Correlation between Surface and Tumour Motion in Lung Cancer - including Deep Learning Perspectives
(2023) MSFT02 20232Medical Physics Programme
- Abstract
- Purpose: The purpose of this master thesis was to retrospectively investigate correlation between surface and tumour motion in lung cancer patients, alongside deep learning applications of the results. Additional correlations such as age, tumour volume and anatomical placement of the tumour were also investigated.
Materials and Methods: 48 lung cancer patients treated with Stereotactic Body Radiation Therapy (SBRT) were included in this study. Delineation of the tumours was made on 4-Dimensional Computed Tomography (4DCT) images where each tumour was delineated in each of the eight respiration phases. Tumour volume and centre of mass coordinates in Left-Right (LR), Anterior-Posterior (AP) and Superior-Inferior (SI) directions were... (More) - Purpose: The purpose of this master thesis was to retrospectively investigate correlation between surface and tumour motion in lung cancer patients, alongside deep learning applications of the results. Additional correlations such as age, tumour volume and anatomical placement of the tumour were also investigated.
Materials and Methods: 48 lung cancer patients treated with Stereotactic Body Radiation Therapy (SBRT) were included in this study. Delineation of the tumours was made on 4-Dimensional Computed Tomography (4DCT) images where each tumour was delineated in each of the eight respiration phases. Tumour volume and centre of mass coordinates in Left-Right (LR), Anterior-Posterior (AP) and Superior-Inferior (SI) directions were retrieved in all respiration phases. The total translational shift from maximum exhale phase was computed. Surface motion data was acquired from a surface imaging system which was recorded during 4DCT simulation of the patient.
The Spearman Correlation Coefficient (SCC) between tumour and surface motion was calculated. An additional Spearman correlation was calculated between the SCC and patient features (age, tumour volume, smallest distance to thoracic wall, distance to thoracic spine and to chest surface, lung volume, and use of abdominal compression belt). Wilcoxon's rank sum test was performed to determine statistical significance between the groups with and without abdominal compression belt.
An artificial neural network (ANN) model was created to investigate the possibility to predict tumour motion given the surface motion as an input to the model. The model was built like a feed-forward ANN with two hidden layers using Rectified Linear Unit (ReLU) activation functions. Training, validation, and test data set split were 36/45, 6/45 and 3/45 respectively.
Results: Strong correlation (0.70<SCC<1.0) was found for 36%, 49%, 67% and 73% of the patients in the LR, AP, SI, and total 3D direction respectively. The correlations between SCC and patient features investigated were weak (SCC<0.30). There was a statistically significant difference in correlation coefficient values (p<0.05) between groups with and without abdominal compression belt in SI direction. A stronger correlation was seen in the patient cohort with abdominal compression belt (SCC=0.79) than in the patient cohort without (SCC=0.70).
When taking the ANN model uncertainty in terms of the mean squared error, 87.5% of the ground truth values were within the predicted interval.
Conclusion: The correlation between surface and tumour motion was strong in majority of the patients in SI and total 3D direction. However, SCC varied amongst the patient data set and the result should therefore be interpreted as an indication of possible further individualisation of treatment and treatment planning. If correlation is to be used for treatment planning, an action plan could be applied for when the patient breathing pattern is irregular or different. (Less) - Popular Abstract (Swedish)
- Av de 70 000 personerna som blev diagnoserade med cancer i Sverige 2021, fick runt 4000 lungcancer. Lungcancer är vanligen elakartade celler som utgår från lungans slemhinna. Behandling innefattar oftast operation eller strålterapi, ibland i kombination med cellgifter eller immunterapi.
Inför strålbehandling gör patienten en röntgenundersökning som ger bilder i tre dimensioner, en så kallad datortomografiundersökning. Läkarna använder sedan dessa bilder för att markera ut området som man vill bestråla, och organ runt om som man vill undvika att bestråla. Efter detta görs en dosplan där behandlingen simuleras i ett datorprogram för att göra en så optimerad behandling som möjligt. Där kan parametrar som strålfältstorlek, strålslag och... (More) - Av de 70 000 personerna som blev diagnoserade med cancer i Sverige 2021, fick runt 4000 lungcancer. Lungcancer är vanligen elakartade celler som utgår från lungans slemhinna. Behandling innefattar oftast operation eller strålterapi, ibland i kombination med cellgifter eller immunterapi.
Inför strålbehandling gör patienten en röntgenundersökning som ger bilder i tre dimensioner, en så kallad datortomografiundersökning. Läkarna använder sedan dessa bilder för att markera ut området som man vill bestråla, och organ runt om som man vill undvika att bestråla. Efter detta görs en dosplan där behandlingen simuleras i ett datorprogram för att göra en så optimerad behandling som möjligt. Där kan parametrar som strålfältstorlek, strålslag och ingångsvinkel bestämmas för att ge en hög och jämn fördelning av strålningen till tumören och så lite som möjligt till intilliggande frisk vävnad.
Vid behandling ligger patienten på en brits under en linjäraccelerator som producerar joniserande strålning. För lungcancer används ibland en speciell teknik som ger mycket hög dos till liten volym med mycket snabbt dosfall i intilliggande vävnad. Detta kallas stereotaktisk strålbehandling. Vid denna typ av behandling är det mycket viktigt att patienten ligger helt still.
Om tumören ligger i bröstkorgen, vilket är fallet för lungcancer, kan den röra sig under behandling på grund av andning. Risken är då att strålningen missar tumören om den rör sig med andningen. Läkarna kompenserar ofta för detta genom att göra det område som ska behandlas något större än den faktiska tumören baserat på datortomografiundersökningen. Detta säkerställer att tumören bestrålas trots rörelsen. En svårighet som kan uppstå är att vi ofta andas olika och rörelsen som uppstår kan variera mycket för samma person vid olika tillfällen. Detta innebär att det finns en risk att patienten andas annorlunda under behandling jämfört med under datortomografiundersökningen.
Vi behöver därför verktyg för att kontrollera hur andningen ser ut före och under behandling. Ett verktyg för att göra detta är optiska ytskanningssystem. Dessa system mäter hur mycket bröstkorgen rör sig och kan eventuellt ge en indikation på tumörrörelsen. Eftersom man inte ser själva tumören blir detta alltid en uppskattning, och för att bli säkrare på att tumören rör sig som förväntat med bröstkorgen eller magen kan en statistisk korrelation göras mellan dessa.
Detta arbete har undersökt denna faktiska statistiska korrelationen mellan bröstkorgens rörelse och tumörens rörelse i kroppen på lungcancerpatienter. Som hjälp har jag använt data från optiska ytskanningssystem samt datortomografibilder och undersökt hur många millimeter som de båda rör sig. Som tillägg till detta har jag även gjort en datormodell som ska hjälpa att förutspå var tumören kommer att ligga utifrån bröstkorgsrörelsen. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9139565
- author
- Kjellström, Caisa
- supervisor
- organization
- course
- MSFT02 20232
- year
- 2023
- type
- H2 - Master's Degree (Two Years)
- subject
- language
- English
- id
- 9139565
- date added to LUP
- 2023-10-05 09:20:55
- date last changed
- 2023-10-05 09:20:55
@misc{9139565,
abstract = {{Purpose: The purpose of this master thesis was to retrospectively investigate correlation between surface and tumour motion in lung cancer patients, alongside deep learning applications of the results. Additional correlations such as age, tumour volume and anatomical placement of the tumour were also investigated.
Materials and Methods: 48 lung cancer patients treated with Stereotactic Body Radiation Therapy (SBRT) were included in this study. Delineation of the tumours was made on 4-Dimensional Computed Tomography (4DCT) images where each tumour was delineated in each of the eight respiration phases. Tumour volume and centre of mass coordinates in Left-Right (LR), Anterior-Posterior (AP) and Superior-Inferior (SI) directions were retrieved in all respiration phases. The total translational shift from maximum exhale phase was computed. Surface motion data was acquired from a surface imaging system which was recorded during 4DCT simulation of the patient.
The Spearman Correlation Coefficient (SCC) between tumour and surface motion was calculated. An additional Spearman correlation was calculated between the SCC and patient features (age, tumour volume, smallest distance to thoracic wall, distance to thoracic spine and to chest surface, lung volume, and use of abdominal compression belt). Wilcoxon's rank sum test was performed to determine statistical significance between the groups with and without abdominal compression belt.
An artificial neural network (ANN) model was created to investigate the possibility to predict tumour motion given the surface motion as an input to the model. The model was built like a feed-forward ANN with two hidden layers using Rectified Linear Unit (ReLU) activation functions. Training, validation, and test data set split were 36/45, 6/45 and 3/45 respectively.
Results: Strong correlation (0.70<SCC<1.0) was found for 36%, 49%, 67% and 73% of the patients in the LR, AP, SI, and total 3D direction respectively. The correlations between SCC and patient features investigated were weak (SCC<0.30). There was a statistically significant difference in correlation coefficient values (p<0.05) between groups with and without abdominal compression belt in SI direction. A stronger correlation was seen in the patient cohort with abdominal compression belt (SCC=0.79) than in the patient cohort without (SCC=0.70).
When taking the ANN model uncertainty in terms of the mean squared error, 87.5% of the ground truth values were within the predicted interval.
Conclusion: The correlation between surface and tumour motion was strong in majority of the patients in SI and total 3D direction. However, SCC varied amongst the patient data set and the result should therefore be interpreted as an indication of possible further individualisation of treatment and treatment planning. If correlation is to be used for treatment planning, an action plan could be applied for when the patient breathing pattern is irregular or different.}},
author = {{Kjellström, Caisa}},
language = {{eng}},
note = {{Student Paper}},
title = {{Correlation between Surface and Tumour Motion in Lung Cancer - including Deep Learning Perspectives}},
year = {{2023}},
}