Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Diffusion-MRI for Characterisation of Head and Neck Tumours

Gandomi, Minoo (2024) MSFT02 20241
Medical Physics Programme
Abstract
Diffusion magnetic resonance imaging (dMRI) in radiotherapy has the potential to facilitate the identification of treatment resistant regions within tumours, such as areas with insufficient oxygen supply. This is particularly interesting in head and neck (H&N) tumours with poor prognosis correlated with unidentified hypoxic areas. However, dMRI is sensitivive to the magnetic susceptibility differences between tissues, particularly when using fast imaging techniques such as echo planar imaging (EPI), which can cause geometric distortions. This problem becomes prominent in the complex H&N anatomy as radiotherapy-MRI applications require high spatial accuracy to ensure safe treatments. There are different methods to minimise the distortions,... (More)
Diffusion magnetic resonance imaging (dMRI) in radiotherapy has the potential to facilitate the identification of treatment resistant regions within tumours, such as areas with insufficient oxygen supply. This is particularly interesting in head and neck (H&N) tumours with poor prognosis correlated with unidentified hypoxic areas. However, dMRI is sensitivive to the magnetic susceptibility differences between tissues, particularly when using fast imaging techniques such as echo planar imaging (EPI), which can cause geometric distortions. This problem becomes prominent in the complex H&N anatomy as radiotherapy-MRI applications require high spatial accuracy to ensure safe treatments. There are different methods to minimise the distortions, either using post-processing distortion corrections (such as reverse phase encoding techniques) or optimising the image acquisition by using for example multi-shot EPI instead of the conventional single-shot EPI (ssEPI). This project aims to compare the efficacy of various distortion correction techniques in improving the accuracy and reliability of dMRI data. The goal is to obtain usable diffusion images for further research, particularly in the development of imaging biomarkers, and provide guidance for optimising H&N dMRI protocols by evaluating the results in-vivo.

ssEPI and multiplexed sensitivity encoding (MUSE) were compared, both employing several reversed phase encoding gradient corrections as a pre-processing technique. The directions applied were anterior-posterior (AP), posterior-anterior (PA), right-left (RL), left-right (LR), and combinations of AP/PA and RL/LR.. The distortion correction was performed either via the FMRIB Software Library or using an embedded vendor- based correction, which uses less reverse phase encoding information. The results were evaluated in both phantom and healthy volunteers through visual and quantitative assessment. Finally, a method based on previous measurements was tested in two patients and the oxygenation of the tumours were mapped.

The results show that the MUSE sequence is less prone to distortions compared to the ssEPI. The quantitative evaluation showed that the distortion corrections notably improved the geometric accuracy of the images. Corrections with more information in the reverse phase encoding direction yield better results compared to corrections which contains less information. The efficacy of the correction depends not only on the sequence, but also the direction of the phase encoding pairs in relation phantom geometry. Post-correction, both ssEPI and MUSE are feasible in-vivo with the best options for phase encoding directions being either AP/PA or all four directions. The four-directional corrections with ssEPI and MUSE appear to perform similarly, although small differences in the pharynx and tongue can be observed. The patient examination produced results similar to those of the healthy volunteer, with the addition of revealing tumour tissue. Oxygen maps were successfully generated, but require additional validation prior to implementation. Furthermore, Multi-shot sequences are sensitive to movements between shots, and the MUSE signal may need to be validated. By addressing distortions, it becomes feasible to advance the development of dependable imaging biomarkers for tumour characterisation in H&N.

In conclusion, MUSE shows reduced susceptibility to distortions compared to ssEPI and yields more accurate corrections in certain areas, particularly along the pharynx. It may be the preferred option when the region of interest neighbours a tissue-air interface. Both sequences, particularly when correcting in AP/PA or all four phase encoding directions, offer viable options for in-vivo imaging, although multi-shot sequences may require additional validation due to potential motion-related inaccuracies. (Less)
Popular Abstract (Swedish)
Varje år diagnostiseras ungefär 1600 personer i Sverige med huvud- och halscancer. De flesta av dessa patienter behandlas med strålbehandling, men tyvärr är prognosen för många ganska dålig. Forskning pågår för att förbättra behandlingen och utveckla mer effektiva sätt att döda tumören, samtidigt som man strävar efter att minimera biverkningarna. Ett sätt att göra detta är att använda avancerade bildtekniker, så som magnetkamera-undersökningar (MR), som kan identifiera områden inom tumören som är resistenta mot behandling och sedan anpassa stråldosen därefter.

Diffusions-MR möjliggör kartläggningen av hur vattenmolekyler rör sig i kroppens vävnader. Det förmedlar information om tumörens mikroskopiska struktur och potentiellt även dess... (More)
Varje år diagnostiseras ungefär 1600 personer i Sverige med huvud- och halscancer. De flesta av dessa patienter behandlas med strålbehandling, men tyvärr är prognosen för många ganska dålig. Forskning pågår för att förbättra behandlingen och utveckla mer effektiva sätt att döda tumören, samtidigt som man strävar efter att minimera biverkningarna. Ett sätt att göra detta är att använda avancerade bildtekniker, så som magnetkamera-undersökningar (MR), som kan identifiera områden inom tumören som är resistenta mot behandling och sedan anpassa stråldosen därefter.

Diffusions-MR möjliggör kartläggningen av hur vattenmolekyler rör sig i kroppens vävnader. Det förmedlar information om tumörens mikroskopiska struktur och potentiellt även dess syrenivåer, vilket är viktigt eftersom tumörer med låg syrenivå ofta är mer motståndskraftiga mot strålningen. Om dessa områden kan identifieras kan behandlingen optimeras för varje patient, så att den blir mer effektiv. Med nuvarande tekniker för diffusions-MR blir bilderna i huvud-halsområdet ofta av dålig kvalitet. Detta är på grund av att MR-bilder blir förvrängda när material med olika magnetiska egenskaper ligger nära varandra, så som luft, mjukvävnad och ben. I huvud- halsområdet finns det flera sådana problematiska övergångar, som t.ex. i bihålorna, munhålan och svalget, där mycket vävnad gränsar till luft.

I det här arbetet har olika metoder för att förbättra bildkvaliteten i diffusions-MR-bilder undersökts med syfte att öka användbarheten inom strålbehandling. Två bildtekniker, enkel och segmenterad datainsamling har jämförts. Dessutom har olika tekniker för att korrigera förvrängningarna testats, där bilder tas med förvrängningar i motsatta riktningar för att möjliggöra efterföljande korrigeringar.

Resultaten visar att den segmenterade datainsamlingen är mindre känslig för förvrängningar jämfört med den enkla. Bildkvaliteten förbättrades betydligt med korrigeringsteknikerna, särskilt när mer data från motstående insamlingsriktningar användes. Med dessa förbättringar kunde mer geometriskt korrekta bilder av tumörer och omkringliggande vävnad skapas, vilket är avgörande för att sedan kunna identifiera områden med låg syrenivå inom tumören.
För att utvärdera metoderna genomfördes MR-undersökningar både på fantom (en glasburk fylld med gelatin och luft), och på friska frivilliga. Slutligen testades de mest lovande metoderna på patienter med huvud- halscancer, där kartor över tumörens relativa syrenivå kunde skapas. Kartläggningen av syrenivåerna är inte utvärderad, utan ytterligare validering krävs innan metoderna kan implementeras kliniskt.
Sammanfattningsvis visar arbetet att kvaliteten på diffusions-MR-bilder i huvud- och halsområdet kan förbättras genom användning av segmenterad insamling och ytterligare korrigeringstekniker. Detta kan bidra till att optimera strålbehandlingen för patienter med huvud- och halscancer genom att möjliggöra mer exakt avbildning av tumörernas egenskaper och därmed ytterligare individanpassa behandlingen med förbättrade resultat. (Less)
Please use this url to cite or link to this publication:
author
Gandomi, Minoo
supervisor
organization
course
MSFT02 20241
year
type
H2 - Master's Degree (Two Years)
subject
language
English
id
9160331
date added to LUP
2024-06-20 08:32:21
date last changed
2024-06-20 08:32:21
@misc{9160331,
  abstract     = {{Diffusion magnetic resonance imaging (dMRI) in radiotherapy has the potential to facilitate the identification of treatment resistant regions within tumours, such as areas with insufficient oxygen supply. This is particularly interesting in head and neck (H&N) tumours with poor prognosis correlated with unidentified hypoxic areas. However, dMRI is sensitivive to the magnetic susceptibility differences between tissues, particularly when using fast imaging techniques such as echo planar imaging (EPI), which can cause geometric distortions. This problem becomes prominent in the complex H&N anatomy as radiotherapy-MRI applications require high spatial accuracy to ensure safe treatments. There are different methods to minimise the distortions, either using post-processing distortion corrections (such as reverse phase encoding techniques) or optimising the image acquisition by using for example multi-shot EPI instead of the conventional single-shot EPI (ssEPI). This project aims to compare the efficacy of various distortion correction techniques in improving the accuracy and reliability of dMRI data. The goal is to obtain usable diffusion images for further research, particularly in the development of imaging biomarkers, and provide guidance for optimising H&N dMRI protocols by evaluating the results in-vivo.

ssEPI and multiplexed sensitivity encoding (MUSE) were compared, both employing several reversed phase encoding gradient corrections as a pre-processing technique. The directions applied were anterior-posterior (AP), posterior-anterior (PA), right-left (RL), left-right (LR), and combinations of AP/PA and RL/LR.. The distortion correction was performed either via the FMRIB Software Library or using an embedded vendor- based correction, which uses less reverse phase encoding information. The results were evaluated in both phantom and healthy volunteers through visual and quantitative assessment. Finally, a method based on previous measurements was tested in two patients and the oxygenation of the tumours were mapped.

The results show that the MUSE sequence is less prone to distortions compared to the ssEPI. The quantitative evaluation showed that the distortion corrections notably improved the geometric accuracy of the images. Corrections with more information in the reverse phase encoding direction yield better results compared to corrections which contains less information. The efficacy of the correction depends not only on the sequence, but also the direction of the phase encoding pairs in relation phantom geometry. Post-correction, both ssEPI and MUSE are feasible in-vivo with the best options for phase encoding directions being either AP/PA or all four directions. The four-directional corrections with ssEPI and MUSE appear to perform similarly, although small differences in the pharynx and tongue can be observed. The patient examination produced results similar to those of the healthy volunteer, with the addition of revealing tumour tissue. Oxygen maps were successfully generated, but require additional validation prior to implementation. Furthermore, Multi-shot sequences are sensitive to movements between shots, and the MUSE signal may need to be validated. By addressing distortions, it becomes feasible to advance the development of dependable imaging biomarkers for tumour characterisation in H&N.

In conclusion, MUSE shows reduced susceptibility to distortions compared to ssEPI and yields more accurate corrections in certain areas, particularly along the pharynx. It may be the preferred option when the region of interest neighbours a tissue-air interface. Both sequences, particularly when correcting in AP/PA or all four phase encoding directions, offer viable options for in-vivo imaging, although multi-shot sequences may require additional validation due to potential motion-related inaccuracies.}},
  author       = {{Gandomi, Minoo}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Diffusion-MRI for Characterisation of Head and Neck Tumours}},
  year         = {{2024}},
}