Adopting transfer learning for neuroimaging : a comparative analysis with a custom 3D convolution neural network model
© 2022. The Author(s)..
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.
Medienart: |
E-Artikel |
---|
Erscheinungsjahr: |
2022 |
---|---|
Erschienen: |
2022 |
Enthalten in: |
Zur Gesamtaufnahme - volume:22 |
---|---|
Enthalten in: |
BMC medical informatics and decision making - 22(2022), Suppl 6 vom: 07. Dez., Seite 318 |
Sprache: |
Englisch |
---|
Links: |
---|
Anmerkungen: |
Date Completed 15.12.2022 Date Revised 10.01.2024 published: Electronic Citation Status MEDLINE |
---|
doi: |
10.1186/s12911-022-02054-7 |
---|
funding: |
|
---|---|
Förderinstitution / Projekttitel: |
|
PPN (Katalog-ID): |
NLM349890161 |
---|
LEADER | 01000caa a22002652 4500 | ||
---|---|---|---|
001 | NLM349890161 | ||
003 | DE-627 | ||
005 | 20240114233713.0 | ||
007 | cr uuu---uuuuu | ||
008 | 231226s2022 xx |||||o 00| ||eng c | ||
024 | 7 | |a 10.1186/s12911-022-02054-7 |2 doi | |
028 | 5 | 2 | |a pubmed24n1256.xml |
035 | |a (DE-627)NLM349890161 | ||
035 | |a (NLM)36476613 | ||
040 | |a DE-627 |b ger |c DE-627 |e rakwb | ||
041 | |a eng | ||
100 | 1 | |a Soliman, Amira |e verfasserin |4 aut | |
245 | 1 | 0 | |a Adopting transfer learning for neuroimaging |b a comparative analysis with a custom 3D convolution neural network model |
264 | 1 | |c 2022 | |
336 | |a Text |b txt |2 rdacontent | ||
337 | |a ƒaComputermedien |b c |2 rdamedia | ||
338 | |a ƒa Online-Ressource |b cr |2 rdacarrier | ||
500 | |a Date Completed 15.12.2022 | ||
500 | |a Date Revised 10.01.2024 | ||
500 | |a published: Electronic | ||
500 | |a Citation Status MEDLINE | ||
520 | |a © 2022. The Author(s). | ||
520 | |a 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 | ||
520 | |a 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 | ||
520 | |a 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 | ||
650 | 4 | |a Journal Article | |
650 | 4 | |a Research Support, N.I.H., Extramural | |
650 | 4 | |a Research Support, U.S. Gov't, Non-P.H.S. | |
650 | 4 | |a Research Support, Non-U.S. Gov't | |
650 | 4 | |a Brain Neurodegenerative Disorders | |
650 | 4 | |a Convolution Neural Networks | |
650 | 4 | |a Medical Image Classification | |
650 | 4 | |a Transfer Learning | |
700 | 1 | |a Chang, Jose R |e verfasserin |4 aut | |
700 | 1 | |a Etminani, Kobra |e verfasserin |4 aut | |
700 | 1 | |a Byttner, Stefan |e verfasserin |4 aut | |
700 | 1 | |a Davidsson, Anette |e verfasserin |4 aut | |
700 | 1 | |a Martínez-Sanchis, Begoña |e verfasserin |4 aut | |
700 | 1 | |a Camacho, Valle |e verfasserin |4 aut | |
700 | 1 | |a Bauckneht, Matteo |e verfasserin |4 aut | |
700 | 1 | |a Stegeran, Roxana |e verfasserin |4 aut | |
700 | 1 | |a Ressner, Marcus |e verfasserin |4 aut | |
700 | 1 | |a Agudelo-Cifuentes, Marc |e verfasserin |4 aut | |
700 | 1 | |a Chincarini, Andrea |e verfasserin |4 aut | |
700 | 1 | |a Brendel, Matthias |e verfasserin |4 aut | |
700 | 1 | |a Rominger, Axel |e verfasserin |4 aut | |
700 | 1 | |a Bruffaerts, Rose |e verfasserin |4 aut | |
700 | 1 | |a Vandenberghe, Rik |e verfasserin |4 aut | |
700 | 1 | |a Kramberger, Milica G |e verfasserin |4 aut | |
700 | 1 | |a Trost, Maja |e verfasserin |4 aut | |
700 | 1 | |a Nicastro, Nicolas |e verfasserin |4 aut | |
700 | 1 | |a Frisoni, Giovanni B |e verfasserin |4 aut | |
700 | 1 | |a Lemstra, Afina W |e verfasserin |4 aut | |
700 | 1 | |a Berckel, Bart N M van |e verfasserin |4 aut | |
700 | 1 | |a Pilotto, Andrea |e verfasserin |4 aut | |
700 | 1 | |a Padovani, Alessandro |e verfasserin |4 aut | |
700 | 1 | |a Morbelli, Silvia |e verfasserin |4 aut | |
700 | 1 | |a Aarsland, Dag |e verfasserin |4 aut | |
700 | 1 | |a Nobili, Flavio |e verfasserin |4 aut | |
700 | 1 | |a Garibotto, Valentina |e verfasserin |4 aut | |
700 | 0 | |a Alzheimer’s Disease Neuroimaging Initiative |e verfasserin |4 aut | |
700 | 1 | |a Ochoa-Figueroa, Miguel |e verfasserin |4 aut | |
773 | 0 | 8 | |i Enthalten in |t BMC medical informatics and decision making |d 2001 |g 22(2022), Suppl 6 vom: 07. Dez., Seite 318 |w (DE-627)NLM112392881 |x 1472-6947 |7 nnns |
773 | 1 | 8 | |g volume:22 |g year:2022 |g number:Suppl 6 |g day:07 |g month:12 |g pages:318 |
856 | 4 | 0 | |u http://dx.doi.org/10.1186/s12911-022-02054-7 |3 Volltext |
912 | |a GBV_USEFLAG_A | ||
912 | |a GBV_NLM | ||
951 | |a AR | ||
952 | |d 22 |j 2022 |e Suppl 6 |b 07 |c 12 |h 318 |