Development of artificial intelligence applied to predictive analysis of recurrence and personalization of medical treatment for patients with breast and lung carcinoma

Abstract

Breast cancer is one of the highest worldwide incidence. In Brazil, according to INCA report, the former mostly affects women, being responsible for around 28% of mortality among diagnosed cases and has by association risk factors such as obesity, sedentary lifestyle, frequent exposure to ionizing radiation, in addition to evidence such as smoking. Moreover, in the three-year period from 2023 to 2025, the estimate is that 70 thousand people will be affected by breast cancer and of these, 40% may experience a recurrence of the disease. Treatments for different type of tumor have distint protocols depending on the grade, stage of the disease and type of neoplasm. In the case of breast cancer, breast surgery followed by chemotherapy or radiotherapy are the mostcommon. Currently, the immunotherapy associated with liquid biopsy has stood out in the evaluation of ctDNA (circulating tumor DNA) which helps in the detection of circulating tumor DNA in the blood in order to evaluate possible diagnoses of recurrence. Within this context, many different treatments when combined, can offer the patient a longer survival free from disease recurrence. Still, many patients who undergo this treatment do not experience clinical benefit and end up presenting relapse or progression of the disease. From the aforementioned analyses, it is interesting to note that the treatment is prescribed based on the clinical data of patients described in the literature. However, there may be other clinical aspects to be considered when detectingbetter treatment prescription in order to avoid relapses that have not been considered. In this way, the patient’s clinical and demographic data with the neoplasm in conjunction with those of other patients can provide valuable information for predictive analyzes that can be carried out by applying Artificial Intelligence (AI) techniques. Furthermore, studies can be carried out on which variables help to increase the probability of recurrence, as well as recommending the treatment protocol personalized based on predictive analysis and generated models. As highlighted, there is currently a set of variables described in the literature which may indicate factors that have recurrence as a prognosis. However, greater number of variables that can corroborate the analysis can allow more assertions are taken for treatment. There are still few published works that apply AI techniques to predict the probability of neoplasm recurrence and which clinical variables are predominantly responsible. Within this context, Personalized Medicine emerged as a way to assist in the diagnostic profile followed by the analysis of the patient’s probability of recurrence in an assertive and accurate way for each case. To this end, the development of a solution based on Artificial Intelligence capable of tracing a patient’s clinical profile and evaluating the best treatment in a personalized way becomes an increasingly pressing need. In this way, data collection have come from Doctor Arnaldo Cancer Institute in Sao Paulo, Brazil through acollaborative platform. This platform have allowed structuring data from patient anamnesis, as well as collecting radiological images so that machine learning and deep learning models can be generated, capable of extracting knowledge from these data and supporting medical decision-making. The solution must, among others, propose personalized forms of treatment, estimate the probability of recurrence, and reduce the length of hospital stay.

Alexandre Ray

Senior Data Scientist
Machine Learning Engineer
AI Engineer

My research interests include leadership, team science, and open science