Objective hypomimia biomarker for Parkinson's disease diagnosis

Background In vivo misdiagnosis in Parkinson's disease is one of the biggest unmet needs in this disease. In the last decades, several studies attempt to improve diagnostic accuracy by means of quantitative evaluations. Hypomimia is one of the earlier motor symptoms in Parkinson's disease that starts 10 years before clinical diagnosis. The aim of the present study is to use a technology able to automatically extract face features, in order to verify the accuracy of objective hypomimia biomarker for Parkinson's disease diagnosis. Materials and Methods Nine healthy subjects (HS), (age 53.3 ± 8.9 years), and twelve patients affected by Parkinson's disease (PD), according to UK PD Society Brain Bank diagnostic criteria (age 68.2 ± 6.4 years), with and H&Y stage range 1.5-2.5, were enrolled in the study. All enrolled subjects, were filmed, with a standard camera, under two different test conditions for 1 min for each task: (1) rest, (2) conversation. Face features were extracted from digitally recorded video images, focusing on blinking intensity and lips distance, and a final index merging both parameters for rest task was created. Results The maximum intensity of blinking showed to be lower in Parkinson's disease patients compared to controls, during rest task. The absolute value of maximum lips distance showed to be higher in Parkinson's disease patients compared to controls, during both rest and conversation task. The combined index, created using only rest parameters, defined as Parkinson's disease hypomimia predictor (PHP), showed a high diagnostic accuracy (95%) in Parkinson's disease vs. healthy subjects discrimination, with an ROC AUC of 0,949, a positive predictive value (PPV) of 92% and a negative predictive value (NPV) of 100%. The correlation between the MDS-UPRS III item 3.2 and PHP was rs(18) = 0,738. Discussion Both literature and the present study data showed that hypomimia in Parkinson's disease patients, is a good candidate as a proxy symptoms for diagnosis of Parkinson's disease. Quantitative and objective assessments make assessments more accurate and reproducible. In the present study, the data were collected in the less intrusive way, by using only a standard camera, without markers placed on the face of subjects. Custom parameters were calculated from the extracted face features, focusing on the most relevant hypomimia features, in line to face features evaluated in clinical practice and MDS-UPDRS scale, i.e. blinking and lips movements. The final predictor (PHP) showed a high diagnostic accuracy and, in addition, the variable MDS-UPRS III item 3.2 and PHP were found to be strongly correlated, showing that PHP could be a useful objective tool to evaluate hypomimia in Parkinson's disease. Conclusion In conclusion, PHP is a new hypomimia measure, which can be an aid tool for the diagnosis of Parkinson's disease. This new tool has a high diagnostic accuracy, positive and negative predictive value. It can be derived from short, cheap, widely available and non-intrusive face recordings at rest, and it is correlated to standard clinical motor scale scoring system.