Skip to main content

Table 2 Performance of the Alzheimer’s disease classifier

From: A practical Alzheimer’s disease classifier via brain imaging-based deep learning on 85,721 samples

 

Accuracy

AUC

Sensitivity

Specificity

Input Feature

ADNI

AIBL

MIRIAD

OASIS

ADNI

AIBL

MIRIAD

OASIS

ADNI

AIBL

MIRIAD

OASIS

ADNI

AIBL

MIRIAD

OASIS

Sex classifier as base model

                

GMV + GMD

0.909

[0.902, 0.915]

0.945

[0.926, 0.965]

0.936

[0.917, 0.955]

0.911

[0.895, 0.928]

0.963

[0.958, 0.967]

0.966

[0.945, 0.987]

0.994

[0.991, 0.997]

0.976

[0.966, 0.985]

0.838

[0.824, 0.854]

0.881

[0.816, 0.943]

0.897

[0.869, 0.928]

0.932

[0.889, 0.972]

0.942

[0.935, 0.948]

0.958

[0.941, 0.974]

1.000

[1.000, 1.000]

0.908

[0.890, 0.926]

GMD

0.879

[0.872, 0.887]

0.941

[0.923, 0.958]

0.944

[0.926, 0.962]

0.906

[0.889, 0.924]

0.940

[0.935, 0.946]

0.952

[0.926, 0.978]

0.990

[0.985, 0.995]

0.946

[0.927, 0.964]

0.762

[0.744, 0.782]

0.803

[0.732, 0.875]

0.959

[0.939, 0.980]

0.793

[0.722, 0.862]

0.933

[0.926, 0.941]

0.968

[0.953, 0.981]

0.920

[0.886, 0.955]

0.922

[0.905, 0.940]

GMV

0.903

[0.895, 0.910]

0.943

[0.926, 0.962]

0.916

[0.894, 0.936]

0.862

[0.841, 0.880]

0.960

[0.955, 0.964]

0.964

[0.940, 0.986]

0.991

[0.986, 0.995]

0.949

[0.932, 0.965]

0.829

[0.812, 0.844]

0.857

[0.783, 0.923]

0.894

[0.865, 0.922]

0.866

[0.806, 0.924]

0.937

[0.930, 0.944]

0.960

[0.943, 0.979]

0.953

[0.927, 0.983]

0.861

[0.840, 0.882]

T1-weighted

0.866

[0.857, 0.874]

0.947

[0.929, 0.965]

0.937

[0.917, 0.955]

0.730

[0.701, 0.757]

0.928

[0.922, 0.934]

0.961

[0.939, 0.983]

0.991

[0.987, 0.995]

0.940

[0.920, 0.963]

0.766

[0.748, 0.785]

0.820

[0.745, 0.893]

0.935

[0.911, 0.959]

0.918

[0.871, 0.969]

0.912

[0.904, 0.920]

0.971

[0.958, 0.985]

0.940

[0.911, 0.972]

0.704

[0.674, 0.735]

Age predictor as base model

                

GMV + GMD

0.901

[0.894, 0.908]

0.950

[0.933, 0.968]

0.942

[0.923, 0.958]

0.892

[0.875, 0.909]

0.957

[0.953, 0.962]

0.965

[0.941, 0.988]

0.994

[0.991, 0.997]

0.965

[0.952, 0.979]

0.812

[0.795, 0.829]

0.881

[0.815, 0.946]

0.907

[0.879, 0.935]

0.866

[0.806, 0.928]

0.943

[0.936, 0.950]

0.964

[0.947, 0.980]

1.000

[1.000, 1.000]

0.896

[0.877, 0.914]

  1. The ADNI dataset contained 2,186 AD samples and 4,671 NC samples. The AIBL dataset contained 115 AD samples and 554 NC samples. The MIRIAD dataset contained 409 AD samples and 235 NC samples. The OASIS dataset contained 137 AD samples and 986 NC samples. The sample sizes shown here are the numbers of T1-weighted brain MRI scans. Abbreviations: AD = Alzheimer’s disease participants, NC = normal control participants, GMD = grey matter density map, GMV = grey matter volume map, T1-weighted = normalized 3D T1-weighted structural image, AUC = area under the curve. The 95% confidence intervals are listed in brackets