Skip to main content

Table 3 Diagnosis prediction application of Expert Systems (ES) in Urology

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

Validation

Statistical outcome

[33]

ANN

CaP

Pre-biopsy diagnosis with TRUS variables

Age, PSA, number of biopsies, clinical diagnosis, PSAD, TRUS variables

Cancer or benign

N = 442 from single centre database

ROC AUC NPV, PPV

½ CV

NPV 97%, PPV 82% better than LR

[34]

ANN

CaP

Diagnosis PSA 2.5–4

Age, tPSA, creatinine phospho kinase, prostatic acid phosphatase

Cancer or benign

Multicentre data 522 (PSA 2.5–4)

ROC AUC

CV 152 cases

AUC 0.74

[35]

ANN

CaP

Diagnosis PSA 4–10

Age, tPSA, %fPSA, TPV, DRE

Risk of cancer

656 data from Finnish trial

ROC, Sp, Se LOO

Se 79%, Sp 57%, Ac 62%, PPV 35, NPV 90

[36]

ANN

CaP

Diagnosis PSA 2–20

Age, tPSA, %fPSA, TPV, DRE

Risk of Cancer

1188 multi centre

ROC, Sp, Se, 1/10 CV

Sp 90%, Se 64%

[36]

ANN

CaP

Diagnosis in trial patients with PSA 4–10

Age, tPSA, %fPSA, TPV, DRE

Risk of Cancer

1188 multi centre

ROC, Sp, Se, 204 trial data PSA 4–10

Se 95%, Sp 23.3%, CI 17.4%–30.2%, P < 0.0002

[37]

ANN

CaP

Diagnosis

fPSA, TZD, PSAV, %f PSA, TZV, t PSA, and PSAD

Cancer or benign

PSA 2.5–4, 272 patients, multicentre data

ROC, AUC

¼ CV

AUC 0.88

[37]

ANN

CaP

Diagnosis

TZD, % f PSA, PSAD and TPV

Cancer or benign

PSA 4–10, 974 patients, multicentre data

ROC, AUC

¼ CV

AUC 0.91

[38]

ANN

CaP

Diagnosis after initial negative biopsy PSA 4–10

t PSA, %f PSA, TPV, TZV, PSAD, TZD

Cancer or benign

820 patients with PSA 4–10 European cancer detection studies

ROC AUC

1/3 CV

AUC 0.83

[39]

ANN

CaP

Diagnosis of BPE and CaP

Age, ethnicity, FH, IPSS, t PSA, %f PSA, DRE

Risk of cancer

Multicentre 354 patients, multicentre

ROC vs LR, 144 test set 40% CV

AUC

ANN 0.8, LR 0.5

[40]

FRB

CaP

Early diagnosis

Age, t PSA, TPV

Risk of cancer

Experts aided in developing 77 fuzzy rules

Not published

None

[41]

ANN

CaP

Diagnosis PSA 2–10

Age, tPSA, %fPSA, TPV, TZV, PSAD, TZD = ANNA 1

Cancer and benign

228 data one centre

ROC, 30% CV

AUC 0.78

[41]

ANN

CaP

Diagnosis PSA 2–10

ANNA 1 + presumed circle area ratio and DRE

Cancer and benign

228 data one centre

ROC

30% CV

AUC 0.79, Sp 45%, Se 90%

[42]

ANN

CaP

Diagnosis

Age, tPSA, TPV, PSAD, DRE, and TRUS findings

Cancer and benign

3814 prostate cancer screening data

ROC AUC 1/3 CV, 2 centres prospective data

AUC: 0.74, 0.76, and 0.75 prospective 0.73, 0.74

[43]

ANN

CaP

Diagnosis

Age, DRE, PSA, PSAD, TZV, TZD = ANNA

Cancer and benign

TRUS, single centre 684 data

ROC AUC

1/4 CV

AUC 0.74

[43]

ANN

CaP

Diagnosis

ANNA + TRUS findings

Cancer and benign

TRUS, single centre 684 data

ROC AUC

1/4 CV

AUC 0.86

[44]

FNM

CaP

Diagnosis, PSA < 20

Age, PSA, %f PSA

Cancer and benign

1030 patients’ data, one centre

ROC, Sp, Se, 1/4 CV

AUC 0.8, Sp 52%, Se 90%

[45]

ANN

CaP

Prostate cancer early diagnosis PSA 4–10

Age, tPSA, %fPSA, TPV, DRE

Cancer or benign

606 multicentre group (PSA 4–10)

ROC AUC, 1/10 CV

AUC 0.83, AUC 0.74 in Finish group

[45]

ANN

CaP

Prostate cancer early diagnosis PSA 4–10

Age, tPSA, %fPSA, TPV, DRE

Cancer or benign

656 Finnish cancer survey group (PSA 4–10)

ROC AUC, 1/10 CV

AUC 0.77

[46]

ANN

CaP

Diagnosis

Age, DRE, t PSA and f PSA

Cancer and benign

1509 with PSA < 20, Single centre

ROC AUC, 1/5 CV

AUC 0.74

[46]

ANN

CaP

Diagnosis

Age, DRE, t PSA, f PSA, TPV and TRUS findings

Cancer and benign

1509 with PSA < 20, Single centre

ROC AUC, 1/5 CV

AUC 0.75

[47]

ANN

CaP

Diagnosis with -2 Pro PSA

Age, TPV, tPSA, %fPSA, p2 PSA, %p2 PSA (-2 proPSA)

Cancer and benign

PSA 1–30, 586 one centre

ROC, Sp, Se LOO 586

AUC 0.85, Sp 62%, Se 90%

[48]

ANN

CaP

Diagnosis pre-biopsy

Age, DRE, tPSA, PSAD, TZD, TRUS findings

Benign and malignant

600 patients with suspected CaP

ROC AUC, 477 random

AUC 0.77

[48]

SVM

CaP

Diagnosis pre-biopsy

Age, DRE, tPSA, PSAD, TZD, TRUS findings

Benign and malignant

600 patients with suspected CaP

ROC AUC, 477 random

AUC 0.85

[49]

ANN

CaP

Diagnosis PSA 2–20

Age, tPSA, %f PSA, DRE, TPV

Cancer and benign

Testing Prostataclass

ROC AUC, 165 patients one centre

AUC (PSA 2–10) 63–69%, (PSA 10–20) 57–88%

[50]

ANN

CaP

Diagnosis

Age, tPSA, %f PSA

Prognosis: cancer or not

121 Patients data from one centre

ROC AUC, 30% CV 29 patients

AUC 0.92

[51]

ANN

CaP

Diagnosis of clinically significant cancer

Age, DRE, PSA, PRV, TRUS, Biopsy cores

Disease clinical significance

3025 multicentre data

Accuracy estimation

Ac 57%

[52]

ANN

CaP

Diagnosis of cancer

Age, DRE, PSA, %fPSA, and TPV

Cancer and benign

204 PSA between 4 -10

ROC AUC

AUC 0.72

[53]

ANN

CaP

PHI index and TPV in diagnosis

Age, TPV, %fPSA, tPSA, PHI, %P2PSA

Cancer and benign

220 cases

PSA < 10

ROC AUC

AUC 0.81

[53]

ANN

CaP

PHI index and TPV in diagnosis

Age, %fPSA, tPSA, PHI, %P2PSA

Cancer and benign

221 cases

PSA < 10

ROC AUC

AUC 0.77

[54]

FRB

CaP

Diagnosis

Age, PSA, TPV

Cancer and benign

78 TRUS cases from Urology clinic

None

None

[55]

ANN

Fert

Sperm count

Age, duration of infertility, FSH, LH, TT and PRL, testicular volume

Presence of spermatozoa

303 patient’s data

ROC AUC then kappa stats of LR, test set 73 random

Se 68%, Sp 87.5%, PPV 73.9%, NPV 84%

[56]

ANN

Fert

Endocrinopathy with low sperm count

Testis volume, total sperm count,

Endocrinopathy

1035 Data from 2 centres

ROC AUC

AUC 0.95

[57]

ANN

Fert

Microdissection testicular sperm extraction

Age, FSH level, cryptorchidism and Klinefelter

Syndrome

Sperm retrieval

1026 data, one centre

ROC AUC

Se 67% Sp 49.5% PPV, 63.9% NPV 52% Ac 60.8%

[58]

ANN

U Dys

Interpretation of U dyn and symptoms

Neurological and physical symptoms, flowmetry, cystometry, U dyn

Areflexia, hyper-reflexive, effort incontinence

400 U Dyn data

80 patients, 1/5 CV,

Accuracy 85%

[59]

ANN

U Dys

Interpretation of U dyn and symptoms

Neurological and physical symptoms, flowmetry, cystometry, U dynamics

Healthy or ill

300 patients with LUT disease

ROC, Ac, 1/5 CV

Accuracy 89%

[60]

ANN

U Dys

Bladder outlet obstruction

values of the average flow rate, Qmax, PVR and TPV

Obstructed, non-obstructed, and equivocal

N = 457 cases from single centre

Accuracy estimation 157 cases

Ac 60% (testing) 75% (training)

[61]

ANN

BPD

IPSS interpretation

IPSS subdomain scores

Obstructed, non-obstructed, and equivocal

N = 460 from single centre

Accuracy estimation 157 cases

Ac 73%

  1. A total of 37 systems identified in this application of Expert Systems in Urology with evident prevalence of ANN as the model and CaP to be the dominant domain