Outlier detection and removal improves accuracy of machine learning approach to multispectral burn diagnostic imaging

Weizhi Li, Weirong Mo, Xu Zhang, John J. Squiers, Yang Lu, Eric W. Sellke, Wensheng Fan, J. Michael Dimaio, Jeffrey E. Thatcher

Research output: Contribution to journalArticle

32 Citations (Scopus)

Abstract

Multispectral imaging (MSI) was implemented to develop a burn tissue classification device to assist burn surgeons in planning and performing debridement surgery. To build a classification model via machine learning, training data accurately representing the burn tissue was needed, but assigning raw MSI data to appropriate tissue classes is prone to error. We hypothesized that removing outliers from the training dataset would improve classification accuracy. A swine burn model was developed to build an MSI training database and study an algorithm's burn tissue classification abilities. After the ground-truth database was generated, we developed a multistage method based on Z-test and univariate analysis to detect and remove outliers from the training dataset. Using 10-fold cross validation, we compared the algorithm's accuracy when trained with and without the presence of outliers. The outlier detection and removal method reduced the variance of the training data. Test accuracy was improved from 63% to 76%, matching the accuracy of clinical judgment of expert burn surgeons, the current gold standard in burn injury assessment. Given that there are few surgeons and facilities specializing in burn care, this technology may improve the standard of burn care for patients without access to specialized facilities.

Original languageEnglish (US)
Article number121305
JournalJournal of Biomedical Optics
Volume20
Issue number12
DOIs
StatePublished - Dec 1 2015

Fingerprint

machine learning
Learning systems
education
surgeons
Tissue
Imaging techniques
ground truth
swine
Surgery
surgery
planning
Planning

Keywords

  • burn
  • machine learning
  • medical imaging
  • model accuracy
  • multispectral imaging
  • outlier detection

ASJC Scopus subject areas

  • Biomedical Engineering
  • Biomaterials
  • Electronic, Optical and Magnetic Materials
  • Atomic and Molecular Physics, and Optics

Cite this

Outlier detection and removal improves accuracy of machine learning approach to multispectral burn diagnostic imaging. / Li, Weizhi; Mo, Weirong; Zhang, Xu; Squiers, John J.; Lu, Yang; Sellke, Eric W.; Fan, Wensheng; Dimaio, J. Michael; Thatcher, Jeffrey E.

In: Journal of Biomedical Optics, Vol. 20, No. 12, 121305, 01.12.2015.

Research output: Contribution to journalArticle

Li, W, Mo, W, Zhang, X, Squiers, JJ, Lu, Y, Sellke, EW, Fan, W, Dimaio, JM & Thatcher, JE 2015, 'Outlier detection and removal improves accuracy of machine learning approach to multispectral burn diagnostic imaging', Journal of Biomedical Optics, vol. 20, no. 12, 121305. https://doi.org/10.1117/1.JBO.20.12.121305
Li, Weizhi ; Mo, Weirong ; Zhang, Xu ; Squiers, John J. ; Lu, Yang ; Sellke, Eric W. ; Fan, Wensheng ; Dimaio, J. Michael ; Thatcher, Jeffrey E. / Outlier detection and removal improves accuracy of machine learning approach to multispectral burn diagnostic imaging. In: Journal of Biomedical Optics. 2015 ; Vol. 20, No. 12.
@article{2b53945c4be54937acad23f6f6b0e953,
title = "Outlier detection and removal improves accuracy of machine learning approach to multispectral burn diagnostic imaging",
abstract = "Multispectral imaging (MSI) was implemented to develop a burn tissue classification device to assist burn surgeons in planning and performing debridement surgery. To build a classification model via machine learning, training data accurately representing the burn tissue was needed, but assigning raw MSI data to appropriate tissue classes is prone to error. We hypothesized that removing outliers from the training dataset would improve classification accuracy. A swine burn model was developed to build an MSI training database and study an algorithm's burn tissue classification abilities. After the ground-truth database was generated, we developed a multistage method based on Z-test and univariate analysis to detect and remove outliers from the training dataset. Using 10-fold cross validation, we compared the algorithm's accuracy when trained with and without the presence of outliers. The outlier detection and removal method reduced the variance of the training data. Test accuracy was improved from 63{\%} to 76{\%}, matching the accuracy of clinical judgment of expert burn surgeons, the current gold standard in burn injury assessment. Given that there are few surgeons and facilities specializing in burn care, this technology may improve the standard of burn care for patients without access to specialized facilities.",
keywords = "burn, machine learning, medical imaging, model accuracy, multispectral imaging, outlier detection",
author = "Weizhi Li and Weirong Mo and Xu Zhang and Squiers, {John J.} and Yang Lu and Sellke, {Eric W.} and Wensheng Fan and Dimaio, {J. Michael} and Thatcher, {Jeffrey E.}",
year = "2015",
month = "12",
day = "1",
doi = "10.1117/1.JBO.20.12.121305",
language = "English (US)",
volume = "20",
journal = "Journal of Biomedical Optics",
issn = "1083-3668",
publisher = "SPIE",
number = "12",

}

TY - JOUR

T1 - Outlier detection and removal improves accuracy of machine learning approach to multispectral burn diagnostic imaging

AU - Li, Weizhi

AU - Mo, Weirong

AU - Zhang, Xu

AU - Squiers, John J.

AU - Lu, Yang

AU - Sellke, Eric W.

AU - Fan, Wensheng

AU - Dimaio, J. Michael

AU - Thatcher, Jeffrey E.

PY - 2015/12/1

Y1 - 2015/12/1

N2 - Multispectral imaging (MSI) was implemented to develop a burn tissue classification device to assist burn surgeons in planning and performing debridement surgery. To build a classification model via machine learning, training data accurately representing the burn tissue was needed, but assigning raw MSI data to appropriate tissue classes is prone to error. We hypothesized that removing outliers from the training dataset would improve classification accuracy. A swine burn model was developed to build an MSI training database and study an algorithm's burn tissue classification abilities. After the ground-truth database was generated, we developed a multistage method based on Z-test and univariate analysis to detect and remove outliers from the training dataset. Using 10-fold cross validation, we compared the algorithm's accuracy when trained with and without the presence of outliers. The outlier detection and removal method reduced the variance of the training data. Test accuracy was improved from 63% to 76%, matching the accuracy of clinical judgment of expert burn surgeons, the current gold standard in burn injury assessment. Given that there are few surgeons and facilities specializing in burn care, this technology may improve the standard of burn care for patients without access to specialized facilities.

AB - Multispectral imaging (MSI) was implemented to develop a burn tissue classification device to assist burn surgeons in planning and performing debridement surgery. To build a classification model via machine learning, training data accurately representing the burn tissue was needed, but assigning raw MSI data to appropriate tissue classes is prone to error. We hypothesized that removing outliers from the training dataset would improve classification accuracy. A swine burn model was developed to build an MSI training database and study an algorithm's burn tissue classification abilities. After the ground-truth database was generated, we developed a multistage method based on Z-test and univariate analysis to detect and remove outliers from the training dataset. Using 10-fold cross validation, we compared the algorithm's accuracy when trained with and without the presence of outliers. The outlier detection and removal method reduced the variance of the training data. Test accuracy was improved from 63% to 76%, matching the accuracy of clinical judgment of expert burn surgeons, the current gold standard in burn injury assessment. Given that there are few surgeons and facilities specializing in burn care, this technology may improve the standard of burn care for patients without access to specialized facilities.

KW - burn

KW - machine learning

KW - medical imaging

KW - model accuracy

KW - multispectral imaging

KW - outlier detection

UR - http://www.scopus.com/inward/record.url?scp=84940421624&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84940421624&partnerID=8YFLogxK

U2 - 10.1117/1.JBO.20.12.121305

DO - 10.1117/1.JBO.20.12.121305

M3 - Article

C2 - 26305321

AN - SCOPUS:84940421624

VL - 20

JO - Journal of Biomedical Optics

JF - Journal of Biomedical Optics

SN - 1083-3668

IS - 12

M1 - 121305

ER -