The AI4POLYPNET Network Enters the Final Stages of Writing the First White Paper on the Use of Artificial Intelligence Systems for Colorectal Cancer Detection and Diagnosis

• The AI4POLYPNET network aims to develop and validate artificial intelligence systems designed to improve the diagnosis of colorectal cancer.
• The CVC has hosted the network’s second working meeting, bringing together eight institutions – four with a clinical profile and four with a technical profile.
• AI4POLYPNET is funded by the Spanish State Research Agency (Agencia Estatal de Investigación), under the Ministry of Science and Innovation.

The national AI4POLYPNET network, dedicated to improving colorectal cancer diagnosis through Artificial Intelligence (AI), held its second working meeting in Bellaterra, Catalonia. Comprising eight institutions – four clinical and four technical – the network represents a collaborative effort to address one of the most prevalent diseases worldwide.

One of the network’s main objectives is to establish standardised data acquisition and annotation protocols that will enable the development and validation of AI-based colonoscopy support systems with real impact on clinical practice, ultimately improving diagnostic capacity for colorectal cancer.

Colorectal cancer is one of the most common forms of cancer, ranking as the second most frequent in women and the third in men in Europe. Globally, around 361,986 new cases are diagnosed annually in Europe, with mortality reaching 161,182 patients in 2022. Despite its high incidence, more than 90% of cases can be cured if the precancerous polyp—the precursor lesion—is detected and treated in time.

Although several detection techniques exist, colonoscopy remains the most effective, as it allows clinicians to detect and remove lesions in a single procedure. However, this technique has limitations: approximately 22% of lesions are not detected during examination, resulting in missed opportunities for prevention and effective treatment.

In recent years, advances in machine learning and computer vision have driven the development of AI-based methods to support clinical staff in detection and diagnosis. Yet, the scarcity of high-quality annotated data and the limited ability to predict histology have posed significant challenges in this field.

It is within this context that AI4POLYPNET emerged—a national network focused on the development and validation of intelligent systems for colorectal cancer detection and diagnosis. The initiative aims to establish common imaging acquisition and annotation protocols, as well as standardised validation systems, to ensure the robust performance of the developed methods.

Over the two years in which the network has been active, significant progress has been made in defining the clinical needs that intelligent systems should address. For each need, the network has identified both the amount and type of data required for system development, as well as the evaluation metrics that best reflect the potential clinical benefit of using such systems. In addition, a set of best-practice guidelines will be established for using these systems in the examination room, including the selection of necessary equipment and compliance with current regulations.

The outcome of the joint work of the eight participating institutions will be the publication of the first white paper on the use of artificial intelligence systems for colorectal cancer detection and diagnosis. The document will cover a wide range of aspects, providing contextualization for research in the field and, most importantly, linking technical developments with real clinical needs to define which systems should be developed in the future to achieve a tangible impact on clinical practice.

During the second AI4POLYPNET working meeting, held on Thursday, November 13 and Friday, November 14, at the Computer Vision Center (CVC), the network advanced in the drafting of the white paper, reviewing chapters nearing completion, with a particular focus on defining the clinical needs to be addressed. The foundations were also laid for the key chapter that will outline the data acquisition and annotation protocols, as well as performance metrics.

AI4POLYPNET is funded by the Spanish State Research Agency of the Ministry of Science and Innovation and is composed of the Computer Vision Center (network coordinator); the Gastrointestinal and Pancreatic Oncology Research Group at Hospital Clínic de Barcelona; the Jesús Usón Minimally Invasive Surgery Centre (CCMIJU); the Digestive System Department of the University Hospital of Cáceres; the Next-Generation Computer Systems Group (SING) at the University of Vigo; the Digestive Oncology Group of Ourense (GIODO) at the Galician Health Service; the eVIDA Research Group at the University of Deusto; and Osakidetza.