Table 6 Comparison of the performances between the model and human experts on pre-injection image and all images immediately after each of the three injections.

From: Prediction of neovascular age-related macular degeneration recurrence using optical coherence tomography images with a deep neural network

 

Input

Accuracy

Sensitivity

Specificity

F1score

Precision

F1a

A + B + C + D

0.5500

0.3529

0.7551

0.4444

0.6000

F2a

A + B + C + D

0.5300

0.3333

0.7347

0.4198

0.5667

F3a

A + B + C + D

0.5600

0.4902

0.6327

0.5319

0.5814

R1b

A + B + C + D

0.5100

0.2941

0.7347

0.3797

0.5357

R2b

A + B + C + D

0.5400

0.6078

0.4694

0.5741

0.5439

RSc

A + B + C + D

0.5100

0.3137

0.7143

0.3951

0.5333

Doctor

Average

A + B + C + D

0.5333

0.3987

0.6735

0.4575

0.5602

Model

A + B + C + D

0.6022

0.6275

0.5510

0.6095

0.5926

  1. A, B, C, and D denote the pre-injection image, optical coherence tomography (OCT) image after the first injection, OCT image after the second injection, and OCT image after the third injection, respectively.
  2. aF1, F2, and F3 denote retinal fellows with 2 years of experience as a retina specialist.
  3. bR1 and R2 denote ophthalmology residents with 1 and 3 years of experience, respectively.
  4. cRS refers to retina specialist with more than 10 years of experience.
Back to article page