Search
Presentation Details
| Poster #287 IMPROVING OLFACTORY SENSITIVITY DETECTION USING TREESCLASSIFICATION |
| Prasanna Karunanayaka2, Pemantha Lakraj1, Senal Peiris2 1Department of Statistics, University of Colombo, Colombo, --, Sri Lanka 2Department of Radiology,Penn State University College of Medicine, Hershey, PA, United States |
Introduction: Olfactory function plays a critical role in daily life, offering advantages such as enhanced threat detection (e.g., environmental hazards, smoke), improved social interactions, and heightened flavor experiences. However, the ability to smell declines with aging and neurological conditions such as Alzheimer’s and Parkinson’s disease. Therefore, developing quick and reliable tools to assess olfactory function is essential in order to detect those changes. Method: Höchenberger et al. (2019) introduced a Bayesian adaptive algorithm called QUEST for estimating olfactory sensitivity using Sniffin’ Stick data. Their study compared sensitivity thresholds derived from QUEST with those obtained through a standard staircase method. Their findings revealed substantial overlap between the two methods, with QUEST demonstrating slightly higher test-retest correlations, reduced measurement variability, and had a shorter testing duration. Building on their work, we incorporated tree classification techniques, including Partial Tree and Partial Random Forest, to further refine threshold estimation under Staircase and QUEST methods. Results: Based on simulation results, there are significant differences between mean thresholds of the staircase, partial tree and partial random forest methods. Mean threshold of partial random forest and partial tree are higher than that of staircase Conclusion: Incorporating tree classification techniques can be a robust approach for olfactory threshold assessment. Further research, however, is needed to optimize these techniques for assessing olfactory performance. |