Using Bayesian neural networks with ARD input selection to detect malignant ovarian masses prior to surgery

Ben Van Calster; Dirk Timmerman; Ian T. Nabney; Lil Valentin; Antonia C. Testa; Caroline Van Holsbeke; Ignace Vergote; Sabine Van Huffel

doi:10.1007/s00521-007-0147-1

Using Bayesian neural networks with ARD input selection to detect malignant ovarian masses prior to surgery

Ben Van Calster, Dirk Timmerman, Ian T. Nabney, Lil Valentin, Antonia C. Testa, Caroline Van Holsbeke, Ignace Vergote, Sabine Van Huffel

Obstetric, Gynaecological and Prenatal Ultrasound Research

Research output: Chapter in Book/Report/Conference proceeding › Paper in conference proceeding › peer-review

Abstract

In this paper, we applied Bayesian multi-layer perceptrons (MLP) using the evidence procedure to predict malignancy of ovarian masses in a large (n = 1,066) multi-centre data set. Automatic relevance determination (ARD) was used to select the most relevant inputs. Fivefold cross-validation (5CV) and repeated 5CV was used to select the optimal combination of input set and number of hidden neurons. Results indicate good performance of the models with area under the receiver operating characteristic curve values of 0.93-0.94 on independent data. Comparison with a linear benchmark model and a previously developed logistic regression model shows that the present problem is very well linearly separable. A resampling analysis further shows that the number of hidden neurons specified in the ARD analyses for input selection may influence model performance. This paper shows that Bayesian MLPs, although not frequently used, are a useful tool for detecting malignant ovarian tumours.

Original language	English
Title of host publication	NEURAL COMPUTING & APPLICATIONS
Publisher	Springer
Pages	489-500
Volume	17
DOIs	https://doi.org/10.1007/s00521-007-0147-1
Publication status	Published - 2008
Event	28th Annual International Conference of the IEEE-Engineering-in-Medicine-and-Biology-Society - New York, NY Duration: 2006 Aug 30 → 2006 Sep 3

Publication series

Name
Number	5-6
Volume	17
ISSN (Print)	0941-0643

Conference

Conference	28th Annual International Conference of the IEEE-Engineering-in-Medicine-and-Biology-Society
Period	2006/08/30 → 2006/09/03

Subject classification (UKÄ)

Obstetrics, Gynecology and Reproductive Medicine

Keywords

ultrasound
automatic relevance determination
netlab
evidence framework
Bayesian
ovarian tumour classification
multi-layer perceptrons

Access to Document

10.1007/s00521-007-0147-1

Cite this

@inproceedings{2e8b93df22074bd18873fbbe9de5fa43,

title = "Using Bayesian neural networks with ARD input selection to detect malignant ovarian masses prior to surgery",

abstract = "In this paper, we applied Bayesian multi-layer perceptrons (MLP) using the evidence procedure to predict malignancy of ovarian masses in a large (n = 1,066) multi-centre data set. Automatic relevance determination (ARD) was used to select the most relevant inputs. Fivefold cross-validation (5CV) and repeated 5CV was used to select the optimal combination of input set and number of hidden neurons. Results indicate good performance of the models with area under the receiver operating characteristic curve values of 0.93-0.94 on independent data. Comparison with a linear benchmark model and a previously developed logistic regression model shows that the present problem is very well linearly separable. A resampling analysis further shows that the number of hidden neurons specified in the ARD analyses for input selection may influence model performance. This paper shows that Bayesian MLPs, although not frequently used, are a useful tool for detecting malignant ovarian tumours.",

keywords = "ultrasound, automatic relevance determination, netlab, evidence framework, Bayesian, ovarian tumour classification, multi-layer perceptrons",

author = "{Van Calster}, Ben and Dirk Timmerman and Nabney, {Ian T.} and Lil Valentin and Testa, {Antonia C.} and {Van Holsbeke}, Caroline and Ignace Vergote and {Van Huffel}, Sabine",

year = "2008",

doi = "10.1007/s00521-007-0147-1",

language = "English",

volume = "17",

publisher = "Springer",

number = "5-6",

pages = "489--500",

booktitle = "NEURAL COMPUTING & APPLICATIONS",

address = "Germany",

note = "28th Annual International Conference of the IEEE-Engineering-in-Medicine-and-Biology-Society ; Conference date: 30-08-2006 Through 03-09-2006",

}

Van Calster, B, Timmerman, D, Nabney, IT, Valentin, L, Testa, AC, Van Holsbeke, C, Vergote, I & Van Huffel, S 2008, Using Bayesian neural networks with ARD input selection to detect malignant ovarian masses prior to surgery. in NEURAL COMPUTING & APPLICATIONS. vol. 17, Springer, pp. 489-500, 28th Annual International Conference of the IEEE-Engineering-in-Medicine-and-Biology-Society, 2006/08/30. https://doi.org/10.1007/s00521-007-0147-1

TY - GEN

T1 - Using Bayesian neural networks with ARD input selection to detect malignant ovarian masses prior to surgery

AU - Van Calster, Ben

AU - Timmerman, Dirk

AU - Nabney, Ian T.

AU - Valentin, Lil

AU - Testa, Antonia C.

AU - Van Holsbeke, Caroline

AU - Vergote, Ignace

AU - Van Huffel, Sabine

PY - 2008

Y1 - 2008

N2 - In this paper, we applied Bayesian multi-layer perceptrons (MLP) using the evidence procedure to predict malignancy of ovarian masses in a large (n = 1,066) multi-centre data set. Automatic relevance determination (ARD) was used to select the most relevant inputs. Fivefold cross-validation (5CV) and repeated 5CV was used to select the optimal combination of input set and number of hidden neurons. Results indicate good performance of the models with area under the receiver operating characteristic curve values of 0.93-0.94 on independent data. Comparison with a linear benchmark model and a previously developed logistic regression model shows that the present problem is very well linearly separable. A resampling analysis further shows that the number of hidden neurons specified in the ARD analyses for input selection may influence model performance. This paper shows that Bayesian MLPs, although not frequently used, are a useful tool for detecting malignant ovarian tumours.

AB - In this paper, we applied Bayesian multi-layer perceptrons (MLP) using the evidence procedure to predict malignancy of ovarian masses in a large (n = 1,066) multi-centre data set. Automatic relevance determination (ARD) was used to select the most relevant inputs. Fivefold cross-validation (5CV) and repeated 5CV was used to select the optimal combination of input set and number of hidden neurons. Results indicate good performance of the models with area under the receiver operating characteristic curve values of 0.93-0.94 on independent data. Comparison with a linear benchmark model and a previously developed logistic regression model shows that the present problem is very well linearly separable. A resampling analysis further shows that the number of hidden neurons specified in the ARD analyses for input selection may influence model performance. This paper shows that Bayesian MLPs, although not frequently used, are a useful tool for detecting malignant ovarian tumours.

KW - ultrasound

KW - automatic relevance determination

KW - netlab

KW - evidence framework

KW - Bayesian

KW - ovarian tumour classification

KW - multi-layer perceptrons

U2 - 10.1007/s00521-007-0147-1

DO - 10.1007/s00521-007-0147-1

M3 - Paper in conference proceeding

VL - 17

SP - 489

EP - 500

BT - NEURAL COMPUTING & APPLICATIONS

PB - Springer

T2 - 28th Annual International Conference of the IEEE-Engineering-in-Medicine-and-Biology-Society

Y2 - 30 August 2006 through 3 September 2006

ER -

Using Bayesian neural networks with ARD input selection to detect malignant ovarian masses prior to surgery

Abstract

Publication series

Conference

Subject classification (UKÄ)

Keywords

Access to Document

Fingerprint

Cite this