Open this publication in new window or tab >>Institution Of Health And Society, Linkoping, Sweden.
Hospital Universitario La Fe, Valencia, Spain.
Universitat Autònoma De Barcelona, Cerdanyola del Valles, Spain.
Irccs San Martino Polyclinic Hospital, Genoa, Italy.
University Hospital, Linkoping, Sweden.
University Hospital, Linkoping, Sweden.
Hospital Universitario La Fe, Valencia, Spain.
Infn Sezione Di Genova, Genoa, Italy.
Klinikum Der Universität München, Munich, Germany.
Inselspital, Bern, Switzerland.
University Of Antwerp, Antwerpen, Belgium.
Department Of Neurosciences, Leuven, Belgium; University Hospitals Leuven, Leuven, Belgium.
University Medical Centre Ljubljana, Ljubljana, Slovenia.
University Medical Centre Ljubljana, Ljubljana, Slovenia; Faculty Of Medicine, Ljubljana, Slovenia.
Geneva University Hospitals, Geneva, Switzerland.
Geneva University Hospitals, Geneva, Switzerland.
Vu University Medical Center, Amsterdam, Netherlands.
Vrije Universiteit Amsterdam, Amsterdam, Netherlands.
University Of Brescia, Brescia, Italy.
Irccs San Martino Polyclinic Hospital, Genoa, Italy.
Stavanger University Hospital, Stavanger, Norway.
University Of Genoa, Genoa, Italy; King's College London, London, United Kingdom.
University Of Genoa, Genoa, Italy.
Geneva University Hospitals, Geneva, Switzerland.
Institution Of Health And Society, Linkoping, Sweden; University Hospital, Linkoping, Sweden; Linköping University, Linkoping, Sweden.
Show others...
2022 (English)In: BMC Medical Informatics and Decision Making, E-ISSN 1472-6947, Vol. 22, p. 1-15, article id 318Article in journal (Refereed) Published
Abstract [en]
Background: In recent years, neuroimaging with deep learning (DL) algorithms have made remarkable advances in the diagnosis of neurodegenerative disorders. However, applying DL in different medical domains is usually challenged by lack of labeled data. To address this challenge, transfer learning (TL) has been applied to use state-of-the-art convolution neural networks pre-trained on natural images. Yet, there are differences in characteristics between medical and natural images, also image classification and targeted medical diagnosis tasks. The purpose of this study is to investigate the performance of specialized and TL in the classification of neurodegenerative disorders using 3D volumes of 18F-FDG-PET brain scans. Results: Results show that TL models are suboptimal for classification of neurodegenerative disorders, especially when the objective is to separate more than two disorders. Additionally, specialized CNN model provides better interpretations of predicted diagnosis. Conclusions: TL can indeed lead to superior performance on binary classification in timely and data efficient manner, yet for detecting more than a single disorder, TL models do not perform well. Additionally, custom 3D model performs comparably to TL models for binary classification, and interestingly perform better for diagnosis of multiple disorders. The results confirm the superiority of the custom 3D-CNN in providing better explainable model compared to TL adopted ones. © 2022, The Author(s).
Place, publisher, year, edition, pages
London: BioMed Central (BMC), 2022
Keywords
Brain Neurodegenerative Disorders, Convolution Neural Networks, Medical Image Classification, Transfer Learning
National Category
Computer Vision and Robotics (Autonomous Systems)
Identifiers
urn:nbn:se:hh:diva-49079 (URN)10.1186/s12911-022-02054-7 (DOI)000904994900001 ()36476613 (PubMedID)2-s2.0-85143570393 (Scopus ID)
2023-01-102023-01-102023-08-21Bibliographically approved