A NEW METHOD BASED ON ARTIFICIAL INTELLIGENCE IN THE DIAGNOSIS OF CERVICAL LESIONS USING A COMBINATION OF MODERATE BINARY PATTERNS DEVELOPED LOCALLY AND SUPPORT VECTOR MACHINE

Invasive cervical cancer is the second most common cancer among women worldwide. There are many methods based on artificial intelligence to accurately diagnose the normality or cancer of the cells, which help the specialist to diagnose cancer cells better and faster. This study aimed to present a new and efficient method for automatically detecting normal and abnormal cells.

This was a descriptive study. In order to create the database, 2600 images were prepared from 150 cytological slides. Images were evaluated, identified, and classified by specialists. In order to evaluate the proposed method in the prepared database, out of 2600 images prepared, 1300 images were considered for system training and 1300 images for testing. This research used MATLAB software version R2014b to evaluate and compare the proposed method with other methods.

Morphological extractors were used to extract the characteristics of the cells in all three stages, and support vector machine, logistic regression and C4.5 classifications were used for classification, respectively. The accuracy of the proposed method in detecting cervical cells in both normal and abnormal groups was 98.23%, which is more than other methods, and also the ratio of false positives (0.92%) and false negatives (0.85%) is lower than other methods.

The proposed method can help significantly in the field of diagnosis in medicine with the early detection of cancer and, in many cases, can be very effective in timely treatment of the patients and prevention of their mortality.