Periodic Reporting for period 1 - OperA (OPTIMISING COLORECTAL CANCER PREVENTION TROUGH PERSONALISED TREATMENT WITH ARTIFICIAL INTELLIGENCE)
Reporting period: 2022-09-01 to 2024-02-29
The OperA project is a multidisciplinary project with eight different work packages (WPs) to optimize colorectal cancer prevention through personalized treatment with artificial intelligence. The following specific goal for each work package will help achieve this overarching goal.
WP1. Establish the value of artificial intelligence (AI)-assisted colonoscopy in colorectal cancer prevention by conducting a pan-European, population-based, randomised trial.
WP2. Develop a colonoscopy AI risk-prediction tool for personalized treatment of colorectal polyps and cancer.
WP3. Develop cost-effectiveness models of AI-assisted colonoscopy in colorectal cancer screening.
WP4. Investigate ethical and legal barriers in AI development and implementation.
WP5. Generate the first trustworthy and rapidly updated (“living”) clinical guidelines for AI in screening colonoscopy.
WP6, 7, and 8. Facilitate patient-oriented dissemination, communication, and management of the project.
WP1. Establish the value of artificial intelligence (AI)-assisted colonoscopy in colorectal cancer prevention by conducting a pan-European, population-based, randomised trial.
WP2. Develop a colonoscopy AI risk-prediction tool for personalized treatment of colorectal polyps and cancer.
WP3. Develop cost-effectiveness models of AI-assisted colonoscopy in colorectal cancer screening.
WP4. Investigate ethical and legal barriers in AI development and implementation.
WP5. Generate the first trustworthy and rapidly updated (“living”) clinical guidelines for AI in screening colonoscopy.
WP6, 7, and 8. Facilitate patient-oriented dissemination, communication, and management of the project.
Here, we provide short summary of the progress of each of the scientific WPs (WP1-5).
WP 1: We are conducting a pan-European, randomized trial to achieve this goal. Establishment of the trial infrastructure (ethics approval, clinical trial registration with the summary of the protocol) has been successfully achieved. Following this, we have included 9,704 patients as of February 2024.
WP2: We are developing an AI-based pathology prediction tool for colorectal lesions. The first one and half years have been spent primarily for preparation of the study infrastructure including ethics approval of hospitals in six countries (Norway, Germany, Poland, Italy, Spain, and Japan), development of the cloud-based data storage system in the UK, and development of the preliminary, semi-supervised learning algorithm. In line with the ethics approval, we have already started collection of more than 60,000 endoscopic images. In line with development of the AI algorithm we are communication with a consultation company which supports smooth applicaiton for EU-MDR approval.
WP3: We are developing a brand-new microsimulation model with the use of trustworthy clinical data derived from large-scale prospective studies. We have already developed the first microsimulation model simulating the effectiveness of AI-assisted colonoscopy. The resulting effectiveness data has been used as an evidence base to develop trustworthy guidelines in a close collaboration with WP5.
WP4: We have organized the ethics monitoring committee to audit if development of the AI tool in WP2 is done in an ethically acceptable way. We have also organized the position statement committee to maximize the dissemination effect of the WP4. We have already published several articles in major journals with a strong focus on ethics and legal issues in AI.
WP 5: We are closely working with MAGIC Foundation to publish trustworthy and rapid recommendations on AI in colonoscopy screening. We have organized a guidelines committee and completed most of the processes required to develop trustworthy guidelines. The guideline-work includes selection of panel members, identification of the PICO question, development of the systematic review, and drafting the recommendations based on the voting by the panel members.
WP 1: We are conducting a pan-European, randomized trial to achieve this goal. Establishment of the trial infrastructure (ethics approval, clinical trial registration with the summary of the protocol) has been successfully achieved. Following this, we have included 9,704 patients as of February 2024.
WP2: We are developing an AI-based pathology prediction tool for colorectal lesions. The first one and half years have been spent primarily for preparation of the study infrastructure including ethics approval of hospitals in six countries (Norway, Germany, Poland, Italy, Spain, and Japan), development of the cloud-based data storage system in the UK, and development of the preliminary, semi-supervised learning algorithm. In line with the ethics approval, we have already started collection of more than 60,000 endoscopic images. In line with development of the AI algorithm we are communication with a consultation company which supports smooth applicaiton for EU-MDR approval.
WP3: We are developing a brand-new microsimulation model with the use of trustworthy clinical data derived from large-scale prospective studies. We have already developed the first microsimulation model simulating the effectiveness of AI-assisted colonoscopy. The resulting effectiveness data has been used as an evidence base to develop trustworthy guidelines in a close collaboration with WP5.
WP4: We have organized the ethics monitoring committee to audit if development of the AI tool in WP2 is done in an ethically acceptable way. We have also organized the position statement committee to maximize the dissemination effect of the WP4. We have already published several articles in major journals with a strong focus on ethics and legal issues in AI.
WP 5: We are closely working with MAGIC Foundation to publish trustworthy and rapid recommendations on AI in colonoscopy screening. We have organized a guidelines committee and completed most of the processes required to develop trustworthy guidelines. The guideline-work includes selection of panel members, identification of the PICO question, development of the systematic review, and drafting the recommendations based on the voting by the panel members.
Because this reporting is the first periodic report as a summary of the first one half years, main analyses of the core clinical studies have not been done in both WP1, WP2, and WP4. On the other hand, we have experienced significant progress in WP3 and WP5, which are microsimulation modelling and guideline development, respectively. Details of these two WPs will be shown in the upcoming publications, but we can emphasize that the methodlogies we took in these WPs are novel and clinicall relevant. To make the simulation model trustworthy in WP3, we did not use transition probabilities between different disease states in the construction of the model. Instead, we retrieved data on the relationship between intervention status and long-term outcomes such as cancer incidence from the prospective trials, which will result in more trustable and transparent simulation results that can be used to inform decision makers. WP5 has also adopted trustworthy and reasonable methodology to develop the guidelines with thorough compliance with GRADE system in developing guidelines, which may be novel in the AI-medicine field. We also included three patients in the panel members to best reflect patients' perspective in the guideline recommendations. Together with the results of WP3, this high-quality guideline is expected to affect decision making process of physicians, payers, and healthcare policy makers.