arXiv subjects

Fabian Kiessling

Publications and source records attributed to Fabian Kiessling.

8 recordsLinked to original sources

arXiv2026

CoM$^3$eT: A foundation model for medical image analysis through federated, multidimensional context integration

Medical foundation models improve generalization when training AI models with limited labeled data, but remain confined to a single specialty, such as pathology or radiology, and to either sparse or dense outputs, such as classification or segmentation. Here, we present CoM$^3$eT (Co-representation Multidimensional Multitask Medical Transformer), a medical vision foundation model that unifies pathology and radiology, sparse and dense predictions, and two- and higher-dimensional inputs by modeling multidimensional context with attention. CoM$^3$eT outperformed other medical foundation models in an open competition spanning five tomographic, four whole-specimen, and three two-dimensional datasets, covering sparse and dense prediction tasks as well as report generation. When adapted across diverse clinical applications, training fewer than 2.5% of parameters achieved performance comparable to full fine-tuning, enabling research without access to high-performance GPU clusters. Applied to federated learning across hospitals, this approach achieved performance comparable to pooled-data training over internet connections and with consumer-grade hardware.

cs.CV↗

arXiv2024

Tissue Concepts: supervised foundation models in computational pathology

Due to the increasing workload of pathologists, the need for automation to support diagnostic tasks and quantitative biomarker evaluation is becoming more and more apparent. Foundation models have the potential to improve generalizability within and across centers and serve as starting points for data efficient development of specialized yet robust AI models. However, the training foundation models themselves is usually very expensive in terms of data, computation, and time. This paper proposes a supervised training method that drastically reduces these expenses. The proposed method is based on multi-task learning to train a joint encoder, by combining 16 different classification, segmentation, and detection tasks on a total of 912,000 patches. Since the encoder is capable of capturing the properties of the samples, we term it the Tissue Concepts encoder. To evaluate the performance and generalizability of the Tissue Concepts encoder across centers, classification of whole slide images from four of the most prevalent solid cancers - breast, colon, lung, and prostate - was used. The experiments show that the Tissue Concepts model achieve comparable performance to models trained with self-supervision, while requiring only 6% of the amount of training patches. Furthermore, the Tissue Concepts encoder outperforms an ImageNet pre-trained encoder on both in-domain and out-of-domain data.

eess.IV↗

arXiv2024

Source Identification by Consensus-Based Optimization

A consensus-based optimization (CBO) algorithm, which enables derivative and mesh-free optimization, is presented to localize a bioluminescent source. The light propagation is modeled by the radiative transfer equation approximated by spherical harmonics. The approach is investigated for a hierarchy of simplified diffusion models in simulated environments and tissue-mimicking phantoms. In simulations, the state-of-the-art diffusive approximation gives reliable results for heavily scattering media. However, higher-order models achieve better localization and more accurate source intensities for deeper sources and in the presence of artificial noise in strongly absorbing, but only moderately scattering media. In phantoms, higher-order models give lower approximation errors and the most accurate localization, even for a high scattering coefficient. These results demonstrate the potential of CBO to render higher-order models at lower computational cost while ensuring accurate localization in bioluminescence tomography.

q-bio.QM↗

arXiv2023

Overcoming Data Scarcity in Biomedical Imaging with a Foundational Multi-Task Model

Foundational models, pretrained on a large scale, have demonstrated substantial success across non-medical domains. However, training these models typically requires large, comprehensive datasets, which contrasts with the smaller and more heterogeneous datasets common in biomedical imaging. Here, we propose a multi-task learning strategy that decouples the number of training tasks from memory requirements. We trained a Universal bioMedical PreTrained model (UMedPT) on a multi-task database including tomographic, microscopic, and X-ray images, with various labelling strategies such as classification, segmentation, and object detection. The UMedPT foundational model outperformed ImageNet pretraining and the previous state-of-the-art models. For tasks related to the pretraining database, it maintained its performance with only 1% of the original training data and without fine-tuning. For out-of-domain tasks it required not more than 50% of the original training data. In an external independent validation imaging features extracted using UMedPT proved to be a new standard for cross-center transferability.

cs.CV↗

arXiv2022

Cell barrier characterization in transwell inserts by electrical impedance spectroscopy

We describe an impedance-based method for cell barrier integrity testing. A four-electrode electrical impedance spectroscopy (EIS) setup can be realized by simply connecting a commercial chopstick-like electrode (STX-1) to a potentiostat allowing monitoring cell barriers cultivated in transwell inserts. Subsequent electric circuit modeling of the electrical impedance results the capacitive properties of the barrier next to the well-known transepithelial electrical resistance (TEER). The versatility of the new method was analyzed by the EIS analysis of a Caco-2 monolayer in response to (a) different membrane coating materials, (b) two different permeability enhancers ethylene glycol-bis(2-aminoethylether)-N,N,N',N'-tetraacetic acid (EGTA) and saponin, and (c) sonoporation. For the different membrane coating materials, the TEERs of the standard and new protocol coincide and increase during cultivation, while the capacitance shows a distinct maximum for three different surface materials (no coating, Matrigel, and collagen I). The permeability enhancers cause a decline in the TEER value, but only saponin alters the capacitance of the cell layer by two orders of magnitude. Hence, cell layer capacitance and TEER represent two independent properties characterizing the monolayer. The use of commercial chopstick-like electrodes to access the impedance of a barrier cultivated in transwell inserts enables remarkable insight into the behavior of the cellular barrier with no extra work for the researcher. This simple method could evolve into a standard protocol used in cell barrier research.

physics.app-ph↗

arXiv2022

Image prediction of disease progression by style-based manifold extrapolation

Disease-modifying management aims to prevent deterioration and progression of the disease, not just relieve symptoms. Unfortunately, the development of necessary therapies is often hampered by the failure to recognize the presymptomatic disease and limited understanding of disease development. We present a generic solution for this problem by a methodology that allows the prediction of progression risk and morphology in individuals using a latent extrapolation optimization approach. To this end, we combined a regularized generative adversarial network (GAN) and a latent nearest neighbor algorithm for joint optimization to generate plausible images of future time points. We evaluated our method on osteoarthritis (OA) data from a multi-center longitudinal study (the Osteoarthritis Initiative, OAI). With presymptomatic baseline data, our model is generative and significantly outperforms the end-to-end learning model in discriminating the progressive cohort. Two experiments were performed with seven experienced radiologists. When no synthetic follow-up radiographs were provided, our model performed better than all seven radiologists. In cases where the synthetic follow-ups generated by our model were available, the specificity and sensitivity of all readers in discriminating progressors increased from $72.3\%$ to $88.6\%$ and from $42.1\%$ to $51.6\%$, respectively. Our results open up a new possibility of using model-based morphology and risk prediction to make predictions about future disease occurrence, as demonstrated in the example of OA.

eess.IV↗

arXiv2020

Advancing diagnostic performance and clinical usability of neural networks via adversarial training and dual batch normalization

Unmasking the decision-making process of machine learning models is essential for implementing diagnostic support systems in clinical practice. Here, we demonstrate that adversarially trained models can significantly enhance the usability of pathology detection as compared to their standard counterparts. We let six experienced radiologists rate the interpretability of saliency maps in datasets of X-rays, computed tomography, and magnetic resonance imaging scans. Significant improvements were found for our adversarial models, which could be further improved by the application of dual batch normalization. Contrary to previous research on adversarially trained models, we found that the accuracy of such models was equal to standard models when sufficiently large datasets and dual batch norm training were used. To ensure transferability, we additionally validated our results on an external test set of 22,433 X-rays. These findings elucidate that different paths for adversarial and real images are needed during training to achieve state of the art results with superior clinical interpretability.

cs.LG↗

arXiv2016

Initial PET Performance Evaluation of a Preclinical Insert for PET/MRI with Digital SiPM Technology

Hyperion-IID is a positron emission tomography (PET) insert which allows simultaneous operation in a clinical magnetic resonance imaging (MRI) scanner. To read out the scintillation light of the employed LYSO crystal arrays with a pitch of 1 mm pitch and 12 mm in height, digital silicon photomultipliers (DPC 3200-22, Philips Digital Photon Counting) (DPC) are used. The basic PET performance in terms of energy resolution, coincidence resolution time (CRT) and sensitivity as a function of operating parameters, such as the operating temperature, the applied overvoltage, activity and configuration parameters of the DPCs, were evaluated on system level. The measured energy resolution did not show a large dependency on the selected parameters and is in the range of 12.4-12.9% for low activities and degrades to ~13.6% at activities of ~100 MBq. The CRT strongly depends on the selected trigger scheme (trig) of the DPCs. We measured approximately 260 ps, 440 ps, 540 ps and 1300 ps for trig 1-4, respectively. The trues sensitivity for a NEMA NU 4 mouse-sized scatter phantom with a 70-mm-long tube of activity was dependent on the operating parameters and was determined to be 0.4-1.4% at low activities. The random fraction stayed below 5% at activities up to 100 MBq and the scatter fraction was evaluated as ~6% for an energy window of 411-561 keV and ~16% for 250-625 keV. Furthermore, we performed imaging experiments using a mouse-sized hot-rod phantom and a large rabbit-sized phantom. In 2D slices of the reconstructed mouse-sized hot-rod phantom (Ø = 28 mm), the rods were distinguishable from each other down to a rod size of 0.8 mm. There was no benefit of the better CRT of trig 1 over trig 3, where in the larger rabbit-sized phantom (Ø = 114 mm), we could show a clear improvement of image quality using the time-of-flight information.

physics.ins-det↗