Katherine J. Hewitt
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Katherine J. Hewitt has a total of 81 co-authors. Here are the top co-authors by number of shared publications:
- Jakob Nikolas Kather
Else Kroener Fresenius Center for Digital Health, Technical University Dresden, Dresden, Germany
31 shared publications - Laura Žigutytė
TUD Dresden University of Technology 1 Else Kröner Fresenius Center for Digital Health (EKFZ), Faculty of Medicine and University Hospital Carl Gustav Carus Dresden, Germany
26 shared publications - Nic Gabriel Reitsam
Pathology, Faculty of Medicine University of Augsburg Augsburg Germany
26 shared publications - Michaela Unger
TUD Dresden University of Technology 1 Else Kröner Fresenius Center for Digital Health (EKFZ), Faculty of Medicine and University Hospital Carl Gustav Carus Dresden, Germany
25 shared publications - Sebastian Foersch
Institute of Pathology, University Medical Center Mainz, Mainz, Germany
25 shared publications - Tim Lenz
Else Kroener Fresenius Center for Digital Health, Faculty of Medicine, TUD Dresden University of Technology, Dresden, Germany
25 shared publications
Latest Publications
- Open Access
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Figure S8 from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
NonMSIH to MSIH counterfactuals with cell segmentation
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Open AccessFigure S8 from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
NonMSIH to MSIH counterfactuals with cell segmentation
- Open Access
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Figure S9 from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
Counterfactuals for borderline vs. top-contributing MSI tiles
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Open AccessFigure S9 from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
Counterfactuals for borderline vs. top-contributing MSI tiles
- Open Access
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Table 1 from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
Evaluation of diffusion autoencoders trained on different datasets.
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Open AccessTable 1 from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
Evaluation of diffusion autoencoders trained on different datasets.
- Open Access
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Data from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
Deep learning can extract predictive and prognostic biomarkers from histopathology whole-slide images. However, explainable artificial intelligence approaches widely used in digital pathology, such as attention heatmaps and class activation mapping, provide limited insight into the image features associated with classifier outputs. In this study, we developed Morphing histoPathology Diffusion (MoPaDi), a framework for generating counterfactual explanations for histopathology images that help identify morphologic or stain-related features linked to model predictions. MoPaDi combined diffusion autoencoders with task-specific multiple instance learning classifiers to manipulate images and induce prediction shifts by modifying classifier-associated features. The framework was evaluated on multiple datasets spanning colorectal, breast, liver, and lung cancers, including tasks for tissue type, cancer subtype, and biomarker [microsatellite instability (MSI)] classification. MoPaDi generated perceptually realistic counterfactual histopathology images, enabling pathologists to identify morphologic features associated with changes in model predictions, complementing the conventional inspection of highly attended regions in digital pathology. In the MSI status prediction task, MoPaDi highlighted morphologic features linked to classifier predictions, including mucinous differentiation, altered glandular architecture, and lymphocytic infiltration, consistent with prior literature. Analyses separating stain-related from morphology-related components suggested that in this setting, prediction changes were predominantly associated with morphology-related rather than stain-related alterations. Overall, MoPaDi is a practical framework for counterfactual explanations in computational pathology that supports the evaluation of model-specific decision cues and hypothesis generation. Significance:MoPaDi is a diffusion-based tool for counterfactual image generation in cancer histopathology that reveals features associated with deep learning classifier predictions and supports transparent auditing of computational models in biomedical research.
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Open AccessData from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
Deep learning can extract predictive and prognostic biomarkers from histopathology whole-slide images. However, explainable artificial intelligence approaches widely used in digital pathology, such as attention heatmaps and class activation mapping, provide limited insight into the image features associated with classifier outputs. In this study, we developed Morphing histoPathology Diffusion (MoPaDi), a framework for generating counterfactual explanations for histopathology images that help identify morphologic or stain-related features linked to model predictions. MoPaDi combined diffusion autoencoders with task-specific multiple instance learning classifiers to manipulate images and induce prediction shifts by modifying classifier-associated features. The framework was evaluated on multiple datasets spanning colorectal, breast, liver, and lung cancers, including tasks for tissue type, cancer subtype, and biomarker [microsatellite instability (MSI)] classification. MoPaDi generated perceptually realistic counterfactual histopathology images, enabling pathologists to identify morphologic features associated with changes in model predictions, complementing the conventional inspection of highly attended regions in digital pathology. In the MSI status prediction task, MoPaDi highlighted morphologic features linked to classifier predictions, including mucinous differentiation, altered glandular architecture, and lymphocytic infiltration, consistent with prior literature. Analyses separating stain-related from morphology-related components suggested that in this setting, prediction changes were predominantly associated with morphology-related rather than stain-related alterations. Overall, MoPaDi is a practical framework for counterfactual explanations in computational pathology that supports the evaluation of model-specific decision cues and hypothesis generation. Significance:MoPaDi is a diffusion-based tool for counterfactual image generation in cancer histopathology that reveals features associated with deep learning classifier predictions and supports transparent auditing of computational models in biomedical research.
- Open Access
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Figure 4 from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
Independent classifier validation and representative bidirectional counterfactual transformations in a multiclass setting. A, External classifier responses to counterfactual morphing. Counterfactual images were generated at increasing morphing amplitudes (α) with MoPaDi and then encoded with three foundation models (UNI2, CONCH, and Virchow2). Independent classifiers were trained on the corresponding encoders’ features extracted from all TCGA-CRC tiles and then used to predict the target probability P(target) on both the original and counterfactual tiles. The resulting change ΔP(target) reflects how strongly the morphing affected class evidence. Bars show the median ΔP(target) with percentile-based variability across tiles. B, Representative examples of bidirectional counterfactual explanations for MSIL patients. We defined MSIL by fitting a two-component Gaussian mixture model to the log-transformed distribution of total MSI events and using the intersection of the two components as the cutoff separating MSIL from MSIH samples.
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Open AccessFigure 4 from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
Independent classifier validation and representative bidirectional counterfactual transformations in a multiclass setting. A, External classifier responses to counterfactual morphing. Counterfactual images were generated at increasing morphing amplitudes (α) with MoPaDi and then encoded with three foundation models (UNI2, CONCH, and Virchow2). Independent classifiers were trained on the corresponding encoders’ features extracted from all TCGA-CRC tiles and then used to predict the target probability P(target) on both the original and counterfactual tiles. The resulting change ΔP(target) reflects how strongly the morphing affected class evidence. Bars show the median ΔP(target) with percentile-based variability across tiles. B, Representative examples of bidirectional counterfactual explanations for MSIL patients. We defined MSIL by fitting a two-component Gaussian mixture model to the log-transformed distribution of total MSI events and using the intersection of the two components as the cutoff separating MSIL from MSIH samples.
- Open Access
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Figure 7 from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
Counterfactual image examples generated for the lung and breast cancer–type classifiers. A, Representative examples of LUSC tile transitioning to its counterfactual lung adenocarcinoma (LUAD) image and vice versa. B, Morphologic feature prevalence in original and counterfactual image pairs (N = 32 transitions; 16 tiles for each class). Horizontal bars show the percentage of image pairs in which at least one of three raters (three board-certified pathologists) identified each morphologic feature as present in the original (dark gray) or counterfactual (light gray) image. Right, Mean pairwise inter-rater agreement (Cohen κ) per feature, computed across all pairs and both directions combined. C, Counterfactual image generation effectiveness, measured as the percentage of generated images predicted as the opposite class across varying manipulation amplitudes. D, Representative examples of ILC tile transitioning to its counterfactual IDC image and vice versa. Difference maps display pixelwise differences between the original and the synthetic tile. Scale bar applies to all images within the panel unless otherwise indicated.
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Open AccessFigure 7 from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
Counterfactual image examples generated for the lung and breast cancer–type classifiers. A, Representative examples of LUSC tile transitioning to its counterfactual lung adenocarcinoma (LUAD) image and vice versa. B, Morphologic feature prevalence in original and counterfactual image pairs (N = 32 transitions; 16 tiles for each class). Horizontal bars show the percentage of image pairs in which at least one of three raters (three board-certified pathologists) identified each morphologic feature as present in the original (dark gray) or counterfactual (light gray) image. Right, Mean pairwise inter-rater agreement (Cohen κ) per feature, computed across all pairs and both directions combined. C, Counterfactual image generation effectiveness, measured as the percentage of generated images predicted as the opposite class across varying manipulation amplitudes. D, Representative examples of ILC tile transitioning to its counterfactual IDC image and vice versa. Difference maps display pixelwise differences between the original and the synthetic tile. Scale bar applies to all images within the panel unless otherwise indicated.
- Open Access
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Figure 2 from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
Results for counterfactual image generation for the CRC-VAL-HE-7K dataset. A, Tissue-type classifier-guided counterfactuals of a dysplastic tile, generated by shifting its feature vector toward the predicted healthy colon mucosa class. Changes reflect what the model requires to flip the prediction, with increasing manipulation amplitude shown above each image. B, The t-SNE projection of the feature extractor’s latent space (training set), colored by class: adipose (ADI), background (BACK), debris (DEB), lymphocytes (LYM), mucus (MUC), smooth muscle (MUS), healthy colon mucosa (NORM), cancer-associated stroma (STR), colorectal adenocarcinoma epithelium/dysplastic tissue (TUM). The features of dysplastic tile in A, initially near the boundary between TUM (aqua) and NORM (gray), shift toward the healthy cluster (pink points in the zoomed-in plot; gray area—arbitrary healthy mucosa zone) as manipulation amplitude increases. C, Morphologic feature prevalence in original and counterfactual image pairs. Horizontal bars show the percentage of image pairs in which at least one of four raters (three board-certified pathologists and one pathology trainee) identified each morphologic feature as present in the original (dark gray) or counterfactual (light gray) image. Right, Mean pairwise inter-rater agreement (Cohen κ) per feature, computed across all pairs and both directions combined. D, Counterfactuals of a healthy mucosa tile shifted toward dysplastic epithelium. Below, Pixel-level difference maps (darker regions indicate greater changes), showing the most change in gland regions, and SSIM values, which quantify the similarity between the original and manipulated tiles. Bottom, Segmented and classified nuclei: pink (epithelial), orange (connective tissue), blue (plasma), deep purple (lymphocytes), aqua (neutrophils), and green (eosinophils). E, Tile-level differences in cell type fractions between (left) real NORM tiles (n = 741) and synthetic (counterfactual images generated from TUM tiles, n = 1,233; full statistical details are provided in Supplementary Table S4). Right, Tile-level differences in cell-type fractions between real TUM tiles (n = 1,233) and synthetic (counterfactual images generated from NORM tiles, n = 741).
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Open AccessFigure 2 from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
Results for counterfactual image generation for the CRC-VAL-HE-7K dataset. A, Tissue-type classifier-guided counterfactuals of a dysplastic tile, generated by shifting its feature vector toward the predicted healthy colon mucosa class. Changes reflect what the model requires to flip the prediction, with increasing manipulation amplitude shown above each image. B, The t-SNE projection of the feature extractor’s latent space (training set), colored by class: adipose (ADI), background (BACK), debris (DEB), lymphocytes (LYM), mucus (MUC), smooth muscle (MUS), healthy colon mucosa (NORM), cancer-associated stroma (STR), colorectal adenocarcinoma epithelium/dysplastic tissue (TUM). The features of dysplastic tile in A, initially near the boundary between TUM (aqua) and NORM (gray), shift toward the healthy cluster (pink points in the zoomed-in plot; gray area—arbitrary healthy mucosa zone) as manipulation amplitude increases. C, Morphologic feature prevalence in original and counterfactual image pairs. Horizontal bars show the percentage of image pairs in which at least one of four raters (three board-certified pathologists and one pathology trainee) identified each morphologic feature as present in the original (dark gray) or counterfactual (light gray) image. Right, Mean pairwise inter-rater agreement (Cohen κ) per feature, computed across all pairs and both directions combined. D, Counterfactuals of a healthy mucosa tile shifted toward dysplastic epithelium. Below, Pixel-level difference maps (darker regions indicate greater changes), showing the most change in gland regions, and SSIM values, which quantify the similarity between the original and manipulated tiles. Bottom, Segmented and classified nuclei: pink (epithelial), orange (connective tissue), blue (plasma), deep purple (lymphocytes), aqua (neutrophils), and green (eosinophils). E, Tile-level differences in cell type fractions between (left) real NORM tiles (n = 741) and synthetic (counterfactual images generated from TUM tiles, n = 1,233; full statistical details are provided in Supplementary Table S4). Right, Tile-level differences in cell-type fractions between real TUM tiles (n = 1,233) and synthetic (counterfactual images generated from NORM tiles, n = 741).
- Open Access
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Figure 3 from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
Interpretation of the molecular biomarker (MSI) classifier in colorectal cancer through counterfactual image analysis. A, Representative examples of MSIH tiles transitioning to their counterfactual non-MSIH image. Pixel-wise difference images compare the original image with the most manipulated one. B, Representative examples of non-MSIH tiles transitioning to their counterfactual MSIH image. Pixel-based difference maps highlight changes in nuclei, stroma, cell density, and tissue architecture. C, Morphologic feature prevalence in original and counterfactual image pairs (N = 52 transitions; 26 tiles from 13 patients for each class). Horizontal bars show the percentage of image pairs in which at least one of three raters (three board-certified pathologists) identified each morphologic feature as present in the original (dark gray) or counterfactual (light gray) image. Right, Mean pairwise inter-rater agreement (Cohen κ) per feature, computed across all pairs and both directions combined. D, Quantitative evaluation of counterfactual image generation effectiveness. Bars show the percentage of successful manipulations (i.e., predicted class equals target class) across different manipulation amplitudes. The scale bar applies to all images within the figure.
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Open AccessFigure 3 from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
Interpretation of the molecular biomarker (MSI) classifier in colorectal cancer through counterfactual image analysis. A, Representative examples of MSIH tiles transitioning to their counterfactual non-MSIH image. Pixel-wise difference images compare the original image with the most manipulated one. B, Representative examples of non-MSIH tiles transitioning to their counterfactual MSIH image. Pixel-based difference maps highlight changes in nuclei, stroma, cell density, and tissue architecture. C, Morphologic feature prevalence in original and counterfactual image pairs (N = 52 transitions; 26 tiles from 13 patients for each class). Horizontal bars show the percentage of image pairs in which at least one of three raters (three board-certified pathologists) identified each morphologic feature as present in the original (dark gray) or counterfactual (light gray) image. Right, Mean pairwise inter-rater agreement (Cohen κ) per feature, computed across all pairs and both directions combined. D, Quantitative evaluation of counterfactual image generation effectiveness. Bars show the percentage of successful manipulations (i.e., predicted class equals target class) across different manipulation amplitudes. The scale bar applies to all images within the figure.
- Open Access
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Figure 5 from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
Style–morphology decomposition for disentangling structural and staining effects on MSI prediction changes. A, Representative examples of counterfactual manipulation between MSIH and non-MSIH classes. For each original image x and its counterfactual xcf, style-hybrid (xstyle) and morphology-hybrid (xmorph) images were generated using Vahadane stain transfer. Each hybrid isolates the effect of either stain or morphology while controlling for the other. The right-hand bars show the Shapley-style decomposition of the logit change (Δf) into stain (φstyle) and morphology (φmorph) contributions, demonstrating that morphologic differences dominate the model’s predictions. Grad-CAM visualizations below provide region-level attribution under MIL. In contrast, MoPaDi produces class-directed “what-if” edits that offer a complementary view of candidate morphologic and style changes associated with prediction shifts. Scale bar applies to all images within the panel. B, Decomposition results across test-set patients, showing median contributions of φstyle, φmorph, and total (Δf) for manipulations toward (↑) and away from (↓) each class. C, Scatter plot of morphology versus style contributions per patient, illustrating consistent dominance of morphologic effects across both manipulation directions and classes.
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Open AccessFigure 5 from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
Style–morphology decomposition for disentangling structural and staining effects on MSI prediction changes. A, Representative examples of counterfactual manipulation between MSIH and non-MSIH classes. For each original image x and its counterfactual xcf, style-hybrid (xstyle) and morphology-hybrid (xmorph) images were generated using Vahadane stain transfer. Each hybrid isolates the effect of either stain or morphology while controlling for the other. The right-hand bars show the Shapley-style decomposition of the logit change (Δf) into stain (φstyle) and morphology (φmorph) contributions, demonstrating that morphologic differences dominate the model’s predictions. Grad-CAM visualizations below provide region-level attribution under MIL. In contrast, MoPaDi produces class-directed “what-if” edits that offer a complementary view of candidate morphologic and style changes associated with prediction shifts. Scale bar applies to all images within the panel. B, Decomposition results across test-set patients, showing median contributions of φstyle, φmorph, and total (Δf) for manipulations toward (↑) and away from (↓) each class. C, Scatter plot of morphology versus style contributions per patient, illustrating consistent dominance of morphologic effects across both manipulation directions and classes.
- Open Access
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Figure 6 from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
Counterfactual image examples generated for the liver cancer type classifier [hepatocellular carcinoma (HCC) vs. cholangiocarcinoma (CCA)]. A, Representative examples of counterfactual transitions generated with two approaches (linear and MIL) and a comparison of corresponding segmentation masks. B, Counterfactual image generation effectiveness, measured as the percentage of generated images predicted as the opposite class across varying manipulation amplitudes. HCC-to-CCA required higher amplitudes to achieve similar results, which likely reflects class imbalance in the dataset. Scale bar applies to all images within the panel.
- 2 September 2026
- Published by American Association for Cancer Research (AACR)
Open AccessFigure 6 from Counterfactual Diffusion Models Provide Interpretable Explanations of Artificial Intelligence Models in Pathology
Counterfactual image examples generated for the liver cancer type classifier [hepatocellular carcinoma (HCC) vs. cholangiocarcinoma (CCA)]. A, Representative examples of counterfactual transitions generated with two approaches (linear and MIL) and a comparison of corresponding segmentation masks. B, Counterfactual image generation effectiveness, measured as the percentage of generated images predicted as the opposite class across varying manipulation amplitudes. HCC-to-CCA required higher amplitudes to achieve similar results, which likely reflects class imbalance in the dataset. Scale bar applies to all images within the panel.