Table 2 Results of the Cox regression analysis

From: Baseline [18F]FP-CIT PET-based deep learning prediction of levodopa-induced dyskinesia in Parkinson’s disease

Model Comparison

C-index

Differences in the C-index (95% CI)

(1)

DeepScore (image-only MTL)

0.643 (0.046)

Test 1: (−0.077, 0.030)

Test 2: (−0.012, 0.078)

Test 3: (−0.083, 0.101)

Test 4: (−0.026, 0.095)

Test 5: (−0.044, 0.064)

DeepScore (image-only single-task)

0.623 (0.041)

(2)

DeepScore (image-clinical variable MTL)

0.652 (0.045)

Test 1: (−0.028, 0.099)

Test 2: (−0.109, 0.052)

Test 3: (−0.026, 0.113)

Test 4: (−0.131, 0.088)

Test 5: (−0.036, 0.101)

DeepScore (image-only MTL)

0.643 (0.046)

(3)

DeepScore (image-only single-task)

0.623 (0.041)

Test 1: (0.013, 0.363)

Test 2: (0.034, 0.406)

Test 3: (−0.043, 0.325)

Test 4: (0.111, 0.478)

Test 5: (0.167, 0.471)

Cox regression (SNBRs)

0.392 (0.036)

(4)

DeepScore (image-only MTL)

0.643 (0.046)

Test 1: (0.008, 0.334)

Test 2: (0.071, 0.424)

Test 3: (−0.021, 0.303)

Test 4: (0.151, 0.503)

Test 5: (0.173, 0.484)

Cox regression (SNBRs)

0.392 (0.036)

(5)

DeepScore (image-clinical variable MTL)

0.652 (0.045)

Test 1: (0.091, 0.429)

Test 2: (0.042, 0.431)

Test 3: (0.065, 0.405)

Test 4: (0.109, 0.512)

Test 5: (0.312, 0.626)

Cox regression (SNBRs + clinical)

0.348 (0.053)

  1. C-index concordance index. CI confidence interval, SNBR specific/nonspecific binding ratio.
Back to article page