Skip to main content

Table 2 Number LP images of training and test data from ALLSTAR database

From: Optimal length of R–R interval segment window for Lorenz plot detection of paroxysmal atrial fibrillation by machine learning

Segment window length, beat

LP images of training data, n (%)

LP images of test data, n (%)a

Non-AF

AF

Non-AF

AF (pure)b

AF (mixed)c

10

590,602 (50.7)

574,014 (49.3)

934,803 (85.5)

156,295 (14.3)

2543 (0.2)

20

295,284 (50.7)

286,996 (49.3)

467,009 (85.4)

77,426 (14.2)

2365 (0.4)

50

118,099 (50.7)

114,781 (49.3)

186,432 (85.2)

30,265 (13.8)

2000 (0.9)

100

59,035 (50.7)

57,373 (49.3)

92,967 (85.0)

14,684 (13.4)

1677 (1.5)

200

29,506 (50.7)

28,670 (49.3)

46,268 (84.7)

7033 (12.9)

1331 (2.4)

500

11,783 (50.7)

11,453 (49.3)

18,314 (83.9)

2633 (12.1)

877 (4.0)

  1. LP Lorenz plot
  2. aLP images were annotated with the strict criteria (LP images were annotated as AF if they contained at least one AF beat, and non-AF otherwise)
  3. bLP images consisting of only AF beats (including premature beats)
  4. cLP images composed of both SR and AF beats (including premature beats)