Kris K. Dreher
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Kris K. Dreher has a total of 59 co-authors. Here are the top co-authors by number of shared publications:
- Lena Maier-Hein
German Cancer Research Center
9 shared publications - Alexander Seitel
German Cancer Research Center (DKFZ), Division of Intelligent Medical Systems, Heidelberg, Germany
8 shared publications - Jan-Hinrich Nölke
German Cancer Research Center (DKFZ), Division of Intelligent Medical Systems, Heidelberg, Germany
6 shared publications - Leonardo Ayala
German Cancer Research Center (DKFZ), Division of Intelligent Medical Systems, Heidelberg, Germany
5 shared publications - Janek Gröhl
Univ. of Cambridge (United Kingdom)
4 shared publications - Christoph Julien Bender
Division of Intelligent Medical Systems (IMSY), German Cancer Research Center (DKFZ), Heidelberg, Germany
3 shared publications
Latest Publications
- Open Access
- 1 March 2026
- Published by SPIE-Intl Soc Optical Eng in Journal of Biomedical Optics
- Vol. 31 (03) , 1
Learning to simulate realistic human diffuse reflectance spectra (Erratum)
The erratum list some minor errors in the originally published article, which have since been corrected.
- 1 March 2026
- Published by SPIE-Intl Soc Optical Eng in Journal of Biomedical Optics
- Vol. 31 (03) , 1
Open AccessLearning to simulate realistic human diffuse reflectance spectra (Erratum)
The erratum list some minor errors in the originally published article, which have since been corrected.
- Open Access
- 1 February 2026
- Published by SPIE-Intl Soc Optical Eng in Journal of Biomedical Optics
- Vol. 31 (02) , 1-20
Learning to simulate realistic human diffuse reflectance spectra
SignificanceHyperspectral imaging is a noninvasive, cost-effective modality with transformative clinical potential. Its adoption is limited by the lack of accurate and efficient methods that relate spectra to tissue parameters, essential for both AI training and validation of imaging methods, as gold standard Monte Carlo (MC) simulations remain prohibitively computationally expensive.AimWe aim to develop a scalable and accurate method for generating realistic tissue reflectance spectra in support of AI development and validation in biomedical imaging.ApproachWe trained a general-purpose neural surrogate model on >50 million MC simulations based on a flexible multilayer tissue model. We validated our model against >5000 open surgery in vivo hyperspectral images, annotated with 23 tissue classes for stratified performance analysis. In addition, we qualitatively evaluated clinical potential by testing whether surrogate-generated spectra enable recovery of organ-specific oxygenation dynamics in a controlled porcine aortic clamping experiment.ResultsThe surrogate model achieved accuracy matching MC simulations with 5–10 million photons while delivering inference five orders of magnitude faster. Across 140 million human tissue spectra, it improved spectral recall by 13–48 percentage points over existing surrogate models. Scaling analyses revealed a power law relationship between training dataset size and test error, enabling the prediction of training data requirements for target accuracy. Our porcine study suggests that the synthetic data generated with the surrogate model is suitable for recovering organ-specific stO2 trajectories.ConclusionNeural surrogate models can achieve MC-level accuracy and in vivo realism at negligible inference cost, enabling large-scale, compute-efficient data generation for biomedical optics and robust AI development for clinical applications.
- 1 February 2026
- Published by SPIE-Intl Soc Optical Eng in Journal of Biomedical Optics
- Vol. 31 (02) , 1-20
Open AccessLearning to simulate realistic human diffuse reflectance spectra
SignificanceHyperspectral imaging is a noninvasive, cost-effective modality with transformative clinical potential. Its adoption is limited by the lack of accurate and efficient methods that relate spectra to tissue parameters, essential for both AI training and validation of imaging methods, as gold standard Monte Carlo (MC) simulations remain prohibitively computationally expensive.AimWe aim to develop a scalable and accurate method for generating realistic tissue reflectance spectra in support of AI development and validation in biomedical imaging.ApproachWe trained a general-purpose neural surrogate model on >50 million MC simulations based on a flexible multilayer tissue model. We validated our model against >5000 open surgery in vivo hyperspectral images, annotated with 23 tissue classes for stratified performance analysis. In addition, we qualitatively evaluated clinical potential by testing whether surrogate-generated spectra enable recovery of organ-specific oxygenation dynamics in a controlled porcine aortic clamping experiment.ResultsThe surrogate model achieved accuracy matching MC simulations with 5–10 million photons while delivering inference five orders of magnitude faster. Across 140 million human tissue spectra, it improved spectral recall by 13–48 percentage points over existing surrogate models. Scaling analyses revealed a power law relationship between training dataset size and test error, enabling the prediction of training data requirements for target accuracy. Our porcine study suggests that the synthetic data generated with the surrogate model is suitable for recovering organ-specific stO2 trajectories.ConclusionNeural surrogate models can achieve MC-level accuracy and in vivo realism at negligible inference cost, enabling large-scale, compute-efficient data generation for biomedical optics and robust AI development for clinical applications.
- Open Access
- 1 July 2025
- Published by SPIE-Intl Soc Optical Eng in Journal of Biomedical Optics
- Vol. 30 (07) , 1-15
Anthropomorphic tissue-mimicking phantoms for oximetry validation in multispectral optical imaging
SignificanceOptical imaging of blood oxygenation (sO2) can be achieved based on the differential absorption spectra of oxy- and deoxyhemoglobin. A key challenge in realizing clinical validation of the sO2 biomarkers is the absence of reliable sO2 reference standards, including test objects.AimTo enable quantitative testing of multispectral imaging methods for assessment of sO2 by introducing anthropomorphic phantoms with appropriate tissue-mimicking optical properties.ApproachWe used the stable copolymer-in-oil base material to create physical anthropomorphic structures and optimized dyes to mimic the optical absorption of blood across a wide spectral range. Using 3D-printed phantom molds generated from a magnetic resonance image of a human forearm, we molded the material into an anthropomorphic shape. Using both reflectance hyperspectral imaging (HSI) and photoacoustic tomography (PAT), we acquired images of the forearm phantoms and evaluated the performance of linear spectral unmixing (LSU).ResultsBased on 10 fabricated forearm phantoms with vessel-like structures featuring five distinct sO2 levels (between 0 and 100%), we showed that the measured absorption spectra of the material correlated well with HSI and PAT data with a Pearson correlation coefficient consistently above 0.8. Further, the application of LSU enabled a quantification of the mean absolute error in sO2 assessment with HSI and PAT.ConclusionsOur anthropomorphic tissue-mimicking phantoms hold potential to provide a robust tool for developing, standardising, and validating optical imaging of sO2.
- 1 July 2025
- Published by SPIE-Intl Soc Optical Eng in Journal of Biomedical Optics
- Vol. 30 (07) , 1-15
Open AccessAnthropomorphic tissue-mimicking phantoms for oximetry validation in multispectral optical imaging
SignificanceOptical imaging of blood oxygenation (sO2) can be achieved based on the differential absorption spectra of oxy- and deoxyhemoglobin. A key challenge in realizing clinical validation of the sO2 biomarkers is the absence of reliable sO2 reference standards, including test objects.AimTo enable quantitative testing of multispectral imaging methods for assessment of sO2 by introducing anthropomorphic phantoms with appropriate tissue-mimicking optical properties.ApproachWe used the stable copolymer-in-oil base material to create physical anthropomorphic structures and optimized dyes to mimic the optical absorption of blood across a wide spectral range. Using 3D-printed phantom molds generated from a magnetic resonance image of a human forearm, we molded the material into an anthropomorphic shape. Using both reflectance hyperspectral imaging (HSI) and photoacoustic tomography (PAT), we acquired images of the forearm phantoms and evaluated the performance of linear spectral unmixing (LSU).ResultsBased on 10 fabricated forearm phantoms with vessel-like structures featuring five distinct sO2 levels (between 0 and 100%), we showed that the measured absorption spectra of the material correlated well with HSI and PAT data with a Pearson correlation coefficient consistently above 0.8. Further, the application of LSU enabled a quantification of the mean absolute error in sO2 assessment with HSI and PAT.ConclusionsOur anthropomorphic tissue-mimicking phantoms hold potential to provide a robust tool for developing, standardising, and validating optical imaging of sO2.
- Open Access
- 9 May 2025
- Published by Springer Nature in International Journal of Computer Assisted Radiology and Surgery
- Vol. 20 (7) , 1325-1333
Shortcut learning leads to sex bias in deep learning models for photoacoustic tomography
Purpose: Shortcut learning has been identified as a source of algorithmic unfairness in medical imaging artificial intelligence (AI), but its impact on photoacoustic tomography (PAT), particularly concerning sex bias, remains underexplored. This study investigates this issue using peripheral artery disease (PAD) diagnosis as a specific clinical application. Methods: To examine the potential for sex bias due to shortcut learning in convolutional neural network (CNNs) and assess how such biases might affect diagnostic predictions, we created training and test datasets with varying PAD prevalence between sexes. Using these datasets, we explored (1) whether CNNs can classify the sex from imaging data, (2) how sex-specific prevalence shifts impact PAD diagnosis performance and underdiagnosis disparity between sexes, and (3) how similarly CNNs encode sex and PAD features. Results: Our study with 147 individuals demonstrates that CNNs can classify the sex from calf muscle PAT images, achieving an AUROC of 0.75. For PAD diagnosis, models trained on data with imbalanced sex-specific disease prevalence experienced significant performance drops (up to 0.21 AUROC) when applied to balanced test sets. Additionally, greater imbalances in sex-specific prevalence within the training data exacerbated underdiagnosis disparities between sexes. Finally, we identify evidence of shortcut learning by demonstrating the effective reuse of learned feature representations between PAD diagnosis and sex classification tasks. Conclusion: CNN-based models trained on PAT data may engage in shortcut learning by leveraging sex-related features, leading to biased and unreliable diagnostic predictions. Addressing demographic-specific prevalence imbalances and preventing shortcut learning is critical for developing models in the medical field that are both accurate and equitable across diverse patient populations.
- 9 May 2025
- Published by Springer Nature in International Journal of Computer Assisted Radiology and Surgery
- Vol. 20 (7) , 1325-1333
Open AccessShortcut learning leads to sex bias in deep learning models for photoacoustic tomography
Purpose: Shortcut learning has been identified as a source of algorithmic unfairness in medical imaging artificial intelligence (AI), but its impact on photoacoustic tomography (PAT), particularly concerning sex bias, remains underexplored. This study investigates this issue using peripheral artery disease (PAD) diagnosis as a specific clinical application. Methods: To examine the potential for sex bias due to shortcut learning in convolutional neural network (CNNs) and assess how such biases might affect diagnostic predictions, we created training and test datasets with varying PAD prevalence between sexes. Using these datasets, we explored (1) whether CNNs can classify the sex from imaging data, (2) how sex-specific prevalence shifts impact PAD diagnosis performance and underdiagnosis disparity between sexes, and (3) how similarly CNNs encode sex and PAD features. Results: Our study with 147 individuals demonstrates that CNNs can classify the sex from calf muscle PAT images, achieving an AUROC of 0.75. For PAD diagnosis, models trained on data with imbalanced sex-specific disease prevalence experienced significant performance drops (up to 0.21 AUROC) when applied to balanced test sets. Additionally, greater imbalances in sex-specific prevalence within the training data exacerbated underdiagnosis disparities between sexes. Finally, we identify evidence of shortcut learning by demonstrating the effective reuse of learned feature representations between PAD diagnosis and sex classification tasks. Conclusion: CNN-based models trained on PAT data may engage in shortcut learning by leveraging sex-related features, leading to biased and unreliable diagnostic predictions. Addressing demographic-specific prevalence imbalances and preventing shortcut learning is critical for developing models in the medical field that are both accurate and equitable across diverse patient populations.
- Open Access
- 1 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Medical Imaging
- Vol. 43 (9) , 3366-3376
Photoacoustic Quantification of Tissue Oxygenation Using Conditional Invertible Neural Networks
Intelligent systems in interventional healthcare depend on the reliable perception of the environment. In this context, photoacoustic tomography (PAT) has emerged as a non-invasive, functional imaging modality with great clinical potential. Current research focuses on converting the high-dimensional, not human-interpretable spectral data into the underlying functional information, specifically the blood oxygenation. One of the largely unexplored issues stalling clinical advances is the fact that the quantification problem is ambiguous, i.e. that radically different tissue parameter configurations could lead to almost identical photoacoustic spectra. In the present work, we tackle this problem with conditional Invertible Neural Networks (cINNs). Going beyond traditional point estimates, our network is used to compute an approximation of the conditional posterior density of tissue parameters given the measurement. To this end, an automatic mode detection algorithm extracts the plausible solution from the sample-based posterior. According to a comprehensive validation study based on both synthetic and real images, our approach is well-suited for exploring ambiguity in quantitative PAT.
- 1 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Medical Imaging
- Vol. 43 (9) , 3366-3376
Open AccessPhotoacoustic Quantification of Tissue Oxygenation Using Conditional Invertible Neural Networks
Intelligent systems in interventional healthcare depend on the reliable perception of the environment. In this context, photoacoustic tomography (PAT) has emerged as a non-invasive, functional imaging modality with great clinical potential. Current research focuses on converting the high-dimensional, not human-interpretable spectral data into the underlying functional information, specifically the blood oxygenation. One of the largely unexplored issues stalling clinical advances is the fact that the quantification problem is ambiguous, i.e. that radically different tissue parameter configurations could lead to almost identical photoacoustic spectra. In the present work, we tackle this problem with conditional Invertible Neural Networks (cINNs). Going beyond traditional point estimates, our network is used to compute an approximation of the conditional posterior density of tissue parameters given the measurement. To this end, an automatic mode detection algorithm extracts the plausible solution from the sample-based posterior. According to a comprehensive validation study based on both synthetic and real images, our approach is well-suited for exploring ambiguity in quantitative PAT.
- Open Access
- 1 October 2023
- Published by Springer Nature
- p. 770-780
Unsupervised Domain Transfer with Conditional Invertible Neural Networks
Synthetic medical image generation has evolved as a key technique for neural network training and validation. A core challenge, however, remains in the domain gap between simulations and real data. While deep learning-based domain transfer using Cycle Generative Adversarial Networks and similar architectures has led to substantial progress in the field, there are use cases in which state-of-the-art approaches still fail to generate training images that produce convincing results on relevant downstream tasks. Here, we address this issue with a domain transfer approach based on conditional invertible neural networks (cINNs). As a particular advantage, our method inherently guarantees cycle consistency through its invertible architecture, and network training can efficiently be conducted with maximum likelihood training. To showcase our method’s generic applicability, we apply it to two spectral imaging modalities at different scales, namely hyperspectral imaging (pixel-level) and photoacoustic tomography (image-level). According to comprehensive experiments, our method enables the generation of realistic spectral data and outperforms the state of the art on two downstream classification tasks (binary and multi-class). cINN-based domain transfer could thus evolve as an important method for realistic synthetic data generation in the field of spectral imaging and beyond. The code is available at https://github.com/IMSY-DKFZ/UDT-cINN.
- 1 October 2023
- Published by Springer Nature
- p. 770-780
Open AccessUnsupervised Domain Transfer with Conditional Invertible Neural Networks
Synthetic medical image generation has evolved as a key technique for neural network training and validation. A core challenge, however, remains in the domain gap between simulations and real data. While deep learning-based domain transfer using Cycle Generative Adversarial Networks and similar architectures has led to substantial progress in the field, there are use cases in which state-of-the-art approaches still fail to generate training images that produce convincing results on relevant downstream tasks. Here, we address this issue with a domain transfer approach based on conditional invertible neural networks (cINNs). As a particular advantage, our method inherently guarantees cycle consistency through its invertible architecture, and network training can efficiently be conducted with maximum likelihood training. To showcase our method’s generic applicability, we apply it to two spectral imaging modalities at different scales, namely hyperspectral imaging (pixel-level) and photoacoustic tomography (image-level). According to comprehensive experiments, our method enables the generation of realistic spectral data and outperforms the state of the art on two downstream classification tasks (binary and multi-class). cINN-based domain transfer could thus evolve as an important method for realistic synthetic data generation in the field of spectral imaging and beyond. The code is available at https://github.com/IMSY-DKFZ/UDT-cINN.
- Open Access
Efficient Photoacoustic Image Synthesis with Deep Learning
Photoacoustic imaging potentially allows for the real-time visualization of functional human tissue parameters such as oxygenation but is subject to a challenging underlying quantification problem. While in silico studies have revealed the great potential of deep learning (DL) methodology in solving this problem, the inherent lack of an efficient gold standard method for model training and validation remains a grand challenge. This work investigates whether DL can be leveraged to accurately and efficiently simulate photon propagation in biological tissue, enabling photoacoustic image synthesis. Our approach is based on estimating the initial pressure distribution of the photoacoustic waves from the underlying optical properties using a back-propagatable neural network trained on synthetic data. In proof-of-concept studies, we validated the performance of two complementary neural network architectures, namely a conventional U-Net-like model and a Fourier Neural Operator (FNO) network. Our in silico validation on multispectral human forearm images shows that DL methods can speed up image generation by a factor of 100 when compared to Monte Carlo simulations with photons. While the FNO is slightly more accurate than the U-Net, when compared to Monte Carlo simulations performed with a reduced number of photons (), both neural network architectures achieve equivalent accuracy. In contrast to Monte Carlo simulations, the proposed DL models can be used as inherently differentiable surrogate models in the photoacoustic image synthesis pipeline, allowing for back-propagation of the synthesis error and gradient-based optimization over the entire pipeline. Due to their efficiency, they have the potential to enable large-scale training data generation that can expedite the clinical application of photoacoustic imaging.Open AccessEfficient Photoacoustic Image Synthesis with Deep Learning
Photoacoustic imaging potentially allows for the real-time visualization of functional human tissue parameters such as oxygenation but is subject to a challenging underlying quantification problem. While in silico studies have revealed the great potential of deep learning (DL) methodology in solving this problem, the inherent lack of an efficient gold standard method for model training and validation remains a grand challenge. This work investigates whether DL can be leveraged to accurately and efficiently simulate photon propagation in biological tissue, enabling photoacoustic image synthesis. Our approach is based on estimating the initial pressure distribution of the photoacoustic waves from the underlying optical properties using a back-propagatable neural network trained on synthetic data. In proof-of-concept studies, we validated the performance of two complementary neural network architectures, namely a conventional U-Net-like model and a Fourier Neural Operator (FNO) network. Our in silico validation on multispectral human forearm images shows that DL methods can speed up image generation by a factor of 100 when compared to Monte Carlo simulations with photons. While the FNO is slightly more accurate than the U-Net, when compared to Monte Carlo simulations performed with a reduced number of photons (), both neural network architectures achieve equivalent accuracy. In contrast to Monte Carlo simulations, the proposed DL models can be used as inherently differentiable surrogate models in the photoacoustic image synthesis pipeline, allowing for back-propagation of the synthesis error and gradient-based optimization over the entire pipeline. Due to their efficiency, they have the potential to enable large-scale training data generation that can expedite the clinical application of photoacoustic imaging. - Open Access
Unsupervised Domain Transfer with Conditional Invertible Neural Networks
Synthetic medical image generation has evolved as a key technique for neural network training and validation. A core challenge, however, remains in the domain gap between simulations and real data. While deep learning-based domain transfer using Cycle Generative Adversarial Networks and similar architectures has led to substantial progress in the field, there are use cases in which state-of-the-art approaches still fail to generate training images that produce convincing results on relevant downstream tasks. Here, we address this issue with a domain transfer approach based on conditional invertible neural networks (cINNs). As a particular advantage, our method inherently guarantees cycle consistency through its invertible architecture, and network training can efficiently be conducted with maximum likelihood training. To showcase our method's generic applicability, we apply it to two spectral imaging modalities at different scales, namely hyperspectral imaging (pixel-level) and photoacoustic tomography (image-level). According to comprehensive experiments, our method enables the generation of realistic spectral data and outperforms the state of the art on two downstream classification tasks (binary and multi-class). cINN-based domain transfer could thus evolve as an important method for realistic synthetic data generation in the field of spectral imaging and beyond.
Open AccessUnsupervised Domain Transfer with Conditional Invertible Neural Networks
Synthetic medical image generation has evolved as a key technique for neural network training and validation. A core challenge, however, remains in the domain gap between simulations and real data. While deep learning-based domain transfer using Cycle Generative Adversarial Networks and similar architectures has led to substantial progress in the field, there are use cases in which state-of-the-art approaches still fail to generate training images that produce convincing results on relevant downstream tasks. Here, we address this issue with a domain transfer approach based on conditional invertible neural networks (cINNs). As a particular advantage, our method inherently guarantees cycle consistency through its invertible architecture, and network training can efficiently be conducted with maximum likelihood training. To showcase our method's generic applicability, we apply it to two spectral imaging modalities at different scales, namely hyperspectral imaging (pixel-level) and photoacoustic tomography (image-level). According to comprehensive experiments, our method enables the generation of realistic spectral data and outperforms the state of the art on two downstream classification tasks (binary and multi-class). cINN-based domain transfer could thus evolve as an important method for realistic synthetic data generation in the field of spectral imaging and beyond.
- Open Access
- 4 April 2022
- Published by SPIE-Intl Soc Optical Eng in Journal of Biomedical Optics
- Vol. 27 (08) , 083010-21
SIMPA: an open-source toolkit for simulation and image processing for photonics and acoustics
Optical and acoustic imaging techniques enable real-time and noninvasive visualisation of structural and functional tissue properties without exposing the patient to harmful ionizing radiation. Nevertheless, the applicability of purely optical and acoustic imaging techniques is limited, for example, by the low penetration depth of near-infrared spectroscopy1 or by the difficulties of measuring functional tissue properties with ultrasound imaging.2 Furthermore, quantitative measurements are challenging as the state-of-the-art model-based approaches to solve the underlying inverse problems rely on assumptions that might not hold when applied to in vivo measurements.
- 4 April 2022
- Published by SPIE-Intl Soc Optical Eng in Journal of Biomedical Optics
- Vol. 27 (08) , 083010-21
Open AccessSIMPA: an open-source toolkit for simulation and image processing for photonics and acoustics
Optical and acoustic imaging techniques enable real-time and noninvasive visualisation of structural and functional tissue properties without exposing the patient to harmful ionizing radiation. Nevertheless, the applicability of purely optical and acoustic imaging techniques is limited, for example, by the low penetration depth of near-infrared spectroscopy1 or by the difficulties of measuring functional tissue properties with ultrasound imaging.2 Furthermore, quantitative measurements are challenging as the state-of-the-art model-based approaches to solve the underlying inverse problems rely on assumptions that might not hold when applied to in vivo measurements.
- Top Cited PapersOpen Access
- 1 June 2021
- Published by Elsevier in Photoacoustics
- Vol. 22, 100241
Deep learning for biomedical photoacoustic imaging: A review
- 1 June 2021
- Published by Elsevier in Photoacoustics
- Vol. 22, 100241
Top Cited PapersOpen AccessDeep learning for biomedical photoacoustic imaging: A review