Comparative Study of a Software With the Gold Standard

Comparative Study of the PRAEVAorta Device in Patients With an Abdominal Aortic Aneurysm (AAA) : Percentage of Adequacy to the Gold Standard

Nurea wants to set-up a pivotal phase clinical study for its software PRAEVAorta, a decision-making support for diagnosis and follow-up of abdominal aorta aneurysm. This study is a comparative study. Its purpose is to demonstrate the adequacy of PRAEVAorta to the current Gold Standard in the care of patients with abdominal aortic aneurysm. The currently benchmark methods are the maximal diameter method (Hand-held measurement of maximum diameter on scan images) and the segmentation method for the aneurysm volume (manual segmentation of scan images, then calculation of the AAA volume by counting the pixels). The hypothesis is that 90% of the segmented patients have a volume ratio and a diameter ration equal to or greater than 90%.

Control and quality assurance will be applied in the investigation site. The monitor mandated by the sponsor Nurea will ensure the follow-up and the proper completion of the study, the collection of data in writing, their documentation, recording and report, in accordance with the procedures in place and in accordance with good clinical practice as well as the laws and regulations in force. Quality control visits realised at regular intervals will be realised by the monitor. During those visits, the following elements could be demanded: Informed consent forms, information note Respect of the plan of the study and the procedures defined in it Quality of the data collected in the Case Report Form (CRF): accuracy, missing data, consistency of data with the "source" documents (medical file, appointment book, originals of laboratory results). Management of any products Security data All contacts made will give rise to a specific written report Three type of visits (quality control) will be realised: The initiation visit (realised in person) will concern every person potentially involved in the investigation. This visit will be about presenting the plan, redefine inclusion objectives, recall everyone's responsibility and the regulation, as well as answering question. The monitoring visits will be realised by the monitor after the first inclusion to ensure the respect of the plan and the regulation, as well as the quality of the collected data. Those visits will be realised every 2weeks, but the frequency can vary regarding the complexity of the study, of the number and the pace of inclusions. Therefore, there will be at least one visit per month. The close-out visit will be realised after the last data have been collected. The purpose of this visit is to verify the quality of the collected data, to update the documents before archiving. In addition to these on-site travels, a regular monitoring over the phone will be realised in order to best support the participating physicians throughout the study. These quality controls are supplemented by regular follow-up audits in order to evaluate the quality of collected data and to crop the study. An audit may be conducted at any time by persons appointed by the sponsor and independent of the persons conducting the research. Its purpose is to verify the participants' safety and respect for their rights, compliance with applicable regulations and the reliability of data. An inspection can also be carried out by a competent authority (ANSM for France or EMA in the context of a European study, for example). The audit, as well as the inspection, can be applied at all stages of the research, from the development of the protocol to the publication of the results and the classification of the data used or produced as part of the research. Data management Access to data Only participating physicians and the monitor (or clinical research staff) will have access to source files during the monitoring visits. During quality control visits, the principal investigator must make the following documents available: informed consent, briefing notes, observation notebooks, source documents (medical records, appointment books, etc.) In accordance with the law in force, patients have a right of access to their medical file. According to the law of January 2002, patients will be told where they can inquire about the results of the study. Guidelines for collecting data The term "Case Report Form (CRF)" designates any document whatever its support (for example paper, optical, magnetic or electronic) intended to collect in writing all the information required by the protocol concerning each person who is suitable for research and to be sent to the sponsor. All the information required by the protocol must be recorded in case report forms and an explanation must be provided for any missing data. The data will have to be collected as and when they are obtained and transcribed in these notebooks in a clear and legible way. It is the responsibility of the investigator to complete the CRF, but he may delegate this task to the clinical trial technicians or to the research nurse, provided that he has previously completed a task delegation form. Data collection will be done using e-CRF that will include patient characteristics and study data. The patient will be identified by an alphanumeric code and the data concerning him/her will be anonymized in the form. The code will be reported on all pages of the observation book. Modification of e-CRF data Corrections of errors which have occurred in an e-CRF will be carried out via the correction module thereof, and must be dated, initialed, and justified. The copies of the observation notebooks will be kept for at least 15 years on the investigation site while the original copy, once the quality control has been carried out, will be archived by the sponsor for the same duration. Standard operating procedures Investigators will determine whether patients meet the inclusion criteria, and whether the use of one or the other device will bring them beneficial results. Screening / pre-selection The selection of the patients is realised by the principal investigator. A first screening of the patient that could be included in the study is realised. To be pre-included, those patients shall: be still followed by the hospital for the duration of the study, have at least one scheduled visit in the Vascular and General Surgery unit over the study period respect the selection criteria (inclusion and exclusion criteria), The discrimination of the group of patients is carried out according to the list above. Pre-inclusion visit The presentation of the clinical study and the proposition to participate to it will be presented to the patient by an investigator doctor during one of his follow-up routine visit (regarding his care pathway). The investigator surgeon verifies the match to the eligibility criteria. The information notes and the consent form are given to the patient at this same visit. A reasonable response period is given to the patient to return the consent form signed. Its inclusion will be effective upon the return of this signed consent. Reception of the consent form / inclusion When the consent of the patient has been received, the following data are registered in the e-CRF: Demographic data: Birth date, age, gender Investigation data: Patient number, date of signature of the consent Images data: Pre-operation scan, post-operation scan The inclusion is validated when all data (cited above) are registered in the e-CRF. Sample size assessment To answer the main objective of this study, we set up a precision of the proportion of volumes calculated by the software PRAEVAorta in adequation with the volumes hand calculated by the healthcare professionals. For this statistical analysis, variables are quantitative: volume of the aneurysm in mm3. We consider over here the proportion of patients' images who have a volume ratio less or greater than 90%. Volume ratio : \frac{PRAEVAorta\ volume-hand\ volume}{PRAEVAorta\ volume} Hypothesis are the following: H0: 90% of the segmented patients have a volume ratio < 90% H1: 90% of the segmented patient have a volume ratio ≥ 90% We would like to have a precision of this proportion of 10%. Therefore, the numbers of patients needed, with an α risk of 5% and a β risk of 0.9% is 35. n=\ \frac{\varepsilon_\alpha^2\bullet p\bullet q}{i^2}=\ \frac{{1.96}^2\times0.9\times0.1}{{0.1}^2}=34.5 With i: precision, p: proportion, q: 1 - proportion, εα: obtained with the reduced table of the normal law. For security reasons and to anticipate the possible withdrawals, we choose to include approximately 10% more subjects, for a total of 40 patients. Patients will be divided in two groups: 20 patients will see their pre-op images analysed and 20 patients will see their post-op images analysed. Satistical analysis plan To evaluate the segmentation obtained with the software, a quantitative comparison with the results obtained from a semi-automatic segmentation manually corrected by a senior and a junior surgeon shall be realised. The Dice Similarity Coefficient (DSC), Jaccard index(JAC), Sensitivity, Specificity, volumetric similarity (VS), Hausdorff distance (HD), maximum aortic transverse diameter, and the duration of segmentation will be calculated between the two methods and, for the semiautomatic software, also between the two observers. A Pearson's coefficient correlation is done to evaluate the correlation between the volumes, surfaces, and diameters measures with the fully automatic and manually corrected segmentation methods.

Automatic segmentation, Abdominal Aortic Aneurysm (AAA), Artificial Intelligence (AI), Deep learning, Automatic volume segmentation, CT scan

This study is comparative, retrospective, monocentric, open, controlled. We compare the SaMD to the Gold Standard. The data used are CT scans which have already been analysed by the doctors. The study will be realised on a group of 40 subjects which is small enough to have one site of investigation. Subjects will be their own witness. This investigation is NRIPH regarding the "Loi Jardé" since it is on retrospective data. Pivotal stage: pivotal, confirmatory, and interventional clinical investigation. This premarket study aims is to confirm the clinical performance, efficiency, and the security of the medical device PRAEVAorta. To do so, a hypothesis has been defined (cf. VI.2.) and the number of patients has been strictly defined and justified (VII.3). CT scan images of patients are analyzed by the surgeons and then by the software under investigation. The results are then compared.

Surgeons extract diameter by hand (Gold Standard method) then realised a semi-manual segmentation of the images to extract volumes and diameters of the aorta (espacially AAA)

CT scans are analysed by the software under investigation : PRAEVAorta. It will extract volumes and diameters of the aorta (AAA espacially)

Inclusion Criteria: Male or female Patient equal or over 18 years old Patient who has given and signed is informed consent Patient with a diagnosed aneurysm Infra-renal AAA ≥50mm or ≥5mm increase in 6 months or Ruptured AAA Patient with legal capacity to make its own decisions, Patient available for set-up and proper implementation of follow-up visits throughout the duration of the registry Exclusion Criteria: Pregnant or planning on becoming pregnant Advancing Alzheimer's disease or socially dependent patient Patient on opposition to the use of their data for this research Patient under the age of 18 Incapable adult Patient under guardianship, curatorship, or safeguard of justice

Schnitzbauer M, Guntner O, Wohlgemuth WA, Zeman F, Haimerl M, Stroszczynski C, Muller-Wille R. CT after Endovascular Repair of Abdominal Aortic Aneurysms: Diagnostic Accuracy of Diameter Measurements for the Detection of Aneurysm Sac Enlargement. J Vasc Interv Radiol. 2018 Feb;29(2):178-187.e3. doi: 10.1016/j.jvir.2017.09.012. Epub 2017 Dec 6.

Kauffmann C, Tang A, Therasse E, Giroux MF, Elkouri S, Melanson P, Melanson B, Oliva VL, Soulez G. Measurements and detection of abdominal aortic aneurysm growth: Accuracy and reproducibility of a segmentation software. Eur J Radiol. 2012 Aug;81(8):1688-94. doi: 10.1016/j.ejrad.2011.04.044. Epub 2011 May 20.

Lareyre F, Adam C, Carrier M, Dommerc C, Mialhe C, Raffort J. A fully automated pipeline for mining abdominal aortic aneurysm using image segmentation. Sci Rep. 2019 Sep 24;9(1):13750. doi: 10.1038/s41598-019-50251-8.