Machine learning outperforms clinical experts in classification of hip fractures

Ellen Murphy; Beate Ehrhardt; Celia Gregson; Otto Von Arx; April Hartley; Michael Whitehouse; Marianna Thomas; Gregor Stenhouse; Tim Chesser; Chris Budd; H S Gill

doi:10.1038/s41598-022-06018-9

Machine learning outperforms clinical experts in classification of hip fractures

Ellen Murphy
, Beate Ehrhardt
, Celia Gregson
, Otto Von Arx
, April Hartley
, Michael Whitehouse
, Marianna Thomas
, Gregor Stenhouse
, Tim Chesser
, Chris Budd
, H S Gill

Royal United Hospitals Bath NHS Foundation Trust
University of Bristol
North Bristol NHS Trust

Research output: Contribution to journal › Article › peer-review

35 Citations (SciVal)

Abstract

Hip fractures are a major cause of morbidity and mortality in the elderly, and incur high health and social care costs. Given projected population ageing, the number of incident hip fractures is predicted to increase globally. As fracture classification strongly determines the chosen surgical treatment, differences in fracture classification influence patient outcomes and treatment costs. We aimed to create a machine learning method for identifying and classifying hip fractures, and to compare its performance to experienced human observers. We used 3,659 hip radiographs, classified by at least two expert clinicians. The machine learning method was able to classify hip fractures with 19% greater accuracy than humans, achieving overall accuracy of 92%.

Original language	English
Article number	2058 (2022)
Journal	Scientific Reports
Volume	12
Issue number	1
DOIs	https://doi.org/10.1038/s41598-022-06018-9
Publication status	Published - 8 Feb 2022

Funding

This study was funded by Arthroplasty for Arthritis Charity, and the NVIDIA Corporation provided the Titan X GPU through their academic grant scheme. We acknowledge the critical assistance of Mr Rich Wood, PACS Manager at Royal United Hospital NHS Foundation Trust in Bath, in identifying, anonymising and extracting the appropriate radiographs. We are also grateful to Professors Carola-Bibiane Schönlieb and Michael Tipping for their insight and advice on machine learning.

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 3 Good Health and Well-being

Access to Document

10.1038/s41598-022-06018-9Licence: CC BY

1 Finished
1 Active

Programme Grant: Mathematics of Deep Learning
Budd, C. (PI) & Ehrhardt, M. (CoI)
EPSRC
31/01/22 → 30/07/27
Project: Research council
ARCHi: Automated Recognition and Classification of Hip Fracture
Gill, R. (PI) & Ehrhardt, B. (Researcher)
Arthroplasty for Arthritis Charity
2/01/17 → 1/09/19
Project: UK charity

Dataset for "Machine learning outperforms clinical experts in classification of hip fractures"
Gill, R. (Creator), Ehrhardt, B. (Creator) & Murphy, E. (Creator), University of Bath, 8 Feb 2022
DOI: 10.15125/BATH-01011
Dataset

Cite this

@article{f8178ade2c4b481c8a1588e10a7a7433,

title = "Machine learning outperforms clinical experts in classification of hip fractures",

abstract = "Hip fractures are a major cause of morbidity and mortality in the elderly, and incur high health and social care costs. Given projected population ageing, the number of incident hip fractures is predicted to increase globally. As fracture classification strongly determines the chosen surgical treatment, differences in fracture classification influence patient outcomes and treatment costs. We aimed to create a machine learning method for identifying and classifying hip fractures, and to compare its performance to experienced human observers. We used 3,659 hip radiographs, classified by at least two expert clinicians. The machine learning method was able to classify hip fractures with 19\% greater accuracy than humans, achieving overall accuracy of 92\%.",

author = "Ellen Murphy and Beate Ehrhardt and Celia Gregson and \{Von Arx\}, Otto and April Hartley and Michael Whitehouse and Marianna Thomas and Gregor Stenhouse and Tim Chesser and Chris Budd and Gill, \{H S\}",

year = "2022",

month = feb,

day = "8",

doi = "10.1038/s41598-022-06018-9",

language = "English",

volume = "12",

journal = "Scientific Reports",

issn = "2045-2322",

publisher = "Nature Research",

number = "1",

}

TY - JOUR

T1 - Machine learning outperforms clinical experts in classification of hip fractures

AU - Murphy, Ellen

AU - Ehrhardt, Beate

AU - Gregson, Celia

AU - Von Arx, Otto

AU - Hartley, April

AU - Whitehouse, Michael

AU - Thomas, Marianna

AU - Stenhouse, Gregor

AU - Chesser, Tim

AU - Budd, Chris

AU - Gill, H S

PY - 2022/2/8

Y1 - 2022/2/8

N2 - Hip fractures are a major cause of morbidity and mortality in the elderly, and incur high health and social care costs. Given projected population ageing, the number of incident hip fractures is predicted to increase globally. As fracture classification strongly determines the chosen surgical treatment, differences in fracture classification influence patient outcomes and treatment costs. We aimed to create a machine learning method for identifying and classifying hip fractures, and to compare its performance to experienced human observers. We used 3,659 hip radiographs, classified by at least two expert clinicians. The machine learning method was able to classify hip fractures with 19% greater accuracy than humans, achieving overall accuracy of 92%.

AB - Hip fractures are a major cause of morbidity and mortality in the elderly, and incur high health and social care costs. Given projected population ageing, the number of incident hip fractures is predicted to increase globally. As fracture classification strongly determines the chosen surgical treatment, differences in fracture classification influence patient outcomes and treatment costs. We aimed to create a machine learning method for identifying and classifying hip fractures, and to compare its performance to experienced human observers. We used 3,659 hip radiographs, classified by at least two expert clinicians. The machine learning method was able to classify hip fractures with 19% greater accuracy than humans, achieving overall accuracy of 92%.

U2 - 10.1038/s41598-022-06018-9

DO - 10.1038/s41598-022-06018-9

M3 - Article

SN - 2045-2322

VL - 12

JO - Scientific Reports

JF - Scientific Reports

IS - 1

M1 - 2058 (2022)

ER -

Machine learning outperforms clinical experts in classification of hip fractures

Abstract

Funding

UN SDGs

Access to Document

Fingerprint

Projects

Programme Grant: Mathematics of Deep Learning

ARCHi: Automated Recognition and Classification of Hip Fracture

Datasets

Dataset for "Machine learning outperforms clinical experts in classification of hip fractures"

Cite this