Foundation model for cancer imaging biomarkers
© The Author(s) 2024..
Foundation models in deep learning are characterized by a single large-scale model trained on vast amounts of data serving as the foundation for various downstream tasks. Foundation models are generally trained using self-supervised learning and excel in reducing the demand for training samples in downstream applications. This is especially important in medicine, where large labelled datasets are often scarce. Here, we developed a foundation model for cancer imaging biomarker discovery by training a convolutional encoder through self-supervised learning using a comprehensive dataset of 11,467 radiographic lesions. The foundation model was evaluated in distinct and clinically relevant applications of cancer imaging-based biomarkers. We found that it facilitated better and more efficient learning of imaging biomarkers and yielded task-specific models that significantly outperformed conventional supervised and other state-of-the-art pretrained implementations on downstream tasks, especially when training dataset sizes were very limited. Furthermore, the foundation model was more stable to input variations and showed strong associations with underlying biology. Our results demonstrate the tremendous potential of foundation models in discovering new imaging biomarkers that may extend to other clinical use cases and can accelerate the widespread translation of imaging biomarkers into clinical settings.
Medienart: |
E-Artikel |
---|
Erscheinungsjahr: |
2024 |
---|---|
Erschienen: |
2024 |
Enthalten in: |
Zur Gesamtaufnahme - volume:6 |
---|---|
Enthalten in: |
Nature machine intelligence - 6(2024), 3 vom: 22., Seite 354-367 |
Sprache: |
Englisch |
---|
Beteiligte Personen: |
Pai, Suraj [VerfasserIn] |
---|
Links: |
---|
Themen: |
---|
Anmerkungen: |
Date Revised 26.03.2024 published: Print-Electronic Citation Status PubMed-not-MEDLINE |
---|
doi: |
10.1038/s42256-024-00807-9 |
---|
funding: |
|
---|---|
Förderinstitution / Projekttitel: |
|
PPN (Katalog-ID): |
NLM370132459 |
---|
LEADER | 01000caa a22002652 4500 | ||
---|---|---|---|
001 | NLM370132459 | ||
003 | DE-627 | ||
005 | 20240327000058.0 | ||
007 | cr uuu---uuuuu | ||
008 | 240325s2024 xx |||||o 00| ||eng c | ||
024 | 7 | |a 10.1038/s42256-024-00807-9 |2 doi | |
028 | 5 | 2 | |a pubmed24n1349.xml |
035 | |a (DE-627)NLM370132459 | ||
035 | |a (NLM)38523679 | ||
040 | |a DE-627 |b ger |c DE-627 |e rakwb | ||
041 | |a eng | ||
100 | 1 | |a Pai, Suraj |e verfasserin |4 aut | |
245 | 1 | 0 | |a Foundation model for cancer imaging biomarkers |
264 | 1 | |c 2024 | |
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 Revised 26.03.2024 | ||
500 | |a published: Print-Electronic | ||
500 | |a Citation Status PubMed-not-MEDLINE | ||
520 | |a © The Author(s) 2024. | ||
520 | |a Foundation models in deep learning are characterized by a single large-scale model trained on vast amounts of data serving as the foundation for various downstream tasks. Foundation models are generally trained using self-supervised learning and excel in reducing the demand for training samples in downstream applications. This is especially important in medicine, where large labelled datasets are often scarce. Here, we developed a foundation model for cancer imaging biomarker discovery by training a convolutional encoder through self-supervised learning using a comprehensive dataset of 11,467 radiographic lesions. The foundation model was evaluated in distinct and clinically relevant applications of cancer imaging-based biomarkers. We found that it facilitated better and more efficient learning of imaging biomarkers and yielded task-specific models that significantly outperformed conventional supervised and other state-of-the-art pretrained implementations on downstream tasks, especially when training dataset sizes were very limited. Furthermore, the foundation model was more stable to input variations and showed strong associations with underlying biology. Our results demonstrate the tremendous potential of foundation models in discovering new imaging biomarkers that may extend to other clinical use cases and can accelerate the widespread translation of imaging biomarkers into clinical settings | ||
650 | 4 | |a Journal Article | |
650 | 4 | |a Biomarkers | |
650 | 4 | |a Cancer imaging | |
650 | 4 | |a Tumour biomarkers | |
700 | 1 | |a Bontempi, Dennis |e verfasserin |4 aut | |
700 | 1 | |a Hadzic, Ibrahim |e verfasserin |4 aut | |
700 | 1 | |a Prudente, Vasco |e verfasserin |4 aut | |
700 | 1 | |a Sokač, Mateo |e verfasserin |4 aut | |
700 | 1 | |a Chaunzwa, Tafadzwa L |e verfasserin |4 aut | |
700 | 1 | |a Bernatz, Simon |e verfasserin |4 aut | |
700 | 1 | |a Hosny, Ahmed |e verfasserin |4 aut | |
700 | 1 | |a Mak, Raymond H |e verfasserin |4 aut | |
700 | 1 | |a Birkbak, Nicolai J |e verfasserin |4 aut | |
700 | 1 | |a Aerts, Hugo J W L |e verfasserin |4 aut | |
773 | 0 | 8 | |i Enthalten in |t Nature machine intelligence |d 2019 |g 6(2024), 3 vom: 22., Seite 354-367 |w (DE-627)NLM294259147 |x 2522-5839 |7 nnns |
773 | 1 | 8 | |g volume:6 |g year:2024 |g number:3 |g day:22 |g pages:354-367 |
856 | 4 | 0 | |u http://dx.doi.org/10.1038/s42256-024-00807-9 |3 Volltext |
912 | |a GBV_USEFLAG_A | ||
912 | |a GBV_NLM | ||
951 | |a AR | ||
952 | |d 6 |j 2024 |e 3 |b 22 |h 354-367 |