Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard1
  • 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
Improved Differential Diagnostics Using Methods in Machine Learning and Regression
KTH, School of Engineering Sciences (SCI), Mathematics (Dept.), Optimization and Systems Theory.
2018 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesisAlternative title
Förbättrad differentialdiagnostik med metoder inom maskininlärning och regression (Swedish)
Abstract [en]

There is a desire both from the patient and the society to have efficient tools for differential diagnostics. Mathematical relationships between diseases and observable consequences are defined in the thesis. Specifically artificial neural networks are considered in the modeling of the doctor’s methodology. To suggest further lab tests or symptoms to look for the network is inverted by looking at a minimization problem where the objective function gradient can be analytically calculated. Due to difficulties in obtaining real life medical data a program was constructed to generate artificial patient data sets. These data sets will be used to establish proof of concepts. Some data set quality measures are defined and used to model the network accuracy and training time. It is then estimated that a problem with 4000 diagnoses and 20 000 observable consequences would require 200 000 patients to obtain a classification accuracy of 99% with a training time of 50 hours depending on the computational power. Overall the solution strategy seems promising but studies on real life data is required for definitive answers.

Abstract [sv]

Det finns ett behov från både patienten och samhället att ha effektiva verktyg för differentialdiagnostik. Matematiska förhållanden mellan sjukdomar och ob-serverbara konsekvenser definieras i uppsatsen. Specifikt så används artificiella neuronnät i modelleringen av läkarens metodik. För att föreslå ytterligare labprover eller symptom att leta efter inverteras nätverket genom att studera ett minimeringsproblem där målfunktionens gradient kan beräknas analytiskt. På grund av svårigheter i att erhålla verklig medicinsk data konstruerades ett program för att generera artificiell patientdata. Denna patientdata kommer att användas för att etablera bevis på koncept. Några mått på kvalitén av patientdata definieras och används för att modellera nätverkets noggrannhet och träningstid. Det uppskattas sedan att ett problem med 4000 diagnoser och 20 000 observerbara konsekvenser skulle kräva 200 000 patienter för att uppnå en klassificeringsnoggrannhet på 99%med en träningstid på 50 timmar beroende på beräkningskraften. I helhet verkar lösningsstrategin lovande men studier på verklig data krävs för definitiva svar.

Place, publisher, year, edition, pages
2018.
Series
TRITA-SCI-GRU ; 2018:433
National Category
Computational Mathematics
Identifiers
URN: urn:nbn:se:kth:diva-240083OAI: oai:DiVA.org:kth-240083DiVA, id: diva2:1272346
External cooperation
BraineHealth
Subject / course
Systems Engineering
Educational program
Master of Science - Applied and Computational Mathematics
Supervisors
Examiners
Available from: 2018-12-19 Created: 2018-12-19 Last updated: 2018-12-19Bibliographically approved

Open Access in DiVA

fulltext(1419 kB)36 downloads
File information
File name FULLTEXT01.pdfFile size 1419 kBChecksum SHA-512
da4c6283eecc7d36adbb7f0eac7a47fb1c0b3195c238c158701c7da6b5a59c3d975ab6540c858cf5026460411ad3e719a7c5c3085bfdd2e4363cc9b4c1d049be
Type fulltextMimetype application/pdf

By organisation
Optimization and Systems Theory
Computational Mathematics

Search outside of DiVA

GoogleGoogle Scholar
Total: 36 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: 288 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard1
  • 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