The role of laboratory data in 'knowledgeable surveillance'

Crawford Revie, Fernanda Dórea

Research output: Contribution to conferenceKeynote

Abstract

Over the past decade the availability of digital data relating to animal health has grown exponentially, and with it an interest in making effective and timely use of these data. In particular the use of syndrome-based indicators to augment traditional laboratory results for the purpose of disease surveillance has been the focus of a number of studies. The volume and semi-structured nature of such data, together with the fact that it must often be processed in real time, have led to methodological challenges in the appropriate interpretation of these novel data sources. In this talk I will discuss a range of techniques ranging from text-mining, times series analyses and clustering algorithms that can be used to identify syndromic signals in laboratory test request data, together with statistical techniques that can be used to detect the various types of temporal aberrations that can occur. These approaches have been implemented in systems linked to animal health laboratory systems in Canada and Sweden, and their use will be illustrated by way of case-based examples.

However, the isolated use of laboratory data is rarely adequate in the context of syndromic surveillance, and a variety of animal health data sources are being explored for early disease detection. In terms of 'next steps' towards successfully using such data, I believe that the integration of evidence from multiple sources is of critical importance. A key challenge in moving forward is the need to ensure that aggregation and comparison across data sources is being made among similar objects. In this context we are exploring the use of knowledge-based ontologies, which provide machine-readable methods for the representation of and inference from data. We will discuss one such pilot ontology – AHSO (Animal Health Surveillance Ontology) – and illustrate the ways in which the availability of frameworks such as this can be complemented by recent advances in computer science, including deep learning and the Semantic Web. Research results from these areas will allow for the integration of information derived from diagnostic data with that extracted from other sources of animal health information, including clinical records, mortality and even regular production data, to create a framework for truly "knowledgeable" surveillance.

Conference

ConferenceXVII International Symposium of the World Association of Veterinary Laboratory Diagnosticians
Abbreviated titleWAVLD
CountryCanada
CitySaskatoon
Period15/06/1518/06/15
Internet address

Fingerprint

Animals
Health
Ontology
Availability
Aberrations
Clustering algorithms
Computer science
Time series
Agglomeration
Semantics

Keywords

  • knowledge surveillance
  • machine learning
  • AI
  • artifical intelligence
  • laboratory data
  • data integration

Cite this

Revie, C., & Dórea, F. (2015). The role of laboratory data in 'knowledgeable surveillance'. 1-59. XVII International Symposium of the World Association of Veterinary Laboratory Diagnosticians, Saskatoon, Canada.
Revie, Crawford ; Dórea, Fernanda. / The role of laboratory data in 'knowledgeable surveillance'. XVII International Symposium of the World Association of Veterinary Laboratory Diagnosticians, Saskatoon, Canada.59 p.
@conference{506a6546f44242aaa5314f5832a43fc4,
title = "The role of laboratory data in 'knowledgeable surveillance'",
abstract = "Over the past decade the availability of digital data relating to animal health has grown exponentially, and with it an interest in making effective and timely use of these data. In particular the use of syndrome-based indicators to augment traditional laboratory results for the purpose of disease surveillance has been the focus of a number of studies. The volume and semi-structured nature of such data, together with the fact that it must often be processed in real time, have led to methodological challenges in the appropriate interpretation of these novel data sources. In this talk I will discuss a range of techniques ranging from text-mining, times series analyses and clustering algorithms that can be used to identify syndromic signals in laboratory test request data, together with statistical techniques that can be used to detect the various types of temporal aberrations that can occur. These approaches have been implemented in systems linked to animal health laboratory systems in Canada and Sweden, and their use will be illustrated by way of case-based examples. However, the isolated use of laboratory data is rarely adequate in the context of syndromic surveillance, and a variety of animal health data sources are being explored for early disease detection. In terms of 'next steps' towards successfully using such data, I believe that the integration of evidence from multiple sources is of critical importance. A key challenge in moving forward is the need to ensure that aggregation and comparison across data sources is being made among similar objects. In this context we are exploring the use of knowledge-based ontologies, which provide machine-readable methods for the representation of and inference from data. We will discuss one such pilot ontology – AHSO (Animal Health Surveillance Ontology) – and illustrate the ways in which the availability of frameworks such as this can be complemented by recent advances in computer science, including deep learning and the Semantic Web. Research results from these areas will allow for the integration of information derived from diagnostic data with that extracted from other sources of animal health information, including clinical records, mortality and even regular production data, to create a framework for truly {"}knowledgeable{"} surveillance.",
keywords = "knowledge surveillance, machine learning, AI, artifical intelligence, laboratory data, data integration",
author = "Crawford Revie and Fernanda D{\'o}rea",
year = "2015",
month = "6",
day = "18",
language = "English",
pages = "1--59",
note = "XVII International Symposium of the World Association of Veterinary Laboratory Diagnosticians, WAVLD ; Conference date: 15-06-2015 Through 18-06-2015",
url = "https://wcvmtoday.usask.ca/articles/2015/symposium-draws-diagnostic-experts-to-saskatoon.php",

}

Revie, C & Dórea, F 2015, 'The role of laboratory data in 'knowledgeable surveillance'' XVII International Symposium of the World Association of Veterinary Laboratory Diagnosticians, Saskatoon, Canada, 15/06/15 - 18/06/15, pp. 1-59.

The role of laboratory data in 'knowledgeable surveillance'. / Revie, Crawford; Dórea, Fernanda.

2015. 1-59 XVII International Symposium of the World Association of Veterinary Laboratory Diagnosticians, Saskatoon, Canada.

Research output: Contribution to conferenceKeynote

TY - CONF

T1 - The role of laboratory data in 'knowledgeable surveillance'

AU - Revie, Crawford

AU - Dórea, Fernanda

PY - 2015/6/18

Y1 - 2015/6/18

N2 - Over the past decade the availability of digital data relating to animal health has grown exponentially, and with it an interest in making effective and timely use of these data. In particular the use of syndrome-based indicators to augment traditional laboratory results for the purpose of disease surveillance has been the focus of a number of studies. The volume and semi-structured nature of such data, together with the fact that it must often be processed in real time, have led to methodological challenges in the appropriate interpretation of these novel data sources. In this talk I will discuss a range of techniques ranging from text-mining, times series analyses and clustering algorithms that can be used to identify syndromic signals in laboratory test request data, together with statistical techniques that can be used to detect the various types of temporal aberrations that can occur. These approaches have been implemented in systems linked to animal health laboratory systems in Canada and Sweden, and their use will be illustrated by way of case-based examples. However, the isolated use of laboratory data is rarely adequate in the context of syndromic surveillance, and a variety of animal health data sources are being explored for early disease detection. In terms of 'next steps' towards successfully using such data, I believe that the integration of evidence from multiple sources is of critical importance. A key challenge in moving forward is the need to ensure that aggregation and comparison across data sources is being made among similar objects. In this context we are exploring the use of knowledge-based ontologies, which provide machine-readable methods for the representation of and inference from data. We will discuss one such pilot ontology – AHSO (Animal Health Surveillance Ontology) – and illustrate the ways in which the availability of frameworks such as this can be complemented by recent advances in computer science, including deep learning and the Semantic Web. Research results from these areas will allow for the integration of information derived from diagnostic data with that extracted from other sources of animal health information, including clinical records, mortality and even regular production data, to create a framework for truly "knowledgeable" surveillance.

AB - Over the past decade the availability of digital data relating to animal health has grown exponentially, and with it an interest in making effective and timely use of these data. In particular the use of syndrome-based indicators to augment traditional laboratory results for the purpose of disease surveillance has been the focus of a number of studies. The volume and semi-structured nature of such data, together with the fact that it must often be processed in real time, have led to methodological challenges in the appropriate interpretation of these novel data sources. In this talk I will discuss a range of techniques ranging from text-mining, times series analyses and clustering algorithms that can be used to identify syndromic signals in laboratory test request data, together with statistical techniques that can be used to detect the various types of temporal aberrations that can occur. These approaches have been implemented in systems linked to animal health laboratory systems in Canada and Sweden, and their use will be illustrated by way of case-based examples. However, the isolated use of laboratory data is rarely adequate in the context of syndromic surveillance, and a variety of animal health data sources are being explored for early disease detection. In terms of 'next steps' towards successfully using such data, I believe that the integration of evidence from multiple sources is of critical importance. A key challenge in moving forward is the need to ensure that aggregation and comparison across data sources is being made among similar objects. In this context we are exploring the use of knowledge-based ontologies, which provide machine-readable methods for the representation of and inference from data. We will discuss one such pilot ontology – AHSO (Animal Health Surveillance Ontology) – and illustrate the ways in which the availability of frameworks such as this can be complemented by recent advances in computer science, including deep learning and the Semantic Web. Research results from these areas will allow for the integration of information derived from diagnostic data with that extracted from other sources of animal health information, including clinical records, mortality and even regular production data, to create a framework for truly "knowledgeable" surveillance.

KW - knowledge surveillance

KW - machine learning

KW - AI

KW - artifical intelligence

KW - laboratory data

KW - data integration

M3 - Keynote

SP - 1

EP - 59

ER -

Revie C, Dórea F. The role of laboratory data in 'knowledgeable surveillance'. 2015. XVII International Symposium of the World Association of Veterinary Laboratory Diagnosticians, Saskatoon, Canada.