Medical decision support system for cancer treatment in precision medicine in developing countries

Abstract

In many developing countries and regions, there are medical problems such as dense populations, lack of medical resources, and shortage of doctors, making it impossible to provide patients with more convenient full-cycle services. Non-small cell lung cancer is a malignant tumor with the highest morbidity and mortality in the world. Research on the auxiliary treatment of intelligent systems is helpful in understanding and evaluate the disease. The system can help doctors provide patients with effective drug treatments and personalized medical services by actively learning the experience of outstanding experts. According to expert knowledge, through quantitative efficacy scores and specific analysis of evaluation indicators, based on the efficacy evaluation matrix and the extraction of key features of patients and drugs, a predictive model of drug efficacy evaluation for adjuvant therapy is established. The model divides into latent feature extraction and curative effect collaborative prediction modules. In the feature extraction module, adding noise to the original data in model training process helps reduce the impact of the sparseness of the patient's medication data. By considering the uncertainty of experts in drug efficacy evaluation modeling, based on probability analysis and efficacy prediction, the proposed method demonstrates the potential options in the face of hesitating choices. According to the predicted efficacy score, candidate drugs are selected to assist doctors in disease analysis and secondary diagnosis. Experiments have shown that drug efficacy prediction methods can provide adjuvant treatments for diseases and quantify the therapeutic effects of targeted drugs. The efficacy information and detection information of patient-drug pairs are helpful to improve decision-making ability, and the proposed medical decision support system framework is superior to other deep learning methods. By adding data, the performance can be significantly improved.