Review of Various AI Based ADR Detection for Medical Diagnostic Applications

Recently, there has been a rise in interest in pharmacovigilance research, which aims to discover more time-effective and a larger variety of adverse drug reactions (ADRs) from social texts connected to beliefs about taking medication. Specifically, this interest has been sparked by the fact that this study has become more time-intensive. When it comes to extracting ADRs from social texts, one of the most important tasks is to identify the postings that contain ADRs. This may be done by searching for terms like “ADR” in the post's content. The effectiveness of the mechanisms that are in place now is unsatisfactory because there is insufficient room for emotional expression, and there is insufficient information offered in social communications that are brief. Despite the fact that these systems utilize emotional characteristics in order to increase the performance of their approaches, the portrayal of emotional scores at the word level is insufficient for the display of emotions. Additionally, fewer systems leverage medical expertise to better identify the potential link between certain prescriptions and negative comments on postings. In this work, various Artificial Intelligence (AI) based ADR detection process is reviewed to study the function and characteristics. Various feature extraction techniques and classifiers were reviewed in this work to study the performance of the various ADR techniques. Sensitivity evaluation metrics were used to evaluate the performance of various conventional and AI-based ARD detection processes.