Digitala Vetenskapliga Arkivet

Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Supervised Machine Learning for Identification of Severely Injured Trauma Patients
KTH, School of Engineering Sciences in Chemistry, Biotechnology and Health (CBH), Biomedical Engineering and Health Systems, Biomedical Engineering and Health Systems.
2024 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesisAlternative title
Övervakad maskininlärning för identifiering av svårt skadade traumapatienter (Swedish)
Abstract [en]

Trauma is one of the leading causes of mortality, and is the leading cause of death among young people. The rapid activation of a trauma team for severely injured trauma patients is critical for improving patient outcomes. Severely injured patients must be rapidly identified upon arrival at the emergency department to activate a trauma team, making the need for triage tools that can rapidly and accurately assess the need for trauma team activation (TTA) important. This project explores the use of supervised machine learning models to predict the need for TTA using data from the Swedish Trauma Registry (SweTrau). The research focuses on the use of two machine learning models. A fine-tuned Large Language Model (LLM) for analyzing triage notes, and an AdaBoost ensemble model utilizing decision trees for structured data selected from the SweTrau by consulting an expert. The models were trained to predict the necessity of TTA, defined by a New Injury Severity Score (NISS) greater than 15. The results demonstrate that both models outperform the current triage practices in terms of sensitivity, ability to discern between patients severely injured patients and non-severely injured patients, with a combination of the models output offering the best performance. Although neither model is yet suitable for clinical use, the findings suggest that machine learning has the potential to enhance trauma triage protocols, particularly in identifying severe trauma cases that require TTA that could otherwise be overlooked.

Abstract [sv]

Trauma är ledande dödsorsak och är den vanligaste dödsorsaken bland unga. Aktiveringen av ett traumateam är avgörande för att förbättra utfallet för svårt skadade traumapatienter. Svårt skadade patienter måste snabbt identifieras vid ankomst till akutmottagningen, vilket kräver ett triageverktyg som snabbt och exakt kan bedöma patientens behov av traumalarm (TTA). Detta projekt utforskar användningen av övervakade maskininlärningsmodeller för att förutsäga behovet av TTA med hjälp av journaldata och register från Svenska Traumaregistret (SweTrau). Forskningen fokuserar på användningen av två maskininlärnings modeller. En finjusterad Large Language Model (LLM) för att analysera triagejournaler och en AdaBoost-ensemblemodell som använder beslutsträd för att utvärdera strukturerad data som SweTrau. Modellerna tränades för att förutsäga behovet av TTA, definierad av en New Injury Severity Score (NISS) högre än 15. Resultaten visar att båda modellerna överträffar nuvarande triage-praxis när det gäller känslighet, förmåga att skilja mellan lindrigt skadade och skadade patienter, där en kombination av modellerna gav bäst resultat. Även om ingen av modellerna ännu är lämpliga för klinisk användning, tyder fyndet på att maskininlärning har potential som triageverktyg för att identifiera patienter i behov av TTA.

Place, publisher, year, edition, pages
2024. , p. 41
Series
TRITA-CBH-GRU ; 2024:338
Keywords [en]
Triage, Trauma, Large Language Model, Classification, Trauma Team Activation, TTA, Machine Learning, SweTrau, NISS
Keywords [sv]
Triage, Trauma, Stor generativ språkmodell, Klassificering, Aktivering av traumateam, TTA, Maskininlärning, SweTrau, NISS
National Category
Medical Engineering Medical Informatics Engineering Artificial Intelligence
Identifiers
URN: urn:nbn:se:kth:diva-362475OAI: oai:DiVA.org:kth-362475DiVA, id: diva2:1952667
External cooperation
Karolinska Institutet
Subject / course
Medical Engineering
Educational program
Master of Science - Medical Engineering
Supervisors
Examiners
Available from: 2025-04-22 Created: 2025-04-16 Last updated: 2025-04-22Bibliographically approved

Open Access in DiVA

Supervised Machine Learning for Identification of Severely Injured Trauma Patients(1070 kB)34 downloads
File information
File name FULLTEXT01.pdfFile size 1070 kBChecksum SHA-512
7c3edd56621e0d3778b0b42edd9786bacd2d16ce0a6329326cab582f8bba2019a9c5127d7b8ca31ca3a81e23f2ecd0f593752f6bb5c9169b8f2fce65c4d3cad7
Type fulltextMimetype application/pdf

By organisation
Biomedical Engineering and Health Systems
Medical EngineeringMedical Informatics EngineeringArtificial Intelligence

Search outside of DiVA

GoogleGoogle Scholar
Total: 34 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 210 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf