PREDICTION OF BREAST CANCER POSSIBILITY IN THE GENERAL HEALTH NETWORK

Aim of this study is to formulate an algorithm based on neural network modelling to calculate the index of individual risk of breast cancer (BC). We examined 1440 women. All subjects aged 18 to 85 years. To achieve the objective used clinical, laboratory and instrumental methods (mammography, ultrasonography). Questionnaire method, methods for constructing artificial neural network models and logistic regression models, as well as the program «oncologist» are based on a neural network. 12 women of 360 clinically healthy women surveyed were assigned to high-risk group. The majority were women 41-50 age group - (75 ± 0,1%). Women aged 51-60 totalled 16,7 ± 0,1%, in 31-40 years - 8,3 ± 0,1%. In the «medium risk» assigned 95 women - 26,4 ± 0,1%, in the « low risk « - 253 women 70,3 ± 0,1%. Statistically significant differences (p < 0,001). Belonging to high-risk groups increases the reliability of the pathological process in 2.2 (CI 1,4-3,5 ) times (p < 0,001). In women at high and medium risk of developing breast cancer as a result of a comprehensive survey found suspected breast cancer one woman (by histological examination - fibroadenoma), fibro-cystosis disease (FCD) in 18 women with one woman - fibroadenoma of the breast. In the low-risk group for breast cancer FCD was detected in 4 women, nodules were not identified. Statistically significant (p < 0,001) risk factors were set up for each age group in the Bukovynskoy region, which make up the largest proportion. An automated system (based on a neural network, which allows you to create « high-risk « of breast cancer and precancerous diseases. Sensitivity of predictive models - 79,8%, specificity - 79,0% with 86,3% diagnostic accuracy.

selective screening, breast cancer, neural network