For the past three years, Professor Adrien Depeursinge and his team have been organizing an international competition bringing together the best researchers in the world to create and test algorithms to improve throat cancer treatment and prevent the risk of recurrence. The HECKTOR Challenge organized by the eHealth unit of the Institute of Informatics of the HES-SO Valais-Wallis is a concrete contribution to the fight against cancer and to the improvement of patients' quality of life.
Head and neck cancers are among the most frequent cancers in the world. Treatments are improving, but therapeutic failures remain a major challenge and occur in up to 40% of cases in the first two years after treatment. Several studies are attempting to better identify patients with a worse prognosis by exploiting already available images such as those acquired for diagnosis and treatment planning. Although very promising, these methods have been validated on an insufficient number of patients because they require a long, complex, and error-prone work of image collection and annotation: the manual and three-dimensional delineation of primary tumors and lymph node metastases for each patient.
Methods using artificial intelligence algorithms currently exist, both to automate tumor contouring and to identify patients with an unfavorable prognosis. However, the performance of these algorithms is not yet good enough to be used by the medical profession. Thus, it is essential to compare the work of the global research community to advance these approaches. The organization of an international competition, the HECKTOR Challenge (HEad and neCK tumOR segmentation and outcome prediction in PET/CT images), aims at comparing and identifying the best algorithms according to the same rules and on the same medical dataset of sufficient quality and quantity. The scientific results and the knowledge generated are made available to all (open science and open data).
For the third consecutive year, this international competition is organized with many Swiss and foreign institutions including a majority of hospital partners. Led by Professor Adrien Depeursinge, assisted by his colleagues Vincent Andrearczyk, Scientific Assistant, and Valentin Oreiller, PhD student at the Computer Science Institute of the HES-SO Valais-Wallis, the HECKTOR challenge has seen the number of participants increase dramatically over the three editions, as cancer treatments are a central issue of public health worldwide.
After the success of the first two editions, the 2022 challenge will be presented at the 25th International Conference on Medical Computer Imaging and Computer-Assisted Interventions (MICCAI) 2022. In this year's edition, researchers will focus on one of the most complex problems of this disease, namely the identification of patients who are at high risk of relapse after treatment.
Interdisciplinarity. This is a term that keeps coming up in cancer research and control. HECKTOR relies on a consortium of clinical experts in cancer imaging and treatment as well as researchers in artificial intelligence. Nine medical centers provide data (CHUV Lausanne, US Zürich, CHB Rouen, CHU Poitiers, MDA Houston, CHU Montreal, CHU Sherbrook, HGJ Montreal, HMR Montreal) and several companies support the challenge such as MIM which provides the data annotation software, Siemens Healthineers, Aquilab and Biomtech which offer prizes to the winners. This community is committed to helping ENT tumor patients benefit from the best treatment options among the many available (surgery, chemotherapy, radiotherapy, immunotherapy).
To do so, artificial intelligence can help in the management of patients through personalized medicine, which aims to improve their quality of life. Medical imaging allows the physician to diagnose a disease and make treatment decisions. However, these images contain a lot of information that is not yet exploited by humans but can be read by algorithms. In the spirit of complementarity, the use of an algorithm in a quantitative approach to medical imaging would offer an additional tool to the physician who could then put the newly acquired information into context and propose better targeted and more effective treatments.
By bringing together the best research teams in the world around a competition, the promising methods provided by the algorithms will allow for greater precision in diagnosis and treatment and an improvement in the quality of life of patients. Artificial intelligence focuses on the contours of the tumor and the algorithms will eventually be able to:
- Predict the risk of recurrence
- Assist in the choice and planning of treatment
- Reduce invasive procedures (biopsy), particularly for diagnosis
The 2022 Challenge should make it possible to find the tumor, outline it, search for metastases in the lymph nodes and predict the risk of recurrence.
Once the challenge is over, the HES-SO Valais-Wallis research team will be able to present the results of the best algorithms to the scientific community to generate new knowledge and improve the work in progress.