Skip to main content

Table 6 Recurrence and progression 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

Knowledge acquisition

Validation

Statistical outcome

[83, 100]

ANN

Bca

Recurrence

Age, gender, smoking, tumour stage and grade, CIS, number, cytology, other mucosal biopsy

Recurrence or no

N = 432 patients’ data, multicentre

Radom set of 200

ROC AUC

Se 76%, Sp 55%, Ac 72%

[101]

ANN

Bca

Tumour progression recurrence

Tumour stage and grade, size, number, gender, eGFR

Stage progression

105 Ta/T1 TCC multicentre

Compare to 4 clinicians McNemar test

80% accuracy

[101]

ANN

Bca

12 months cancer specific survival

Tumour stage and grade, size, number, gender, eGFR, smoking, cis, dysplasia tumour site, architecture, c-erbB2 (oncogene), p53 (tumour suppressor gene)

6 months recurrence 12 months survival

56 Ta/T1 (6 months recurrence), 40 T2-T4 (12 months survival)

Compare to 4 clinicians McNemar test

Accuracy to predict recurrence (75%) and to predict survival (82%)

[102]

ANN

Bca

Progression of non-invasive TCC

Age, gender, tumour (grade, stage, number and architecture) and mean nuclear volume

Tumour progression and recurrence

68 patients’ specimen from one centre

22 Random test set

ROC, Sp, Se

Recurrence: Se 33%, Sp 40%, PPV 40%, NPV 33%

Progression: Se 100%, Sp 67%, PPV 40%, NPV 100%

[103]

FNM

Bca

Recurrence classifier

Age, gender, grade, smoking, previous cancer, p53, hMLH1, hMSH2

Recurrence or not

109 patients from one centre with TCC

10% cross validation ROC, LR

AUC 0.98, Se 90%, Sp 80%, PPV 92%, NPV 74%, Ac 88%

[103]

FNM

Bca

Survival predictor

Age, gender, grade, smoking, previous cancer, p53, hMLH1, hMSH2

Survival in months

109 patients from one centre with TCC

10% cross validation

Root mean square

RMS = 4.8

[103]

ANN

Bca

Recurrence classifier

Age, gender, grade, smoking, previous cancer, p53, hMLH1, hMSH2

Recurrence or not

109 patients from one centre with TCC

ROC, LR

10% cross validation

AUC 0.91, Se 94%, Sp 96%, PPV 99%, NPV 84%, Ac 95%

[103]

ANN

Bca

Survival predictor

Age, gender, grade, smoking, previous cancer, p53, hMLH1, hMSH2

Survival in months

109 patients from one centre with bladder

10% cross validation RMS

RMS = 11.7

[104]

ANN

Bca

Survival predictor

Age, stage, Grade, smoking, previous cancer

Risk of relapse

109 patients with primary TCC

Difference in RMS 1/4 CV ROC AUC

Se 90%, Sp 89%, PPV 98, NPV, 64%, Ac 90%, RMS 8.8

[104]

ANN

Bca

Recurrence predictor

Stage, Grade, age, smoking, previous cancer, p53, hMLH1, hMLH2

Time to relapse

109 patients with primary TCC

Difference in RMS 1/4 CV ROC AUC

Se 94, Sp 96%, NPV 99%,PPV 84%, Ac 95%, RMS 7.6

[104]

FNM

Bca

Survival predictor

Stage, Grade, age, smoking, previous cancer

Risk of relapse

109 patients with primary TCC

Difference in RMS 1/4 CV ROC AUC

Se 92%, Sp 90%, PPV 98% NPV 72%, Ac 92%, RMS 8.5

[104]

FNM

Bca

Recurrence predictor

Stage, Grade, age, smoking, previous cancer, p53, hMLH1, hMLH2

Time to relapse

109 patients with primary TCC

Difference in RMS 1/4 CV ROC AUC

Se 90% Sp 80%, NPV 92%,PPV 74%, Ac 88%, RMS 7.3

[105]

FNM

Bca

Recurrence (classifier)

Age, gender, grade, smoking, previous cancer, p53, methylation index (% of loci on chromosomes), RARB methylation (gene locus)

Recurrence or not

117 patients with 1ry TCC or UCC from one centre

10% cross validation ROC, LR

AUC 0.98, Se 88–100%, Sp 94–100%, Ac 100%

[105]

FNM

Bca

Survival predictor

Age, gender, grade, smoking, previous cancer, p53, methylation index (% of loci on chromosomes), RARB methylation

Survival in months

117 patients with 1ry TCC or UCC from one centre

10% CV

Kaplan Maier for survival

Average error = 5 months

[105]

ANN

Bca

Recurrence (classifier)

Age, gender, grade, smoking, previous cancer, p53, methylation index (% of loci on chromosomes), RARB methylation

Recurrence or not

117 patients with 1ry TCC or UCC from one centre

10% cross validation ROC, LR

Ac 89–90%, Se 81–87%, Sp 95–100%

[105]

ANN

Bca

Survival predictor

Age, gender, grade, smoking, previous cancer, p53, methylation index (% of loci on chromosomes), RARB methylation

Survival in months

117 patients with 1ry TCC or UCC from one centre

10% CV

Kaplan Maier for survival

Average error = 9 months

[106]

ANN

Bca

Recurrence

Age, sex, previous recurrence, response to adjuvant therapy, number of lesions, adjuvant therapy

Recurrence or no

403 patients

1/3 CV

123 patients ROC AUC

AUC 0.87,Se 79%, Sp 98%

[107]

ANN

Bca

5 Years survival cystectomy

Age, gender, tumour stage, grade, ln, vascular in, perineural in, prostatic invasion, CaP

Survival at 5 years

369 patients

ROC, Cox proportional hazard 1/3 CV

Se 63%, Sp 86%, PPV 76%, NPV 77%

[108]

FNM

Bca

Recurrence classifier

Gender, pathological stage, grade, CIS, lymph vascular invasion

Recurrence or not

609 patients from multiple centres

ROC, LR

10% CV

Se 93%, Sp 68%

[108]

FNM

Bca

Survival predictor

Gender, pathological stage, grade, CIS, lymph vascular invasion

Survival in months

172 multicentre data

ROC, LR

10% CV

Kaplan–Meier survival plots, median error of 8.15 months

[109]

ANN

Bca

Survival post cystectomy

Age, gender, bilhariziasis, histology, grade, lymph nodes, lymph vascular, type of diversion

Patient survival

871 patients’ data post cystectomy

30% CV

ROC vs LR

AUC 0.86, Se 79%, Sp 81%

[110]

ANN

Bca

bladder cancer 5 years survival

Age, gender, histology grade, tumour stage, positive LN, removed LN

5 years survival

cystectomy data base, single centre 106 patients

Prediction error percent 11 and 29 patients

prediction error rate, > 90% efficiency

[111]

ANN

Bca

Recurrence and survival

Age, gender, tumour stage, grade, CIS, ln, lymph vascular invasion

5 years recurrence and cancer specific death

cystectomy data base, multicentre 2111

ROC, Kaplan Maier for survival, Cox Proportional Hazard

Se 59%, Sp 77%, PPV 67%, NPV 70% (30% cross validation)

[112]

ANN

Bca

Survival post cystectomy

Age, gender, albumin, surgical approach, tumour stage, follow up period, type of diversion

5 years survival

117 patients with post cystectomy from one centre

10 Folds CV

ROC, Se, Sp Ac

Ac 72–80% RELM and ELM had best performance

[113]

ANN

Bca

Recurrence of G3 pTa after TURBT

Age, sex, previous histopathological data, previous recurrence rate response to previous BCG adjuvant therapy, number of lesions, size of lesions presence of inflammatory reaction and adjuvant therapy

Recurrence or No

143 patients with G3 pTa at one centre

AUC, Se, Se

1/3 cv 43 cases

AUC 0.81, Se 82%, Sp 96%

[114]

ANN

RCC

RCC survival 36 months

Age, gender, BMI, performance status, histopathology, time interval between primary tumour and detection of Mets, type of systemic therapy, number and sites of Met

Recurrence within 36 months

175 single centre

30% CV

ROC sensitivity analysis

AUC 0.95

(95% CI 0.87–0.98)

[115]

ANN

Nsc

Disease recurrence in five years

(32 variables) age, tumour type, grade, invasion, Mets, ln, treatment, FBC, kidney function

Recurrence within five years

202 multicentre cases

1/4 CV

ROC, Sensitivity analysis

AUC 0.87

[116]

FNM

CaP

Prognosis and survival

Age, BMI, PSA, DRE, Gl, clinical stage and treatment methods

Disease specific survival

100 cases single centre

Sensitivity analysis mean square error,

MSE = 0.068907 (1/10 CV)

[117]

ANN

Nlt

Recurrence of Upper tract stones

Age, sex, history of previous calculi, radiologic type, location and composition of previous calculi, 24-h urine assay urine culture, treatment

Recurrence of Upper tract stones

168 cases, single centre

PPV estimation 68 cases

PPV 100%

  1. The majority of the Expert Systems in this application were artificial neural network predicting recurrence and survival following bladder cancer treatment. Other systems were applied in non-seminomatous testicular cancer, prostate cancer, renal cell carcinoma and recurrence of upper renal tract stones