kth.sePublications KTH
910111213141512 of 22
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
Quantitative Computed Tomography in Health and Chronic Airflow Limitation: Regional Analysis and Deep Learning Methods
KTH, School of Engineering Sciences in Chemistry, Biotechnology and Health (CBH), Biomedical Engineering and Health Systems, Medical Imaging. Department of Clinical Science, Intervention & Technology Karolinska Institutet Stockholm Sweden;Department of Biomedical Engineering Karolinska University Hospital Stockholm Sweden.ORCID iD: 0000-0003-2850-6604
2026 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

Quantitative computed tomography (QCT) enables objective assessment of lung structure and may provide information complementary to spirometry in chronic airflow limitation (CAL). Inspiratory and expiratory chest CT can assess lung volume, lung density, low-attenuation volume, and ventilation-related changes. However, their role in population-based cohorts remains less extensively studied, particularly in individuals without established lung disease or with mild airflow limitation. Furthermore, regional analysis requires accurate lung lobe segmentation, and the value of radiomic and deep learning-derived features for identifying CAL remains incompletely understood.

The overall aim of this thesis was to extract, analyse, automate, and regionalise quantitative measures from inspiratory and expiratory chest CT images, and to study their relationship with normal lung structure, ventilation-related changes, and spirometry-defined CAL. Data were obtained from the Stockholm cohort of the Swedish CArdioPulmonary bioImage Study (SCAPIS), a population-based cohort of men and women aged 50–64 years. Inspiratory and expiratory chest CT, post-bronchodilator spirometry, and questionnaire data were used. CAL was defined as a post-bronchodilator FEV1/FVC ratio < 0.70.

Study I evaluated global inspiratory and expiratory CT-derived lung volumes, mean lung density, low-attenuation volume, and density gradients. Inspiratory CT lung volumes were lower than literature-based total lung capacity reference values, whereas expiratory CT lung volumes exceeded residual volume reference values. Participants with CAL had higher inspiratory and expiratory lung volumes, lower mean lung density, and greater low-attenuation volume than participants without CAL. Expiratory CT measures showed better discriminatory performance than inspiratory measures, with the highest performance observed for expiratory low-attenuation measures. A dorsal–ventral attenuation gradient was observed during expiration in participants without CAL but not in participants with CAL.

Study II trained and evaluated a deep learning-based lung lobe segmentation method incorporating prior anatomical information from lung vessel connectivity. The method was evaluated in both inspiratory and expiratory CT, including expiratory scans, in which segmentation is typically more demanding. Prior anatomical information mainly improved boundary accuracy, with the clearest benefit in expiratory CT. The best overall segmentation performance was achieved using a multitask model segmenting both lung lobes and fissures. Complementary downstream analyses showed that the choice of segmentation method had a limited effect on CAL discrimination in the full cohort, although variability was greater in smaller cohorts.

Study III extended the global analyses in Study I to lobar CT measures in a larger cohort. Participants with CAL had higher inspiratory and expiratory lung volumes, higher low-attenuation volumes, and more negative mean lung density than participants without CAL. Expiratory measures again showed stronger discriminatory ability than inspiratory measures. Lobar analyses demonstrated regional heterogeneity in lung volume, density, and low-attenuation measures, while complementary thesis analyses examined dorsal–ventral variation. In regularised logistic regression models, combined inspiratory–expiratory measures contributed to CAL discrimination; performance improved when expiratory measures were added to inspiratory measures and showed a slight additional improvement with lobar variables. The best Study III model achieved an area under the ROC curve (AUC) of 0.81.

Study IV assessed whether lobe-level radiomic and deep learning-derived features improved CAL discrimination beyond established quantitative CT measures. Classical radiomics, SegResNet, and 3D U-Net feature models were compared using inspiratory, expiratory, and combined inspiratory–expiratory data. Expiratory feature models consistently outperformed inspiratory models. The best performance was achieved by the classical radiomics model combining inspiratory and expiratory features, which outperformed the handcrafted quantitative CT model from Study III. Feature selection showed that classical radiomics relied mainly on expiratory texture features, whereas deep learning-derived models showed more balanced contributions from inspiratory and expiratory images. Combining radiomic, deep learning-derived, and handcrafted feature sets did not improve performance beyond the best individual model.

In conclusion, inspiratory and expiratory chest CT provide quantitative measures associated with spirometry-defined CAL in a population-based cohort. Expiratory CT was consistently more informative for discriminating CAL than inspiratory CT, supporting its value for assessing ventilation-related abnormalities and CAL-related lung changes relevant to early COPD. However, regression analyses showed that combinations of inspiratory and expiratory measures contributed to CAL classification, suggesting that the two respiratory phases provided complementary information. Automated lung lobe segmentation enabled scalable regional analysis, while lobar and radiomic approaches added information beyond whole-lung measures. Together, the findings support QCT as a complementary imaging-based approach for describing lung structure, regional heterogeneity, and CAL-related abnormalities in population-based research.

Abstract [sv]

Kronisk luftflödesbegränsning (på engelska chronic airflow limitation, CAL) innebär en varaktigt nedsatt förmåga att effektivt andas ut luft ur lungorna. Det kan orsakas av förträngningar i de små luftvägarna, minskad elasticitet i lungvävnaden eller att luft blir kvar i lungorna efter utandning. Kronisk luftflödesbegränsning är ett centralt kännetecken vid kroniskt obstruktiv lungsjukdom (KOL; på engelska chronic obstructive pulmonary disease, COPD), en av de vanligaste kroniska lungsjukdomarna och en betydande orsak till sjuklighet och dödlighet i världen. Vid KOL bidrar strukturella förändringar i luftvägarna och lungvävnaden till försämrad andningsfunktion. Dessa förändringar utvecklas ofta gradvis och kan vara svåra att upptäcka i ett tidigt skede. Tidig upptäckt och behandling kan dock bidra till att bromsa sjukdomsutvecklingen och förbättra prognosen. Spirometri, ett andningstest som mäter hur mycket luft en person kan andas ut och hur snabbt detta sker, är standardmetoden för att påvisa kronisk luftflödesbegränsning. Spirometri ger dock ett övergripande mått på lungfunktionen och visar inte var i lungorna avvikelserna finns.

Datortomografi (DT; på engelska computed tomography, CT) är en bilddiagnostisk metod som kan avbilda lungorna med hög detaljrikedom. I denna avhandling användes DT-bilder inte enbart för visuell bedömning, utan även för att beräkna numeriska mått på lungorna. Detta tillvägagångssätt kallas kvantitativ DT. Metoden kan användas för att mäta lungvolym, densitet i lungan och områden i lungan som innehåller ovanligt mycket luft. Sådana mått kan ge ytterligare information om tidiga lungförändringar kopplade till kronisk luftflödesbegränsning och KOL.

Studierna i denna avhandling baserades huvudsakligen på data från Stockholmsdelen av Swedish CArdioPulmonary bioImage Study, SCAPIS. SCAPIS är en stor befolkningsbaserad studie av medelålders kvinnor och män. Deltagarna genomgick DT-undersökning av lungorna både vid inandning och utandning, samt spirometri och frågeformulärsbaserade undersökningar.

Den första delen av avhandlingen undersökte om kvantitativa DT-mått från inandnings- och utandningsbilder kunde identifiera tecken på kronisk luftflödesbegränsning. Resultaten visade att deltagare med kronisk luftflödesbegränsning hade större lungvolymer, lägre medeldensitet och mer lungvävnad med låg densitet än deltagare utan luftflödesbegränsning. Mätvärden från utandnings-DT kunde bättre skilja personer med kronisk luftflödesbegränsning från personer utan sådana förändringar än mätvärden från inandnings-DT. Detta tyder på att utandningsbilder kan vara särskilt användbara för att identifiera förändringar kopplade till kronisk luftflödesbegränsning och relevanta för tidig KOL.

Avhandlingen studerade även hur förändringar skiljer sig mellan olika delar av lungan. Lungorna är indelade i lober, och lungsjukdom kan påverka vissa av dessa regioner mer än andra. För att möjliggöra storskalig regional analys omfattade avhandlingen automatisk lunglobssegmentering med hjälp av djupinlärning. Detta innebär att en datormodell tränades för att dela in lungorna i deras anatomiska lober. Metoden utformades för att fungera inte bara på inandningsbilder, utan även på utandningsbilder, där lungorna är mindre och gränserna mellan loberna kan vara svårare att identifiera.

Med hjälp av automatisk segmentering kunde DT-mått analyseras separat i olika lunglober och regioner. Detta gjorde det möjligt att beskriva var i lungorna förändringar kopplade till kronisk luftflödesbegränsning var lokaliserade. Regionala och lobära analyser gav ytterligare information om fördelningen av dessa lungförändringar, men endast en mindre förbättring i identifieringen av luftflödesbegränsning.

Slutligen undersökte avhandlingen om mer avancerad information från DT-bilderna kunde förbättra identifieringen av kronisk luftflödesbegränsning. Detta gjordes med radiomik, en metod där medicinska bilder analyseras med dator för att fånga kvantitativa mönster och egenskaper som inte nödvändigtvis är synliga för det mänskliga ögat eller kan fångas med enklare mätningar. I denna avhandling analyserades datortomografibilder med både traditionella radiomikmetoder och metoder baserade på djupinlärning. Resultaten visade att sådana mått var möjliga att använda. Modeller baserade på klassisk radiomik förbättrade identifieringen av kronisk luftflödesbegränsning jämfört med tidigare använda kvantitativa DT-mått. 

Sammanfattningsvis visar denna avhandling att kvantitativ DT kan ge användbar bildbaserad information om lungstruktur och kronisk luftflödesbegränsning i ett befolkningsbaserat material av medelålders män och kvinnor. Utandnings-DT var särskilt informativt, och automatisk lunglobssegmentering gjorde det möjligt att studera regionala skillnader i lungorna. Dessa metoder är inte avsedda att ersätta spirometri, men de kan komplettera andningstester genom att visa strukturella och regionala lungförändringar. Efter ytterligare validering kan kvantitativ DT bidra till ökad förståelse av tidiga KOL-relaterade lungförändringar och stödja mer detaljerad bildbaserad karakterisering av kroniskt obstruktiv lungsjukdom.

Place, publisher, year, edition, pages
Huddinge: Karolinska Institutet , 2026. , p. 101
Series
TRITA-CBH-FOU ; 2026:32
Keywords [en]
computed tomography, chronic airflow limitation, deep learning
National Category
Medical Imaging
Research subject
Medical Technology
Identifiers
URN: urn:nbn:se:kth:diva-387489DOI: 10.69622/32509146ISBN: 978-91-8141-166-9 (print)OAI: oai:DiVA.org:kth-387489DiVA, id: diva2:2094264
Public defence
2026-09-25, NEO, Erna Möllersalen, Blickagången 16, 141 52 Huddinge, Huddinge, 09:00 (English)
Opponent
Supervisors
Note

QC 2026-08-21

Available from: 2026-08-21 Created: 2026-08-21 Last updated: 2026-09-07Bibliographically approved
List of papers
1. Pulmonary volumes and signs of chronic airflow limitation in quantitative computed tomography
Open this publication in new window or tab >>Pulmonary volumes and signs of chronic airflow limitation in quantitative computed tomography
Show others...
2024 (English)In: Clinical Physiology and Functional Imaging, ISSN 1475-0961, E-ISSN 1475-097X, Vol. 44, no 4, p. 340-348Article in journal (Refereed) Published
Abstract [en]

Background

Computed tomography (CT) offers pulmonary volumetric quantification but is not commonly used in healthy individuals due to radiation concerns. Chronic airflow limitation (CAL) is one of the diagnostic criteria for chronic obstructive pulmonary disease (COPD), where early diagnosis is important. Our aim was to present reference values for chest CT volumetric and radiodensity measurements and explore their potential in detecting early signs of CAL.

Methods

From the population-based Swedish CArdioPulmonarybioImage Study (SCAPIS), 294 participants aged 50–64, were categorized into non-CAL (n = 258) and CAL (n = 36) groups based on spirometry. From inspiratory and expiratory CT images we compared lung volumes, mean lung density (MLD), percentage of low attenuation volume (LAV%) and LAV cluster volume between groups, and against reference values from static pulmonary function test (PFT).

Results

The CAL group exhibited larger lung volumes, higher LAV%, increased LAV cluster volume and lower MLD compared to the non-CAL group. Lung volumes significantly deviated from PFT values. Expiratory measurements yielded more reliable results for identifying CAL compared to inspiratory. Using a cut-off value of 0.6 for expiratory LAV%, we achieved sensitivity, specificity and positive/negative predictive values of 72%, 85% and 40%/96%, respectively.

Conclusion

We present volumetric reference values from inspiratory and expiratory chest CT images for a middle-aged healthy cohort. These results are not directly comparable to those from PFTs. Measures of MLD and LAV can be valuable in the evaluation of suspected CAL. Further validation and refinement are necessary to demonstrate its potential as a decision support tool for early detection of COPD.

Place, publisher, year, edition, pages
Wiley, 2024
Keywords
medical image processing
National Category
Radiology, Nuclear Medicine and Medical Imaging Medical Imaging
Research subject
Technology and Health; Medical Technology
Identifiers
urn:nbn:se:kth:diva-350134 (URN)10.1111/cpf.12880 (DOI)001196740700001 ()38576112 (PubMedID)2-s2.0-85189452440 (Scopus ID)
Funder
Swedish Heart Lung Foundation
Note

QC 20240708

Available from: 2024-07-06 Created: 2024-07-06 Last updated: 2026-08-21Bibliographically approved
2. Lung vessel connectivity map as anatomical prior knowledge for deep learning-based lung lobe segmentation
Open this publication in new window or tab >>Lung vessel connectivity map as anatomical prior knowledge for deep learning-based lung lobe segmentation
Show others...
2024 (English)In: Journal of Medical Imaging, ISSN 2329-4302, E-ISSN 2329-4310, Vol. 11, no 4Article in journal (Refereed) Published
Abstract [en]

Purpose Our study investigates the potential benefits of incorporating prior anatomical knowledge into a deep learning (DL) method designed for the automated segmentation of lung lobes in chest CT scans. Approach We introduce an automated DL-based approach that leverages anatomical information from the lung's vascular system to guide and enhance the segmentation process. This involves utilizing a lung vessel connectivity (LVC) map, which encodes relevant lung vessel anatomical data. Our study explores the performance of three different neural network architectures within the nnU-Net framework: a standalone U-Net, a multitasking U-Net, and a cascade U-Net. Results Experimental findings suggest that the inclusion of LVC information in the DL model can lead to improved segmentation accuracy, particularly, in the challenging boundary regions of expiration chest CT volumes. Furthermore, our study demonstrates the potential for LVC to enhance the model's generalization capabilities. Finally, the method's robustness is evaluated through the segmentation of lung lobes in 10 cases of COVID-19, demonstrating its applicability in the presence of pulmonary diseases. Conclusions Incorporating prior anatomical information, such as LVC, into the DL model shows promise for enhancing segmentation performance, particularly in the boundary regions. However, the extent of this improvement has limitations, prompting further exploration of its practical applicability.

Place, publisher, year, edition, pages
SPIE-Intl Soc Optical Eng, 2024
Keywords
pulmonary lobe segmentation, computed tomography, deep learning, 3D segmentation
National Category
Medical Imaging
Identifiers
urn:nbn:se:kth:diva-353003 (URN)10.1117/1.JMI.11.4.044001 (DOI)001304656700024 ()38988990 (PubMedID)2-s2.0-85202919207 (Scopus ID)
Note

QC 20240911

Available from: 2024-09-11 Created: 2024-09-11 Last updated: 2026-08-21Bibliographically approved
3. Added Value of Expiratory CT in Chronic Airflow Limitation
Open this publication in new window or tab >>Added Value of Expiratory CT in Chronic Airflow Limitation
Show others...
2026 (English)In: Respiratory Medicine, ISSN 0954-6111, E-ISSN 1532-3064, Vol. 263, article id 109098Article in journal (Refereed) Published
Abstract [en]

Background

Expiratory CT and lobar analysis may provide complementary information to inspiratory CT and spirometry for identifying chronic airflow limitation (CAL), but their value in population-based cohorts remains insufficiently characterised.

Methods

We analysed 2579 participants from the Stockholm cohort of the Swedish CArdioPulmonary bioImage Study, including 430 with CAL, defined as post-bronchodilator FEV1/FVC < 0.7. Inspiratory and expiratory chest CT scans were segmented into lung lobes using a deep-learning model. Total and lobar lung volumes, low-attenuation volume below −950 HU and below −856 HU, inspiratory − expiratory volume difference, and inspiratory fraction were calculated (inspiratory-expiratory volume/inspiratory volume). Discrimination of CAL was assessed using receiver operating characteristic analysis and regularised logistic regression.

Results

Participants with CAL had higher inspiratory and expiratory lung volumes and higher low-attenuation volumes than those without CAL, while inspiratory fraction was lower. Expiratory CT measures showed stronger discrimination for CAL than inspiratory measures. A model using inspiratory CT measures alone achieved an AUC of 0.66, compared with 0.77 when inspiratory and expiratory total lung measures were combined. Adding lobar measures further improved performance to an AUC of 0.81, with 79% sensitivity, 69% specificity, 34% positive predictive value and 94% negative predictive value.

Conclusions

Quantitative expiratory CT measures showed stronger discriminatory ability for CAL than inspiratory measures in this population-based cohort. Lobar analysis provided additional regional information and modestly improved classification performance, supporting integrated inspiratory–expiratory CT assessment for identifying CAL.

Place, publisher, year, edition, pages
Elsevier BV, 2026
Keywords
Chest CT, Chronic airflow limitation, COPD, Expiratory CT, Lung lobe segmentation, Quantitative computed tomography
National Category
Medical Imaging
Research subject
Medical Technology
Identifiers
urn:nbn:se:kth:diva-387486 (URN)10.1016/j.rmed.2026.109098 (DOI)42570787 (PubMedID)
Funder
University of GothenburgUmeå UniversitySwedish Heart Lung FoundationSwedish Research CouncilLund UniversityVinnovaKnut and Alice Wallenberg FoundationKarolinska InstituteUppsala UniversityLinköpings universitetStockholm County Council
Note

QC 20260907

Available from: 2026-08-21 Created: 2026-08-21 Last updated: 2026-09-07Bibliographically approved
4. Lobar Analysis of Classical and Deep Learning2 Radiomic Features in Chronic Airflow Limitation
Open this publication in new window or tab >>Lobar Analysis of Classical and Deep Learning2 Radiomic Features in Chronic Airflow Limitation
Show others...
(English)Manuscript (preprint) (Other academic)
Abstract [en]

Background: Quantitative chest computed tomography (CT) enables detailed assessment of structural lung abnormalities in chronic airflow limitation (CAL). We evaluated whether lobe-based classical radiomics and frozen deep learning transfer features from inspiratory and expiratory CT improve discrimination of CAL compared with handcrafted quantitative measures. Methods: In this population-based study, 2,579 participants underwent inspiratory and expiratory CT and post-bronchodilator spirometry. Chronic airflow limitation (CAL), defined as post-bronchodilator FEV₁/FVC <0.70, was the primary modeling endpoint and was present in 430 participants; 2,149 participants without CAL served as controls. Lung lobes were segmented using nnU-Net. Classical radiomic features were extracted with PyRadiomics, and deep learning transfer features were obtained from frozen 3D U-Net and SegResNet encoders pre-trained on the non-lung AbdomenAtlas dataset. A previously developed handcrafted feature set was included as a reference. Elastic-net regularized logistic regression with 10-fold cross-validation was used for classification, and performance was assessed by the area under the receiver operating characteristic curve (AUC). Results: Models based on expiratory features outperformed inspiratory-only models (AUC improvement 0.03–0.08). The best performance was achieved by classical radiomics combining inspiratory and expiratory features (AUC 0.87; positive predictive value 49%; negative predictive value 95%), exceeding the handcrafted model (AUC 0.81; PPV 34%; NPV 94%). Frozen transfer features from the non-lung-pretrained encoders showed slightly lower discrimination than classical radiomics, and combining feature types did not improve performance. Conclusion: Lobe-based classical radiomics from combined inspiratory and expiratory CT improved cross-validated discrimination of spirometry-defined CAL in this population-based cohort, with expiratory imaging contributing substantially to optimal performance.

Keywords
Radiomics; CT; deep learning; chronic airflow limitation
National Category
Medical Imaging
Research subject
Medical Technology
Identifiers
urn:nbn:se:kth:diva-387487 (URN)
Note

QC 20260821

Under review

Available from: 2026-08-21 Created: 2026-08-21 Last updated: 2026-08-21Bibliographically approved

Open Access in DiVA

Kappa(11520 kB)24 downloads
File information
File name FULLTEXT01.pdfFile size 11520 kBChecksum SHA-512
50afea616c015e16602e0e30c2915b5d5c28c499d8a20bb58e2f07ba818a07f32185af65b41425db99298cfdcf156b833aaa12d85c1daa808aa6ff74aa6f0918
Type fulltextMimetype application/pdf

Other links

Publisher's full texthttps://doi.org/10.69622/32509146

Authority records

Bäcklin, Emelie

Search in DiVA

By author/editor
Bäcklin, Emelie
By organisation
Medical Imaging
Medical Imaging

Search outside of DiVA

GoogleGoogle Scholar
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

doi
isbn
urn-nbn

Altmetric score

doi
isbn
urn-nbn
Total: 416 hits
910111213141512 of 22
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