Optimizing Sample Size and Statistical Methods for Probabilistic Sweet Spot Mapping in Deep Brain Stimulation.

Area of Science:

• Neurosurgery
• Neurology
• Biostatistics

Background:

• Deep Brain Stimulation (DBS) utilizes probabilistic mapping for movement disorder treatment.
• Current probabilistic mapping methods face challenges in generalizability due to data and statistical method variability.
• Establishing minimum sample sizes and consistent statistical approaches is crucial for reliable probabilistic mapping.

Purpose of the Study:

• To determine the minimum sample size required for stable probabilistic mapping results in DBS.
• To identify the statistical approach that yields the highest consistency in probabilistic mapping outcomes.
• To enhance the clinical relevance and generalizability of probabilistic mapping in movement disorder treatment.

Main Methods:

• Intra-operative stimulation data from 36 Parkinson's Disease patients undergoing DBS surgery were analyzed.
• Probabilistic Sweet Spots (PSS) were computed using sample sizes from 4 to 36 patients.
• Statistical methods included Bayesian t-test, Wilcoxon test with FDR correction, and Wilcoxon test with permutation correction, with 10 repetitions.

Main Results:

• The Bayesian t-test consistently produced stable PSS across all tested metrics.
• Stable PSS size and centroid location were achieved with a minimum of 14 patients.
• PSS volume (Dice coefficient) stability was reached with 18 patients; Bayesian t-test showed superior consistency, especially with smaller sample sizes.

Conclusions:

• The Bayesian t-test is the most suitable method for calculating stable PSS in DBS, particularly with small sample sizes.
• A minimum sample size of 14-18 patients is recommended for reliable probabilistic mapping in Parkinson's Disease.
• These findings are critical for improving the reliability and clinical applicability of probabilistic mapping, refining diagnostics, and optimizing interventions.