Digitization of Gynecology Using Artificial Intelligence: Cervical Mapping Corroborated With Clinical Data for Conization Necessity

Abstract

Abstract Background Cervical cancer is the fourth most common female malignancy worldwide. In developing countries, it is the most common subtype of cancer and the third leading cause of cancer mortality among women. Artificial intelligence has the potential to be of real use in the prevention and prompt diagnosis of cervical cancer. The aim of our study was to develop a medical platform consisting of an automated observation sheet containing colposcopy data, a software that would use a machine learning module based on clinical and image data for diagnosis and treatment, and a telemedicine module to enable collaboration between gynecologists. Materials and methods Clinical and colposcopy image data from 136 patients were introduced into a machine learning module designed to generate an algorithm for proposing a preliminary diagnosis and treatment. The clinical and imaging data were corroborated to generate six options: ‘Follow-up’, ‘Pharmacotherapy’, ‘Biopsy’, ‘Curettage’, ‘DTC’, and ‘Conization’. Results Data generated by the machine learning module regarding treatment options were compared with the opinion of gynecologists and yielded an accuracy of 78% for ‘Follow-up’, 81% for ‘Pharmacotherapy’, 84% for ‘Biopsy’, 90% for ‘Curettage’, 96% for ‘DTC’, and 81% for ‘Conization’. Conclusions The developed software can be an important step towards the digitization of existing gynecology offices and the creation of intelligently automated gynecology offices related to prevention and treatment of cervical cancer. More data is needed to improve the accuracy of the developed software.