Skip to main content

Table 2 Mean and std deviation of amplitude errors with 0.4 s and 0.6 s prediction window for 21 liver cancer patients and 10 lung cancer patients

From: Online prediction for respiratory movement compensation: a patient-specific gating control for MRI-guided radiotherapy

PW

Model

Liver organs

Lung tumors

MAE

(mm)

RMSE

(mm)

R2

(None)

MAE

(mm)

RMSE

(mm)

R2

(None)

0.4 s

(j = 2)

Linear

0.60 ± 0.15

1.61 ± 0.64

0.96 ± 0.03

0.86 ± 0.56

1.84 ± 0.84

0.97 ± 0.02

Ridge

0.60 ± 0.15

1.61 ± 0.64

0.96 ± 0.03

0.90 ± 0.60

1.87 ± 0.87

0.97 ± 0.02

L2-L1

1.26 ± 0.33

2.16 ± 0.54

0.94 ± 0.04

1.41 ± 0.38

2.35 ± 0.66

0.96 ± 0.01

LSTM

1.87 ± 0.38

2.94 ± 0.56

0.87 ± 0.12

2.51 ± 1.39

3.93 ± 1.67

0.92 ± 0.03

Bi-LSTM

1.70 ± 0.63

2.85 ± 0.75

0.89 ± 0.11

2.26 ± 0.99

3.57 ± 1.09

0.93 ± 0.02

GRU

1.58 ± 0.46

2.69 ± 0.60

0.90 ± 0.08

2.18 ± 0.92

3.46 ± 1.04

0.94 ± 0.02

KF

2.07 ± 0.48

2.66 ± 0.62

0.95 ± 0.03

2.34 ± 0.83

2.77 ± 0.95

0.93 ± 0.03

0.6 s

(j = 3)

Linear

1.19 ± 0.40

1.55 ± 0.51

0.97 ± 0.02

1.65 ± 0.87

2.11 ± 1.06

0.97 ± 0.03

Ridge

1.2 ± 0.39

1.56 ± 0.50

0.97 ± 0.02

1.69 ± 0.92

2.16 ± 1.12

0.97 ± 0.03

L2-L1

2.23 ± 0.71

2.85 ± 0.90

0.91 ± 0.06

2.52 ± 0.71

3.29 ± 0.96

0.94 ± 0.03

LSTM

1.87 ± 0.38

2.94 ± 0.56

0.87 ± 0.12

3.32 ± 1.40

4.36 ± 1.97

0.91 ± 0.04

Bi-LSTM

1.70 ± 0.63

2.85 ± 0.75

0.89 ± 0.11

3.39 ± 1.64

4.56 ± 2.27

0.89 ± 0.07

GRU

1.58 ± 0.46

2.69 ± 0.60

0.90 ± 0.08

3.50 ± 1.48

4.62 ± 1.98

0.88 ± 0.06

KF

2.55 ± 0.99

3.41 ± 1.21

0.89 ± 0.01

2.89 ± 1.13

3.60 ± 1.32

0.87 ± 0.03

  1. The best performing model is shown in bold for each prediction window
  2. PW prediction window, KF Kalman filter