Dennis Schlegel, Yasin Uenal, 'A Perceived Risk Perspective on Narrow Artificial Intelligence' [2021] Conference: 25th Pacific Asia Conference on Information Systems (PACIS)

[3] NHS, 'What is a colonoscopy?' [2024]

[4] Sarah E W Briggs, Philip Law, James E East et al, 'Integrating genome-wide polygenic risk scores and non-genetic risk to predict colorectal cancer diagnosis using UK Biobank data: population based cohort study' (2022) 379 British Medical Journal DOI: 10.1136/bmj-2022-071707

[5] Medtronic, 'Prestigious NHS AI award paves way for real-world trial of GI GeniusTM intelligent endoscopy module' [2023]

[6] Cancer Research UK, 'Why is early cancer diagnosis important?' [2024]

[7] NHS England, 'Artificial Intelligence in Health and Care Award' [2023]

[8] NHS England, 'AI in Health and Care Award winners' [2023]

[9] NHS England, 'Artificial Intelligence in Health and Care Award' [2023]

[10] Medtronic, 'Prestigious NHS AI award paves way for real-world trial of GI GeniusTM intelligent endoscopy module' [2023]

[11] Cesare Hassan, Michael B Wallace et al, 'New artificial intelligence system: first validation study versus experienced endoscopists for colorectal polyp detection' (2020) 69(5) British Medical Journal DOI: 10.1136/gutjnl-2019-319914

[12] Cesare Hassan, Michael B Wallace et al, 'New artificial intelligence system: first validation study versus experienced endoscopists for colorectal polyp detection' (2020) 69(5) British Medical Journal DOI: 10.1136/gutjnl-2019-319914

​[13] Chantal M C le Clercq, Mariëlle W E Bouwens, 'Postcolonoscopy colorectal cancers are preventable: a population-based study' (2014) 63(6) doi: 10.1136/gutjnl

[14] Van Rijn, Jeroen, Reitsma, Johannes et al, 'Polyp Miss Rate Determined by Tandem Colonoscopy: A Systematic Review' (2006) 101(2) American Journal of Gastroenterology

[15] The London Clinic, 'GI Genius' [2024]

[16] A. L. Samuel, et al, 'Some Studies in Machine Learning Using the Game of Checkers' (1959) 3(3) IBM Journal of Research and Development

[17] Rina Dechter 'Learning While Searching in Constraint-Satisfaction-Problems' [1986] Conference: Proceedings of the 5th National Conference on Artificial Intelligence.

[18] Gov.UK, 'Medical devices: conformity assessment and the UKCA mark' [2024]

[19] MEDDEV 2.4/1 – rev. 8 'Part 2: Guidelines for the classification of medical devices' [2001]

​[20] 'MDCG 2021-24 Guidance on classification of medical devices' [2021] Medical Device Coordination Group Document

​[21] Gov.UK 'Register medical devices to place on the market' [2024]

[22] NHS Health Research Authority 'Medical devices and software applications' [2023]

[23] NHS Health Research Authority, 'Research Ethics Committees overview' [2020]

​[24] NHS Health Research Authority 'Medical devices and software applications' [2023] ​