Generative AI in glioma: Ensuring diversity in training image phenotypes to improve diagnostic performance for IDH mutation prediction
Neuro-Oncology · 2024
DOI: 10.1093/neuonc/noae012 ↗Kurz gefasst
Fachartikel in der Zeitschrift Neuro-Oncology, 2024. Verfasst von Hye Hyeon Moon und 9 weiteren Autor:innen. Beteiligte Einrichtung: Department of Radiology and Research Institute of Radiology, University of Ulsan College of Medicine, Asan Medical Center, Seoul, Korea. Hauptthemen laut MEDLINE: Artificial Intelligence, Brain Neoplasms, Glioma, Isocitrate Dehydrogenase. Die Arbeit wurde bisher 19-mal zitiert (Crossref) und nennt 36 Literaturangaben. Der Volltext ist frei in PubMed Central verfügbar. Gefördert von: Korean government, Ministry of Health & Welfare, Republic of Korea.
Abstract (Originalsprache)
Background
This study evaluated whether generative artificial intelligence (AI)-based augmentation (GAA) can provide diverse and realistic imaging phenotypes and improve deep learning-based classification of isocitrate dehydrogenase (IDH) type in glioma compared with neuroradiologists.
Methods
For model development, 565 patients (346 IDH-wildtype, 219 IDH-mutant) with paired contrast-enhanced T1 and FLAIR MRI scans were collected from tertiary hospitals and The Cancer Imaging Archive. Performance was tested on internal (119, 78 IDH-wildtype, 41 IDH-mutant [IDH1 and 2]) and external test sets (108, 72 IDH-wildtype, 36 IDH-mutant). GAA was developed using a score-based diffusion model and ResNet50 classifier. The optimal GAA was selected in comparison with the null model. Two neuroradiologists (R1, R2) assessed realism, diversity of imaging phenotypes, and predicted IDH mutation. The performance of a classifier trained with optimal GAA was compared with that of neuroradiologists using the area under the receiver operating characteristics curve (AUC). The effect of tumor size and contrast enhancement on GAA performance was tested.
Results
Generated images demonstrated realism (Turing’s test: 47.5–50.5%) and diversity indicating IDH type. Optimal GAA was achieved with augmentation with 110 000 generated slices (AUC: 0.938). The classifier trained with optimal GAA demonstrated significantly higher AUC values than neuroradiologists in both the internal (R1, P = .003; R2, P < .001) and external test sets (R1, P < .01; R2, P < .001). GAA with large-sized tumors or predominant enhancement showed comparable performance to optimal GAA (internal test: AUC 0.956 and 0.922; external test: 0.810 and 0.749).
Conclusions
The application of generative AI with realistic and diverse images provided better diagnostic performance than neuroradiologists for predicting IDH type in glioma.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Substanzen
Autor:innen und Institutionen
- Department of Radiology and Research Institute of Radiology, University of Ulsan College of Medicine, Asan Medical Center, Seoul, Korea.
- Department of Convergence Medicine, University of Ulsan College of Medicine, Asan Medical Center, Seoul, Korea.
- Department of Biomedical Engineering, Asan Medical Institute of Convergence Science of Technology, Asan Medical Center, College of Medicine, University of Ulsan, Seoul, Korea.
- Department of Neurosurgery, University of Ulsan College of Medicine, Asan Medical Center, Seoul, Korea.
ORCID-Profile: Hye Hyeon Moon, Jiheon Jeong, Ji Eun Park, Namkug Kim, Young-Hoon Kim, Sang Woo Song, Chang-Ki Hong, Jeong Hoon Kim, Ho Sung Kim
Förderung
- Korean government
- Ministry of Health & Welfare, Republic of Korea
Zitieren
APA
Moon, H. H., Jeong, J., Park, J. E., Kim, N., Choi, C., Kim, Y. H., Song, S. W., Hong, C. K., Kim, J. H., & Kim, H. S. (2024). Generative AI in glioma: Ensuring diversity in training image phenotypes to improve diagnostic performance for IDH mutation prediction. Neuro-Oncology. https://doi.org/10.1093/neuonc/noae012
Vancouver
Moon HH, Jeong J, Park JE, Kim N, Choi C, Kim YH, et al. Generative AI in glioma: Ensuring diversity in training image phenotypes to improve diagnostic performance for IDH mutation prediction. Neuro-Oncology. 2024. doi:10.1093/neuonc/noae012
BibTeX
@article{Moon2024,
author = {Moon, Hye Hyeon and Jeong, Jiheon and Park, Ji Eun and Kim, Namkug and Choi, Changyong and Kim, Young-Hoon and Song, Sang Woo and Hong, Chang-Ki and Kim, Jeong Hoon and Kim, Ho Sung},
title = {Generative AI in glioma: Ensuring diversity in training image phenotypes to improve diagnostic performance for IDH mutation prediction},
journal = {Neuro-Oncology},
year = {2024},
doi = {10.1093/neuonc/noae012}
}
EUVIMED