Longitudinal Boundary Sharpness Coefficient Slopes Predict Time to Alzheimer's Disease Conversion in Mild Cognitive Impairment: A Survival Analysis Using the ADNI Cohort

Abstract

Predicting whether someone with mild cognitive impairment (MCI) will progress to Alzheimer's disease (AD) is crucial in the early stages of neurodegeneration. This uncertainty limits enrollment in clinical trials and delays urgent treatment. The Boundary Sharpness Coefficient (BSC) measures how well-defined the gray-white matter boundary looks on structural MRI. This study measures how BSC changes over time, namely, how fast the boundary degrades each year works much better than looking at a single baseline scan for predicting MCI-to-AD conversion. This study analyzed 1,824 T1-weighted MRI scans from 450 ADNI subjects (95 converters, 355 stable; mean follow-up: 4.84 years). BSC voxel-wise maps were computed using tissue segmentation at the gray-white matter cortical ribbon. Previous studies have used CNN and RNN models that reached 96.0% accuracy for AD classification and 84.2% for MCI conversion, but those approaches disregard specific regions within the brain. This study focused specifically on the gray-white matter interface. The approach uses temporal slope features capturing boundary degradation rates, feeding them into Random Survival Forest, a non-parametric ensemble method for right-censored survival data. The Random Survival Forest trained on BSC slopes achieved a test C-index of 0.63, a 163% improvement over baseline parametric models (test C-index: 0.24). Structural MRI costs a fraction of PET imaging ($800--$1,500 vs. $5,000--$7,000) and does not require CSF collection. These temporal biomarkers could help with patient-centered safety screening as well as risk assessment.

Figures (4)

• Figure 1: Complete BSC slope-based MCI-to-AD conversion prediction pipeline. Longitudinal T1-weighted MRI scans from four time points (baseline, m12, m24, m36) undergo preprocessing (N4 bias correction, skull stripping, resampling), Atropos segmentation to derive GM/WM probability maps, and BSC computation at the gray-white matter boundary. Per-scan features (35 total) are extracted from each time point, then organized into per-subject longitudinal sequences. Linear regression across time points yields 182 slope-based features per subject capturing annual rates of boundary degradation. The top 20 most variable slopes are selected and input to a Random Survival Forest model, which outputs individual risk scores, Kaplan-Meier survival curves stratified by risk group, and performance metrics (C-index: 0.63).
• Figure 2: Random Survival Forest algorithm schematic. The original training data is repeatedly bootstrap sampled (typically 63% of subjects with replacement, leaving 37% as out-of-bag samples). For each bootstrap sample, a survival tree is grown by recursively splitting nodes using randomly selected candidate variables that maximize survival differences between daughter branches (measured by log-rank statistics). This process generates an ensemble of diverse trees (Ntree = 1,000 in this implementation). For prediction, each tree gives a cumulative hazard estimate based on which terminal node the subject falls into, and the final ensemble prediction averages across all trees. The out-of-bag (OOB) data gives internal validation for calculating error rates and variable importance without needing a separate validation set.
• Figure 3: Cross-validation performance comparison across 8 model variants. Box plots show C-index distributions from 5-fold cross-validation for training (left), test (middle), and overfitting gap (right) sets. RSF models with different feature counts (10, 15, 20, 25, 30) and parametric AFT models (Weibull, LogNormal, LogLogistic) all using top-variance slope features. RSF-20 hits the best balance: median training C-index of 0.82, test of 0.63, though substantial overfitting remains (gap: 0.19). Parametric models show more stable train-test gaps but lower overall performance.
• Figure 4: Kaplan-Meier survival curves stratified by Random Survival Forest predicted risk scores. Training set (left) and test set (right) show probability of remaining MCI (not converting to AD) over time for three risk groups defined by risk score tertiles. High-risk subjects (red, top tertile) show rapid conversion with median survival 2.1 years. Medium-risk subjects (orange) exhibit intermediate progression rates. Low-risk subjects (green, bottom tertile) show slow conversion with median survival 8.5 years. Shaded regions indicate 95% confidence intervals. Log-rank test confirms significant separation between groups (p=0.3245 for test set, high vs. low comparison). Event counts per group shown in legend demonstrate adequate statistical power for hazard estimation.