Editorial by Prof Manolis Tsiknakis, Coordinator of ProCAncer-I project
The ProCAncer-I project has completed the first year of its lifecycle at the end of October 2021. With this brief note key results and achievements of the first year are summarized in the sequel.
The main focus of our work has been devoted to crystallizing the various clinical endpoints on which the project focuses, the so-called clinical Use Cases (UCs). The UCs within the ProCAncer-I project represent the drivers to answer Prostate Cancer (PCa) relevant clinical questions, ranging from PCa diagnosis and characterization to prediction of treatment response and occurrence of side effects after treatment.
Nine such UCs have been defined that cover the whole prostate cancer care continuum. I will provide a more detailed description of a couple of our UCs, so that our readers understand the clinical rationale and importance of the work performed in ProCAncer-I.
- UC2 focuses on the characterization of cancer according to its biological aggressiveness into clinically significant and non-significant disease. UC2 aims to stratify men with suspicious findings on MRI into high-risk cases, which need radical treatments to ensure that cancer will not grow and spread to remote parts of the body becoming a deadly disease, from low-risk cases which could be safely followed up with active surveillance, avoiding comorbidities of treatment and ensuring the highest possible quality of life for the patients;
- UC5 focuses on the prediction of the risk of disease recurrence after radical prostatectomy, based on imaging data and AI techniques. In UC5 post-surgery findings, as positive surgical margins and extracapsular extension, will be considered in a nomogram comprising also radiomics and clinical variables to predict disease recurrence. UC5 will help clinicians to “modulate” radical prostatectomy (in terms of different nerve sparing approaches, according to the site of extracapsular extension and/or seminal vesicle invasion), tailoring treatment to the predicted risk of disease recurrence;
In the context of this stream of activity, a very detailed definition of related clinical and histopathological data that needs to accompany the imaging data has been defined and technical decisions regarding the standards to be used for its representation supporting interoperability with other similar repositories and infrastructures.
A second main area of work has focused on establishing the legal framework of the project. All collected data during the lifetime of the ProCAncer-I project will be anonymized on the premises of the clinical partners (Data Controllers). All data elements, either imaging data, or/and clinical data and metadata, will be anonymized prior to their upload and storage on the ProCAncer-I cloud platform. Significant efforts, both analytical and development have taken place in developing the ProCAncer-I Upload Tool, which is the communication gateway with the envisaged web platform, to execute the defined data anonymization process. In parallel, a detailed analysis of the risks of Re-identification within the ProCAncer-I Health Data Sets has taken place. Study protocols have been developed for both the retrospective as well as the prospective data collection. Most clinical partners have completed the process to obtain approvals from their corresponding Ethical Committees. This was a very important milestone, successfully achieved.
The third main area of work has been the definition and implementation of the cloud-based platform of the project and its integrated tools. User access to the ProCAncer-I platform will be granted via an Authentication and Authorization Infrastructure (AAI) implementation and the de-identified data will be accessed in a community-driven fashion. A fault-tolerant, elastic, and high-performance computing infrastructure provides the services and tools for the implementation of AI algorithms, execution of advanced processing steps, and visualization of results. Tools have been developed and will be provided to various user groups, i.e. clinicians/radiologists and other clinical staff, in order to support the efficient upload, curation, and annotation of imaging and related clinical data. In specific, the image curation tool developed guides users through a set of steps for the data curation, i.e. motion-correction, co-registration, quality check, final approval, and storage of the derived images. Also, for the annotation of the collected DICOM studies uploaded in the ProCancer-I platform, a web-based annotation tool has been developed and integrated with the rest of the ProCAncer-I platform components. All relevant metadata are stored in the Metadata catalogue of the project using an OMOP-based common data model that appropriately documents and catalogues data and model curation steps enabling traceability and trustworthiness.
Finally, along with the brief summary of activities and achievements, key work was also been performed, jointly with all other AI4MI projects (EUCanImage, Chaimeleon, INCISIVE and PRIMAGE), through a number of working groups that have already produced important and internationally visible results.
Having successfully completed the first, very important, phase of the project’s lifecycle, the second, equally important, phase starts that focuses on uploading initially the retrospective data from the ProCAncer-I clinical partners and subsequently the relevant prospective data. In parallel, AI-base model development will begin targeting the initially described UCs.