-
On the Impact of Cross-Domain Data on German Language Models
Authors:
Amin Dada,
Aokun Chen,
Cheng Peng,
Kaleb E Smith,
Ahmad Idrissi-Yaghir,
Constantin Marc Seibold,
Jianning Li,
Lars Heiliger,
Xi Yang,
Christoph M. Friedrich,
Daniel Truhn,
Jan Egger,
Jiang Bian,
Jens Kleesiek,
Yonghui Wu
Abstract:
Traditionally, large language models have been either trained on general web crawls or domain-specific data. However, recent successes of generative large language models, have shed light on the benefits of cross-domain datasets. To examine the significance of prioritizing data diversity over quality, we present a German dataset comprising texts from five domains, along with another dataset aimed…
▽ More
Traditionally, large language models have been either trained on general web crawls or domain-specific data. However, recent successes of generative large language models, have shed light on the benefits of cross-domain datasets. To examine the significance of prioritizing data diversity over quality, we present a German dataset comprising texts from five domains, along with another dataset aimed at containing high-quality data. Through training a series of models ranging between 122M and 750M parameters on both datasets, we conduct a comprehensive benchmark on multiple downstream tasks. Our findings demonstrate that the models trained on the cross-domain dataset outperform those trained on quality data alone, leading to improvements up to $4.45\%$ over the previous state-of-the-art. The models are available at https://huggingface.co/ikim-uk-essen
△ Less
Submitted 13 October, 2023; v1 submitted 11 October, 2023;
originally announced October 2023.
-
Multimodal Interactive Lung Lesion Segmentation: A Framework for Annotating PET/CT Images based on Physiological and Anatomical Cues
Authors:
Verena Jasmin Hallitschke,
Tobias Schlumberger,
Philipp Kataliakos,
Zdravko Marinov,
Moon Kim,
Lars Heiliger,
Constantin Seibold,
Jens Kleesiek,
Rainer Stiefelhagen
Abstract:
Recently, deep learning enabled the accurate segmentation of various diseases in medical imaging. These performances, however, typically demand large amounts of manual voxel annotations. This tedious process for volumetric data becomes more complex when not all required information is available in a single imaging domain as is the case for PET/CT data. We propose a multimodal interactive segmentat…
▽ More
Recently, deep learning enabled the accurate segmentation of various diseases in medical imaging. These performances, however, typically demand large amounts of manual voxel annotations. This tedious process for volumetric data becomes more complex when not all required information is available in a single imaging domain as is the case for PET/CT data. We propose a multimodal interactive segmentation framework that mitigates these issues by combining anatomical and physiological cues from PET/CT data. Our framework utilizes the geodesic distance transform to represent the user annotations and we implement a novel ellipsoid-based user simulation scheme during training. We further propose two annotation interfaces and conduct a user study to estimate their usability. We evaluated our model on the in-domain validation dataset and an unseen PET/CT dataset. We make our code publicly available: https://github.com/verena-hallitschke/pet-ct-annotate.
△ Less
Submitted 24 January, 2023;
originally announced January 2023.
-
AutoPET Challenge: Combining nn-Unet with Swin UNETR Augmented by Maximum Intensity Projection Classifier
Authors:
Lars Heiliger,
Zdravko Marinov,
Max Hasin,
André Ferreira,
Jana Fragemann,
Kelsey Pomykala,
Jacob Murray,
David Kersting,
Victor Alves,
Rainer Stiefelhagen,
Jan Egger,
Jens Kleesiek
Abstract:
Tumor volume and changes in tumor characteristics over time are important biomarkers for cancer therapy. In this context, FDG-PET/CT scans are routinely used for staging and re-staging of cancer, as the radiolabeled fluorodeoxyglucose is taken up in regions of high metabolism. Unfortunately, these regions with high metabolism are not specific to tumors and can also represent physiological uptake b…
▽ More
Tumor volume and changes in tumor characteristics over time are important biomarkers for cancer therapy. In this context, FDG-PET/CT scans are routinely used for staging and re-staging of cancer, as the radiolabeled fluorodeoxyglucose is taken up in regions of high metabolism. Unfortunately, these regions with high metabolism are not specific to tumors and can also represent physiological uptake by normal functioning organs, inflammation, or infection, making detailed and reliable tumor segmentation in these scans a demanding task. This gap in research is addressed by the AutoPET challenge, which provides a public data set with FDG-PET/CT scans from 900 patients to encourage further improvement in this field. Our contribution to this challenge is an ensemble of two state-of-the-art segmentation models, the nn-Unet and the Swin UNETR, augmented by a maximum intensity projection classifier that acts like a gating mechanism. If it predicts the existence of lesions, both segmentations are combined by a late fusion approach. Our solution achieves a Dice score of 72.12\% on patients diagnosed with lung cancer, melanoma, and lymphoma in our cross-validation. Code: https://github.com/heiligerl/autopet_submission
△ Less
Submitted 14 October, 2022; v1 submitted 2 September, 2022;
originally announced September 2022.
