Skip to main content

Table 4 Disease stage prediction

From: A systematic review of the applications of Expert Systems (ES) and machine learning (ML) in clinical urology

Art

Mdl

Dom

Subdomain

Variables

Output

System training

Statistical outcome

Validation set

[62]

ANN

CaP

staging of localised disease

Age, race, DRE, tPSA, size of tumour on ultrasound, Gl, bilaterality of cancer and number of positive cores and perineural infiltration

Margin, seminal vesicle and lymph node positivity

1200, patients’ data from multicentre

AUC 0.77, 0.79, 0.8

20% CV

[63]

FSS

CaP

Localised disease staging

Age, PSA, PSAD, DRE, TRUS, Gl, CT, bone scan, chest x-ray, MRI

Localised or advanced

16 Cases

Se 92%, Sp 84%, Ac 82%

43 cases RRP

[64]

ANN

CaP

Lymph node staging in CaP post RPP

Age, Gl, clinical stage

Lymph node spread

736 data from one centre clinically localised CaP

Se 64%, Sp 81.5%, PPV 14%, NPV 98%

1840 and 316 cases from 2 centres

[65]

ANN

CaP

Prostate cancer staging post RRP

Age, tPSA, Gl, clinical stage

Lymph node spread or organ confinement

5744 data from one centre clinically localised CaP

AUC 77%, 88% for LN

25% CV random

[66]

ANN

CaP

Stage prediction post RRP

Age, histological variables from biopsy

CaP stage

97 cases with non-organ confined

Prediction accuracy ranged from 82 to 90%

 

[66]

ANN

CaP

Stage prediction post RRP

Age, histological variables from biopsy, tPSA and TPV

CaP stage

77 cases with non-organ confined and extracapsular spread

Prediction accuracy ranged from 82 to 90%

 

[67]

ANN

CaP

Prostate cancer staging post RRP PSA 2–10

tPSA, TNM, Gl (ANNA1)

localised disease

124 data from 2 centres Clinically localised CaP

AUC 0.82

20% (n = 36 patients)

[67]

ANN

CaP

Prostate cancer staging post RRP PSA 2–10

tPSA, TNM, Gl, maximum tumour length (ANNA2)

localised disease

124 data from 2 centres Clinically localised CaP

AUC 0.88

20% (n = 36 patients)

[67]

ANN

CaP

Prostate cancer staging post RRP PSA 2–10

tPSA, TNM, Gl, maximum tumour length, PSAD (ANNA3)

localised disease

124 data

2 centres Clinically localised CaP

Ac 83.3%, Se 85%, Sp 83%, PPV 73%, NPV 90% AUC 0.9

20% (n = 36 patients)

[67]

ANN

CaP

Prostate cancer staging post RRP PSA 2–10

tPSA, TNM, Gl, maximum tumour length PSAD, age (ANNA4)

localised disease

124 data

2 centres Clinically localised CaP

AUC 0.87

20% 36 patients

[68]

ANN

CaP

Prostate cancer staging post RRP

tPSA, TPV, TZV, PSAD, TZ, Gl

Pathological stage t2-4

201 cases from multinational European cancer data base (PSA 10 or less)

AUC 0.87

61 prospective set

[69]

ANN

CaP

diagnosis of skeletal metastasis

Age, tPSA

skeletal Mets

111 retrospective cases in one centre

AUC 0.88, Se 87.5%, Sp 83.3%

Bootstrap CV

[70]

ANN

CaP

Stage prediction post RRP

DRE, % of cancer, sum of tumour length, % cancer length and maximum cancer core length

advanced cancer (> pT3a)

300 randomly selected from retrospective data

AUC 0.71, Se 63%, Sp 81%, Ac78%

232 random selected set

[70]

SVM

CaP

Stage prediction post RRP

DRE, % of cancer, sum of tumour length, % cancer length and maximum cancer core length

advanced caner (> pT3a)

300 randomly selected from retrospective data

AUC 0.81, Se 67%, Sp 79%, Ac77%

232 random selected set

[71]

ANN

CaP

Define precise stage

PSA, clinical stage, pathological stage, Gl (other added for different set: erection, IPSS, TRUS size, MRI stage

margin, seminal vesicle and lymph node positivity

From 7500 patients’ data from BAUS database and remodelled with external data of 85 patients

AUC 0.38–0.67, concordance index for variables

10 folds CV

[71]

BN

CaP

Define precise stage

PSA, clinical stage, pathological stage, Gl (other added for different set: erection, IPSS, TRUS size, MRI stage

margin, seminal vesicle and lymph node positivity

From 7500 patients’ data from BAUS database and remodelled with external data of 85 patients

AUC 0.01–0.67 concordance index for variables

10 folds CV

[71]

kNN

CaP

Define precise stage

PSA, clinical stage, pathological stage, Gl (other added for different set: erection, IPSS, TRUS size, MRI stage

margin, seminal vesicle and lymph node positivity

From 7500 patients’ data from BAUS database and remodelled with external data of 85 patients

AUC 0.33–0.6 concordance index for variables

10 folds CV

[71]

RBF

CaP

Define precise stage

PSA, clinical stage, pathological stage, Gl (other added for different set: erection, IPSS, TRUS size, MRI stage

margin, seminal vesicle and lymph node positivity

From 7500 patients’ data from BAUS database and remodelled with external data of 85 patients

AUC 0.45–0.5 concordance index for variables

10 folds CV

[71]

SVM

CaP

Define precise stage

PSA, clinical stage, pathological stage, Gl (other added for different set: erection, IPSS, TRUS size, MRI stage

margin, seminal vesicle and lymph node positivity

From 7500 patients’ data from BAUS database and remodelled with external data of 85 patients

AUC 0.5 concordance index for variables

10 folds CV

[72]

ANN

CaP

Staging post RRP

Age, tPSA, n Positive cores, involvement per core, % of positive core

Organ confinement and metastasis

870 multicentre data

Ac 60%

120 cases, Accuracy estimation

[73]

FNM

CaP

Cancer staging of organ confinement

Age, PSA, Primary Gleason Pattern, secondary Gleason pattern, clinical stage

Organ confinement and metastasis

399 cases from research network database

AUC 0.8, FNM outperformed ANN, FCM, LR

ROC AUC vs other models

[74]

ANN

Nsc

staging

vascular, lymphatic, tunical invasion, percentage of embryonal carcinoma, yolk sac carcinoma, teratoma and seminoma

Stage one or two

93 cancer specimen, single centre

Prediction accuracy 79.6 to 87.1%,

10 folds CV

  1. This table demonstrated Expert Systems predicting urological diagnosis from variable clinical and radiological date. Artificial neural networks (ANN) diagnosing localised prostate cancer (CaP) before surgery were the most common systems in this application