Objective: 1) To compare variables derived from pen-on-paper spirals using an automated algorithm with clinical rating scales for tremor. 2) To analyse the diagnostic utility of derived variables in differentiating tremor spirals from controls.

Background: Spiral drawings using pen-on-paper are commonly used in the clinic to identify and quantify tremor. Scoring systems such as the Bain and Findley scale rate the spirals for tremor severity but are rater dependent. Although tablet based systems can analyse spiral characteristics, lack of temporal domain information makes analysis of pen-on-paper spirals challenging. We aimed to develop an algorithm to objectively analyse pen-on-paper spirals without specialized digitising equipment for data capture, which has the advantage of patient familiarity and requiring minimal hardware.

Methods: 23 subjects with postural/action tremor (dystonic tremor n= 20, essential tremor n= 3) and 20 healthy volunteers were recruited. Subjects were asked to produce a free hand clockwise spiral using the dominant hand on standard pen and paper. Scanned images were subject to pre-processing followed by unravelling of the spiral from Cartesian co-ordinates to radial. The resulting graph was analysed to compute tremor variability (TV) and mean deviation (MD) from a computer generated ideal spiral. The derived parameters were compared with Bain and Findley scale (3 raters) and in patients, the Fahn-Tolosa-Marin tremor rating scale (FTM). Receiver Operating Curve (ROC) statistics were applied to identify cut-offs to detect tremor spirals from controls.

Results: Both tremor variability and mean deviation showed statistically significant strong correlation with Bain and Findley ratings (Spearman’s correlation: TV, rho=0.55, p<0.001; MD, rho= 0.56, p<0.001). The FTM total and dominant upper extremity composite scores did not correlate significantly with TV and MD in patients. Both TV (p<0.001) and MD (p=0.001) were significantly different between patients and controls. TV was the better parameter to detect tremor (AUC: 0.79, 95% CI 0.67-0.93) with a cut-off value of 35.7units showing 74% sensitivity and 85% specificity to detect tremor.

Conclusions: Automated analysis of pen-on-paper spirals without specialized digitizing equipment is feasible and can generate clinically relevant variables to detect and quantify tremor.

« Back to 2018 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/an-automated-algorithm-for-detection-and-quantification-of-tremor-from-pen-on-paper-archimedes-spirals/