Identifying pathological facial weakness using fuzzy inference

Stroke is the second largest cause of death and disability-adjusted life-years in the world. Minimising the time to treatment for patients is extremely important. Facial weakness is a core symptom that medical professionals consider when identifying cases of stroke. This is a subjective assessment of asymmetry in the face. Due to this subjectivity, it is challenging to articulate the decision making process of a neurologist. This work presents novel computational approaches to accurately model the detection of pathological facial weakness from images of people with and without pathological facial weakness. Instance segmentation is first used to isolate key facial features that inform the decision making of a fuzzy inference system. This proof of concept study shows the feasibility of automated feature extraction and the effectiveness of fuzzy inference systems in identifying facial weakness. Furthermore, the transparent nature of the instance segmentation model and the fuzzy rule base has enabled the model to be compared against the real-world decision-making process of a neurologist. The findings motivate future investigations to develop fuzzy inference systems to detect other common deficits of stroke including limb weakness and drift as well as dysarthria.