DeepFake for Life Sciences

dc.contributor.authorLUONGO, Francesca
dc.contributor.authorNTAVELIS, Evangelos
dc.contributor.authorKASTANIS, Iason
dc.contributor.authorSCHMID, Philipp A.E
dc.date.accessioned2022-05-17T12:29:05Z
dc.date.available2022-05-17T12:29:05Z
dc.date.issued2022-05-16
dc.description.abstractSpheroids are three-dimensional cellular aggregates and one of the most common and versatile way to culture cells in 3D. In order to scale laboratory tests, automated processes are needed, including robust classification. The spheroids need to be sorted with a high accuracy as misclassification can pollute the entire batch of healthy spheroids. The automation often uses deep learning algorithms to sort healthy and unhealthy spheroids. Nevertheless, the spheroids grown under normal conditions exhibit a high-class imbalance. It is estimated that only 1-10% of unhealthy spheroids are contained in one cell culture batch. This imbalance can bias the classification algorithm. Classical data warping augmentations and weighted loss are state-of-the-art methods when dealing with class imbalance. However, they are limited to the extent they can increase the performance of the classifier. Thus, the underrepresented class is oversampled with synthetic images generated with Generative Adversarial Networks (GAN). The results show significant improvement of the classification performance of an imbalanced dataset with this novel method of data augmentation.en_US
dc.identifier.citationPDA conference 2022 poster presentationen_US
dc.identifier.urihttps://hdl.handle.net/20.500.12839/1018
dc.language.isoenen_US
dc.subjectdeep fakeen_US
dc.subjectGANsen_US
dc.subjectsynthetic imagesen_US
dc.subjectdeep learningen_US
dc.subjectgenerative data augmentationen_US
dc.subjectclassification performance improvementen_US
dc.titleDeepFake for Life Sciencesen_US
dc.typeConferenceen_US
dc.type.csemdivisionsDiv-Ren_US
dc.type.csemresearchareasData & AIen_US
dc.type.csemresearchareasDigital Healthen_US
dc.type.csemresearchareasIndustry 4.0en_US
dc.type.csemresearchareasTools for Life Sciencesen_US
Files
Original bundle
Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
CSEM_DeepFake for Life Sciences.pdf
Size:
3.89 MB
Format:
Adobe Portable Document Format
Description:
Poster
License bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
license.txt
Size:
2.82 KB
Format:
Item-specific license agreed upon to submission
Description: