Paige Prostate for prostate cancer

Innovations

Paige Prostate is a deep learning system that is a type of machine learning software. It has learned directly from thousands of slides using clinical diagnostic reports, without the need for per‑pixel annotation by using multiple instance learning. This approach has the potential to improve the accuracy and speed of detecting prostate cancer. While alternative artificial intelligence (AI)‑based systems exist, the company claim that the Paige Prostate algorithm has been developed to be highly robust to variations in slide preparation from different institutions and does not need a per-site calibration using per‑pixel annotations or other forms of calibration data. With Paige's FullFocus viewer pathologists can view the biopsy tissue and simultaneously access Paige Prostate results for AI‑assisted diagnostic reporting.

Current care pathway

People with suspected prostate cancer are usually seen within the primary care setting and offered a blood test that looks for raised prostate specific antigen levels. If these are raised, NICE's guideline on prostate cancer recommends offering multiparametric MRI (mpMRI) as the first line investigation. Results should be reported using a radiological scoring system such as the Prostate Imaging Reporting and Data System (PI-RADS) or the 5-point Likert scale. Within the secondary care setting, urologists will consider if a biopsy is appropriate depending on the results of the mpMRI. People whose mpMRI score is 1 or 2 may opt out or opt in for a systematic prostate biopsy after discussing the risk-benefit ratio of the procedure with a healthcare professional. Individuals with a score of 3 or more should be offered a prostate biopsy.

There are several ways biopsies can be done, which include transrectal ultrasound‑guided biopsy and transperineal template biopsies. The transrectal biopsy is usually done using local anaesthetic and takes 5 to 10 minutes. A needle is inserted through the wall of the back passage to obtain 10 to 12 small pieces of tissue from different areas of the prostate. The transperineal biopsy is done under local or general anaesthetic and the approach is through the skin of the perineum. The samples are then processed and stained, before its morphology is studied under a microscope by a histopathologist. Biopsy review might include conventional microscopy (using a standard microscope) and digital pathology (a computer‑based viewing of the whole slide digital image of a glass slide), and experts advise it is becoming more common these methods are used in parallel. The tissue may be examined using a digital system and computer monitor instead of a standard microscope. If cancer is detected, it is graded according to the Gleason grading system (and the Grade Group category) and additional quantitative information, such as the numbers of biopsy cores with cancer and the maximum length of the cancer (in mm), is normally provided as part of the report, to further inform the management. Cases may be processed with one viewing, but in some instance other opinions from colleagues may be warranted as well as further staining (immunohistochemistry) or deeper sections may be requested. Cases may also be reviewed at cancer centres as part of multidisciplinary team meetings to guide disease management decisions such as active surveillance or radical treatment.

The NHS rapid diagnostic and research pathways handbook for implementing a timed prostate cancer diagnostic pathway set out that, if appropriate, a prostate biopsy should be done within 9 days from GP referral and a target of 5 days turnaround for reported pathology should be agreed as a minimum standard. This is a 14‑day turnaround from GP referral to prostate biopsy result. Many services adhere to the The Royal College of Pathology (RCP) key assurance indicators for laboratories. According to Prostate Cancer UK, the diagnostic pathway can take up to 28 days before a definitive diagnosis is made. There is an acknowledged capacity challenge in the area with an increasing complexity and volume of pathology requests but with a lack of pathologists (RCP workforce census, 2018).

The following publications have been identified as relevant to this care pathway:

Population, setting and intended user

Paige Prostate is intended to be used in addition to current methods of assessing, detecting and characterising prostate cancer. There are reportedly over 40,000 new cases of prostate cancer a year in the UK (NPCA, 2017). If a person has signs and symptoms which suggest prostate cancer, a referral is made to a urology specialist. Diagnosing prostate cancer begins in a secondary care setting using biopsies. Paige Prostate is for use in cellular pathology departments by the pathologists to assist in the analysis of biopsies for the detection and characterisation of prostate cancer.

The company provide initial in person or remote training on Paige Prostate to all users as part of the standard subscription fees, which usually takes less than half a day to do. Additional advanced application or specific feature training can be provided upon request. Training materials including tutorial videos are also provided.

Costs

Resource consequences

Following on from being awarded Artificial Intelligence in Health and Care Award, the technology is understood to be in the process of being deployed for prospective clinical use at 3 NHS trusts.

Paige Prostate can be used by cellular pathology laboratories that have partial or fully digital operations. Those with partial digital operations might need a clinical grade scanner to use this technology, depending on what types of digital pathology slide scanner(s) are already available. The NHS long term plan identified digital transformation as a priority area across the NHS. The number of laboratories with digital pathology within the NHS is increasing through several different initiatives, such as the AI Centres of Excellence and Upscaling AI centres programme including the work of the PathLAKE and PathLake Plus consortium. The benefits of digital pathology in enabling remote analysis and supporting high demand have been seen during the Covid‑19 pandemic and the Royal College of Pathologists provided guidance for the remote reporting of digital pathology slides during exceptional service pressure. Digital pathology offers greater potential for collaborative work and learning as well as health and safety benefits with no physical movement of slides, avoiding potential for damage or infection transmission. The work process does involve an additional task of glass slide scanning into the workflow.

Adopting the technology is likely to cost more than standard care, but the company claim it has greater benefits which may lead to releasing resources and producing cost savings overall. Cost savings may be produced because of improved productivity, where there is published evidence to support these claims, as well as improved patient outcomes, eliminating the costs of unnecessary treatment or progression of a disease and its related costs. There are no published studies on prospective use of this technology to improve patient outcomes or optimise treatment. The investment in scanning infrastructure may affect multiple other workflow improvements and AI applications across all areas of cancer and non-cancer diagnostics in the NHS.