With approximately 14 mln new cases and 8 mln cancer-related deaths in 2012, cancer is a major cause of morbidity and mortality worldwide. The Baltic States have one of the highest incidences of cervical cancer (CC) among European countries ranging from 46/100000 in Lithuania, 35/100000 in Estonia, and 27/100000 in Latvia. CC mortality is higher in Baltic/central/eastern European countries than elsewhere in Europe and rising.[3, 4] Effective CC screening has a proven, strong effect for preventing cancer at the population level, but it is not satisfactory in the Baltic States.
Cervical cytology based (PAP test) cancer mass screening is one of the 20th century success stories.  Almost all Nordic countries have established mass screening for CC. However, the participation in screening is suboptimal, with a 3-year coverage below, or close to, 70% in Denmark, Sweden and Norway in 2015. Furthermore, in many countries including Estonia, Latvia, Lithuania, Belarus, Bulgaria, and Russia, an increase in CC incidence has recently been observed despite screening efforts. Mass-screening, with its one-size-fits-all strategy, is an infrastructural challenge in low resource settings which face about a five times higher burden of CC compared to high-income countries [8, 9] and establishing more effective and cost-effective personalised screening programmes is highly relevant. The introduction of HPV vaccines, the use of HPV testing in screening, and the use of HPV typing and other biomarkers after a positive screening test [10-11] makes a personalised risk assessment pertinent. The concept of personalised cancer prevention is attracting increasing interest as the screening, diagnostic and treatment choices are increasing due to scientific discoveries. The added complexity requires computerised assistance for proper management of population segments with different risk levels (i.e. WLWH) and challenges the conceptual and logistical framework of delivering the existing cancer screening which is designed to deliver preventive health care in a “one-size-fits-all” approach.
Allocation of accurate individual risks could have an immediate application to screening programmes by allowing the determination of optimal screening intervals for each individual based on their individual data/risk. This is in contrast to the current practice where recommended screening intervals are the same for large groups of women. Probabilistic models, such as the Hidden Markov Model, can determine patient-level probabilities of being in a particular disease state. To improve emerging fitting issue, models require retraining. Biological data mining approaches, particularly those related to artificial intelligence (AI) and machine learning, could address current epidemiologic limitations and are starting to be explored in populationbased studies that include patient and biologic level data. These approaches are model-free, nonparametric, and allow for high performance computing that can incorporate AI approaches with human knowledge. Recent successful examples show the feasibility of combining such data and analytic approaches for translational and clinical research.[14, 15] This project seeks to implement AI in medicine together with “real-world” validation of AI algorithms for sustainable and effective cancer control.