Deep attention networks for the prediction of pouch of Douglas obliteration from transvaginal ultrasound sliding sign videos

Standard

Deep attention networks for the prediction of pouch of Douglas obliteration from transvaginal ultrasound sliding sign videos. / O'Shea, K.; Leonardi, M.; Bolton, R. A.; Condous, G.; Lu, C.

In: Ultrasound in Obstetrics and Gynecology, Vol. 54, No. S1, EP34.29, 30.09.2019, p. 448-448.

Research output: Contribution to journalMeeting abstract

Harvard

Author

O'Shea, K. ; Leonardi, M. ; Bolton, R. A. ; Condous, G. ; Lu, C. / Deep attention networks for the prediction of pouch of Douglas obliteration from transvaginal ultrasound sliding sign videos. In: Ultrasound in Obstetrics and Gynecology. 2019 ; Vol. 54, No. S1. pp. 448-448.

Bibtex - Download

@article{afb805397906456393e6eed3079f7abc,
title = "Deep attention networks for the prediction of pouch of Douglas obliteration from transvaginal ultrasound sliding sign videos",
abstract = "ObjectivesRecent studies have demonstrated that the transvaginal sonography sliding sign technique is an accurate method of determining whether the pouch of Douglas (POD) is obliterated in women with suspected endometriosis. In this work we extend earlier work in which state of the art machine learning approaches to image recognition, particularly deep neural networks, were used for automated interpretation of the {\textquoteleft}sliding sign{\textquoteright} videos for automated detection of POD obliteration.MethodsThe data set used for model development consists of 88 ultrasound recordings from different sonographic machines to determine POD obliteration of women presenting with chronic pelvic pain, using the dynamic real‐time {\textquoteleft}sliding sign{\textquoteright} technique. Long term recurrent attention‐based convolutional neural networks were developed for this video classification task. The trained neural networks would take a sequence of frames, and extract spatial‐temporal features hierarchically through multiple layers of information processing, and output scores for the whether the video is interpreted as negative or positive sliding sign.ResultsOur best performing model achieved an accuracy of 80.3%, sensitivity 78.7%, and specificity of 74.7% in hold‐one‐out cross‐validation for predicting POD obliteration.ConclusionsIn spite of the limited dataset, we have demonstrated the potential of using deep attention‐based neural networks for the preoperative prediction of POD obliteration from a number of {\textquoteleft}sliding‐sign{\textquoteright} studies. Further work is needed to improve and evaluate the models using larger datasets.",
author = "K. O'Shea and M. Leonardi and Bolton, {R. A.} and G. Condous and C. Lu",
year = "2019",
month = sep,
day = "30",
doi = "10.1002/uog.21819",
language = "English",
volume = "54",
pages = "448--448",
journal = "Ultrasound in Obstetrics and Gynecology",
issn = "0960-7692",
publisher = "Wiley",
number = "S1",
note = "29th World Congress on Ultrasound in Obstetrics and Gynecology ; Conference date: 12-10-2019 Through 16-10-2019",

}

RIS (suitable for import to EndNote) - Download

TY - JOUR

T1 - Deep attention networks for the prediction of pouch of Douglas obliteration from transvaginal ultrasound sliding sign videos

AU - O'Shea, K.

AU - Leonardi, M.

AU - Bolton, R. A.

AU - Condous, G.

AU - Lu, C.

PY - 2019/9/30

Y1 - 2019/9/30

N2 - ObjectivesRecent studies have demonstrated that the transvaginal sonography sliding sign technique is an accurate method of determining whether the pouch of Douglas (POD) is obliterated in women with suspected endometriosis. In this work we extend earlier work in which state of the art machine learning approaches to image recognition, particularly deep neural networks, were used for automated interpretation of the ‘sliding sign’ videos for automated detection of POD obliteration.MethodsThe data set used for model development consists of 88 ultrasound recordings from different sonographic machines to determine POD obliteration of women presenting with chronic pelvic pain, using the dynamic real‐time ‘sliding sign’ technique. Long term recurrent attention‐based convolutional neural networks were developed for this video classification task. The trained neural networks would take a sequence of frames, and extract spatial‐temporal features hierarchically through multiple layers of information processing, and output scores for the whether the video is interpreted as negative or positive sliding sign.ResultsOur best performing model achieved an accuracy of 80.3%, sensitivity 78.7%, and specificity of 74.7% in hold‐one‐out cross‐validation for predicting POD obliteration.ConclusionsIn spite of the limited dataset, we have demonstrated the potential of using deep attention‐based neural networks for the preoperative prediction of POD obliteration from a number of ‘sliding‐sign’ studies. Further work is needed to improve and evaluate the models using larger datasets.

AB - ObjectivesRecent studies have demonstrated that the transvaginal sonography sliding sign technique is an accurate method of determining whether the pouch of Douglas (POD) is obliterated in women with suspected endometriosis. In this work we extend earlier work in which state of the art machine learning approaches to image recognition, particularly deep neural networks, were used for automated interpretation of the ‘sliding sign’ videos for automated detection of POD obliteration.MethodsThe data set used for model development consists of 88 ultrasound recordings from different sonographic machines to determine POD obliteration of women presenting with chronic pelvic pain, using the dynamic real‐time ‘sliding sign’ technique. Long term recurrent attention‐based convolutional neural networks were developed for this video classification task. The trained neural networks would take a sequence of frames, and extract spatial‐temporal features hierarchically through multiple layers of information processing, and output scores for the whether the video is interpreted as negative or positive sliding sign.ResultsOur best performing model achieved an accuracy of 80.3%, sensitivity 78.7%, and specificity of 74.7% in hold‐one‐out cross‐validation for predicting POD obliteration.ConclusionsIn spite of the limited dataset, we have demonstrated the potential of using deep attention‐based neural networks for the preoperative prediction of POD obliteration from a number of ‘sliding‐sign’ studies. Further work is needed to improve and evaluate the models using larger datasets.

U2 - 10.1002/uog.21819

DO - 10.1002/uog.21819

M3 - Meeting abstract

VL - 54

SP - 448

EP - 448

JO - Ultrasound in Obstetrics and Gynecology

JF - Ultrasound in Obstetrics and Gynecology

SN - 0960-7692

IS - S1

M1 - EP34.29

T2 - 29th World Congress on Ultrasound in Obstetrics and Gynecology

Y2 - 12 October 2019 through 16 October 2019

ER -

View graph of relations
Citation formats