Towards Reliable AI-Based Histological Staining: A Systematic Study of Scaling and Uncertainty in Unpaired Generative Models

Qasim Siddiqui; Adrian Friebel; Maiju Myllys; Zaynab Hobloss; Daniela Gonzalez; Ahmed Ghallab; Stefan Hoehme

cs.CVAug 25, 2026

Towards Reliable AI-Based Histological Staining: A Systematic Study of Scaling and Uncertainty in Unpaired Generative Models

Authors: Qasim Siddiqui, Adrian Friebel, Maiju Myllys, Zaynab Hobloss, Daniela Gonzalez, Ahmed Ghallab, Stefan Hoehme

Organizations: Interdisciplinary Centre of Bioinformatics (IZBI) Universität Leipzig Leipzig, Germany · Center for Scalable Data Analytics and Artificial Intelligence (ScaDS.AI) University of Leipzig Leipzig, Germany · Department of Toxicology Technical University Dortmund Dortmund, Germany · Faculty of Veterinary Medicine Qena University Qena, Egypt

Abstract

Liver fibrosis, the principal predictor of long-term outcome in chronic liver disease, is staged from histological estimates of collagen content. Sirius Red (SR) provides the standard quantitative readout (collagen proportionate area, CPA) but is not acquired at every clinical centre and consumes tissue, time, and reagent cost beyond the routine Hematoxylin and eosin (H&E) stain. AI-based virtual staining can generate SR directly from H&E, yet systematic benchmarks of unsupervised models are scarce and their predictive uncertainty has not been quantified, even though visually plausible outputs may not faithfully reproduce the underlying tissue structure. We therefore benchmark six unsupervised image-to-image architectures (GAN-based and diffusion-based) across 54 scaling configurations on a newly released paired H&E to SR mouse liver dataset, the first open resource for this translation task. Each configuration is evaluated jointly on perceptual, distributional, and task-specific axes plus a blinded expert reader study; the best per family is then retrained as a deep ensemble, the first systematic comparison of epistemic uncertainty across unsupervised stain-to-stain architectures. Across families, perceptual quality, task-specific error, and ensemble agreement measure largely independent axes of model fitness: GAN-based methods cluster tightly on perceptual metrics yet differ substantially on task error and ensemble agreement, while the diffusion-based method (CycleDiffusion) is qualitatively different on all three. No single metric captures these differences, so reliable virtual staining requires reporting and selecting on all three jointly. The dataset, tiling pipeline, models, and evaluation code are released publicly.

Figures & tables

Figure 1: Overview of the virtual H&E $\to$ SR staining pipeline studied in this work. A whole-slide image (WSI) acquired with H&E staining is decomposed into $256{\times}256$ pixel tiles, which are passed through an unpaired image-to-image (I2I) translation model to produce virtual Sirius Red tiles. The translated tiles are reassembled into a virtual SR WSI, from which collagen proportionate area (CPA) can be quantified without requiring a second, physical SR-stained section from the same tissue block.

Figure 2: All four evaluation metrics across the 54 scaling configurations ( $6~\text{models}\times 3~\text{sizes}\times 3~\text{data fractions}$ ). Colour encodes training data fraction (blue: 25%, orange: 50%, green: 100%); marker shape encodes generator size (circle: small S, square: medium M, triangle: large L). Large star markers indicate the best-performing model size at the 100% data fraction per model family, selected for ensemble uncertainty estimation in Section 6 . Error bars on Patch-SSIM, LPIPS, and CPA MAE denote $\pm$ std over $N{=}5$ held-out paired specimens (10 WSIs); FID is a set-level statistic with no per-sample variance. Vertical axis directions: Patch-SSIM higher is better; LPIPS, FID, CPA MAE lower is better.

Family	Size	Patch-SSIM $\uparrow$	LPIPS $\downarrow$	FID $\downarrow$	CPA MAE $\downarrow$
CycleGAN	M	$0.334\pm 0.041$	$\mathbf{0.038}\pm 0.001$	$201.8$	$\mathbf{0.008}\pm 0.007$
UNIT	M	$0.333\pm 0.041$	$\mathbf{0.038}\pm 0.002$	$\mathbf{195.2}$	$0.037\pm 0.033$
MUNIT	M	$0.310\pm 0.022$	$0.040\pm 0.001$	$209.5$	$0.043\pm 0.014$
DCLGAN	S	$\mathbf{0.339}\pm 0.040$	$0.039\pm 0.001$	$207.4$	$0.028\pm 0.037$
UVCGAN	S	$0.311\pm 0.032$	$0.039\pm 0.001$	$242.5$	$0.035\pm 0.021$
CycleDiffusion	S	$0.217\pm 0.009$	$0.044\pm 0.000$	$210.4$	$0.051\pm 0.029$

Table 1: Best configuration per model family at 100% training data, selected on lowest CPA MAE (the rule also used for ensemble uncertainty estimation in Section 6 ). Bold marks the best value per metric across families. Patch-SSIM higher is better; LPIPS, FID, and CPA MAE lower is better. Patch-SSIM, LPIPS, and CPA MAE are mean

\pm

std over the

N{=}5

held-out paired specimens; FID is a set-level statistic.

	Patch-SSIM	LPIPS	FID	CPA MAE
Patch-SSIM	1.00
LPIPS	0.82	1.00
FID	0.60	0.49	1.00
CPA MAE	0.63	0.64	0.55	1.00

Table 2: Spearman rank correlation (

r_{s}

) between the four metric orderings over all 54 configurations (rank 1 = best per metric, so each metric is oriented in its own better-is-first direction and no

\uparrow

\downarrow

applies). All off-diagonal values significant at

p{<}0.001

Panel of Figure 3 — Figure 3: Qualitative comparison on two held-out H&E patches. For each patch, the top row shows the H&E input, the loosely registered real Sirius Red ground truth, and the virtual SR produced by CycleGAN’s medium generator at 100% training data and CycleDiffusion’s small generator at 25% training data. The bottom row shows the nnU-Net v2 collagen segmentation masks of the corresponding real and virtual SR images, from which CPA is computed. Because the H&E and SR sections are non-adjacent cuts, visible differences between the real and virtual SR images reflect both true translation error and irreducible inter-section variation. These two configurations bracket the study: the CycleGAN-M 100% configuration records the study-best CPA MAE across all 54 configurations, while the CycleDiffusion-S 25% configuration records the study-worst.

Figure 4: Distribution of per-tile epistemic uncertainty $\bar{\sigma}$ across $9{,}365$ test tiles (those with at least $0.1\%$ tissue mask coverage) for each of the six model families; median values are annotated. GAN-based families cluster within a narrow band (median $37.68$ – $41.42$ , IQR $1.48$ – $3.80$ ); CycleDiffusion sits below the cluster (median $23.35$ ) yet shows substantially wider spread (IQR $5.89$ ).

Split	Paired specimens	H&E tiles	SR tiles	Data fraction
Train (100%)	30	201,257	196,415	100%
Train (50%)	15	109,049	96,292	50%
Train (25%)	7	49,312	49,293	25%
Test	5	16,290	16,290	—

Table S1: Tile counts after tissue filtering. Each paired specimen contributes two WSIs, one H&E and one SR. Training subsets (25%, 50%) are strict subsets of the full 100% training set.

Split	Specimens	WSIs	BDL stage (specimens)
			Sham	Day 3	Day 21	Day 42	Day 63
Train	30	60	13	4	4	4	5
Test	5	10	2	1	1	1	0

Table S2: Dataset split summary. BDL stage indicates the post-ligation day defining the disease stage; sham = control mice without bile-duct ligation. Each animal contributes one paired specimen, and each paired specimen contributes two WSIs, one H&E and one SR, so the stage columns sum to the specimen count rather than the WSI count. The test row carries no day-63 specimen: the only candidate could not be co-registered across its H&E and SR sections and was excluded.

Model	Size	Params	$n_{gf}$	$n_{\text{blocks}}$	Additional hyperparameters
CycleGAN	Small	10.20 M	64	8	—
	Medium	50.18 M	128	10	—
	Large	102.26 M	192	9	—
UNIT	Small	10.33 M	64	8	shared blocks = 2
	Medium	50.71 M	128	10	shared blocks = 2
	Large	103.44 M	192	9	shared blocks = 3

Table S3: Generator size configurations. Parameter counts are for the single A

\to

B generator direction. “Shared” denotes shared bottleneck blocks in UNIT.

Model	Loss weights
CycleGAN	$\lambda_{\mathrm{cyc}}{=}10$ , $\lambda_{\mathrm{id}}{=}0.5$
UNIT	$\lambda_{\mathrm{GAN}}{=}1$ , $\lambda_{\mathrm{recon}}{=}10$ , $\lambda_{\mathrm{KL}}{=}0.01$
MUNIT	$\lambda_{\mathrm{img}}{=}10$ , $\lambda_{c}{=}1$ , $\lambda_{s}{=}1$
DCLGAN	$\lambda_{\mathrm{cyc}}{=}10$ , $\lambda_{\mathrm{id}}{=}0$ , $\lambda_{\mathrm{DCL}}{=}1$
UVCGAN	$\lambda_{\mathrm{cyc}}{=}10$ , $\lambda_{\mathrm{id}}{=}0.5$
CycleDiffusion	$\varepsilon$ -prediction objective only (no auxiliary weights)

Table S4: Loss weights used for each model family. Loss terms vary across architectures; each row lists only the weights applicable to that family.

Model	Size	Data %	FID $\downarrow$	LPIPS $\downarrow$	Patch-SSIM $\uparrow$	CPA MAE $\downarrow$
CycleGAN	S	25	$206$	$0.039\pm 0.001$	$0.331\pm 0.041$	$0.045\pm 0.052$
	S	50	$219$	$0.039\pm 0.001$	$0.329\pm 0.039$	$0.063\pm 0.058$
	S	100	$210$	$0.039\pm 0.001$	$0.339\pm 0.040$	$0.049\pm 0.042$
	M	25	$217$	$0.039\pm 0.001$	$0.331\pm 0.046$	$0.053\pm 0.051$
	M	50	$245$	$0.038\pm 0.002$	$0.323\pm 0.039$	$0.057\pm 0.033$
	M	100	$202$	$0.038\pm 0.001$	$0.334\pm 0.041$	$0.008\pm 0.007$

Table S5: Scaling configurations, part 1 of 2: CycleGAN, UNIT, MUNIT. S/M/L = Small/Medium/Large generator. Data fraction is the percentage of training specimens used. FID is rounded to the nearest integer here and reported to one decimal in the main paper. Configurations meeting FID

>300

and Patch-SSIM

<0.20

are flagged as collapsed (asterisk); a high FID alone is not sufficient.

\downarrow

= lower is better;

\uparrow

= higher is better.

Model	Size	Data %	FID $\downarrow$	LPIPS $\downarrow$	Patch-SSIM $\uparrow$	CPA MAE $\downarrow$
DCLGAN	S	25	$212$	$0.039\pm 0.002$	$0.322\pm 0.039$	$0.049\pm 0.044$
	S	50	$203$	$0.038\pm 0.001$	$0.341\pm 0.042$	$0.049\pm 0.030$
	S	100	$207$	$0.039\pm 0.001$	$0.339\pm 0.040$	$0.028\pm 0.037$
	M	25	$196$	$0.039\pm 0.001$	$0.316\pm 0.039$	$0.047\pm 0.057$
	M	50	$207$	$0.038\pm 0.001$	$0.327\pm 0.034$	$0.019\pm 0.017$
	M	100	$208$	$0.039\pm 0.001$	$0.334\pm 0.042$	$0.042\pm 0.058$

Table S6: Scaling configurations, part 2 of 2: DCLGAN, UVCGAN, CycleDiffusion. Notation and conventions as in Table S5 .

Figure S1: Epistemic uncertainty estimation via deep ensembles. An H&E input tile is forwarded through $K{=}10$ independently trained generators; the pixel-wise sample variance across the $K$ SR outputs yields the uncertainty map $\mathbf{U}(x)$ .

Model	Size	Within-tile $r_{s}$ $\uparrow$	Across-tile Pearson $\uparrow$	Across-tile Spearman $\uparrow$	ECE tile $\downarrow$
CycleGAN	M	$0.147\pm 0.109$	$0.507$	$-0.178$	$0.101$
UNIT	M	$0.139\pm 0.088$	$0.497$	$0.169$	$0.182$
MUNIT	M	$0.165\pm 0.173$	$0.618$	$0.186$	$0.058$
DCLGAN	S	$0.157\pm 0.084$	$0.659$	$-0.016$	$0.055$
UVCGAN	S	$0.069\pm 0.130$	$0.515$	$0.205$	$0.169$
CycleDiffusion	S	$0.058\pm 0.095$	$-0.025$	$-0.038$	$0.214$

Table S7: Uncertainty calibration results. Within-tile

r_{s}

: mean

\pm

std of per-tile Spearman rank correlation between pixel-level uncertainty

U

and cycle-reconstruction error

E

over tissue pixels. Across-tile Pearson and Spearman: correlations between tile-mean

\overline{U}

and tile-mean

\overline{E}

. ECE: tile-level expected calibration error after joint

p_{1}

–

p_{99}

normalisation.

\uparrow

= higher is better;

\downarrow

= lower is better. All results on

N{=}9{,}365

tiles from the 5 held-out H&E WSIs.

Figure S2: Tile-level reliability diagrams for all six model families (top row: CycleGAN, UNIT, MUNIT; bottom row: DCLGAN, UVCGAN, CycleDiffusion). Each panel plots mean normalised cycle-reconstruction error per uncertainty bin ( $y$ -axis) against mean normalised uncertainty ( $x$ -axis); the dashed diagonal is the identity line. ECE is annotated in each panel title. Reliability curves lying below the identity indicate that ensembles assign higher uncertainty than the cycle-reconstruction error warrants; the lowest-uncertainty bin is an exception in every family, where error exceeds uncertainty. CycleDiffusion shows a near-flat error line across all bins, indicating uninformative uncertainty.

Figure S3: Distribution of within-tile Spearman $r_{s}$ between pixel-level uncertainty $U$ and cycle-reconstruction error $E$ across the $N{=}9{,}365$ test tiles for each model family (top row: CycleGAN, UNIT, MUNIT; bottom row: DCLGAN, UVCGAN, CycleDiffusion). The dashed vertical line marks the mean $r_{s}$ . GAN-based families peak around $r_{s}\approx 0.15$ – $0.17$ ; UVCGAN and CycleDiffusion are centred near zero, confirming near-absent spatial calibration. The broad spread in all panels reflects high tile-to-tile variability.

Figure S4: Across-tile scatter plots of tile-mean uncertainty $\overline{U}$ ( $x$ -axis) versus tile-mean cycle-reconstruction error $\overline{E}$ ( $y$ -axis) for each model family (top row: CycleGAN, UNIT, MUNIT; bottom row: DCLGAN, UVCGAN, CycleDiffusion). Uncertainty is on the same linear $0$ – $255$ intensity scale as the per-tile $\bar{\sigma}$ of Figure 4 in the main paper. Pearson $\rho$ and Spearman $r_{s}$ are annotated per panel. GAN-based families show moderate Pearson correlations ( $0.50$ – $0.66$ ) driven by high-leverage tiles at large uncertainty values; Spearman correlations are substantially weaker, indicating the relationship is non-monotonic. CycleDiffusion (bottom right) shows no meaningful correlation in either metric.

Measure	Best config.	Worst config.
Task 1: identified as generated (pooled) $\downarrow$	$38\%$	$62\%$
Task 2: plausibility, reader 1 $\uparrow$	$2.7$	$1.1$
Task 2: plausibility, reader 2 $\uparrow$	$3.2$	$1.4$
Task 2: plausibility, reader 3 $\uparrow$	$3.7$	$1.7$
CPA MAE (Tables S5 , S6 ) $\downarrow$	$0.008$	$0.171$

Table S8: Blinded reader study on the study-best (CycleGAN-M, 100% data) and study-worst (CycleDiffusion-S, 25% data) configurations. Task 1 reports the fraction of generated tiles correctly identified as generated, pooled over the three readers; values close to chance indicate that generated tiles are not distinguishable from real SR. Task 2 reports each reader’s mean plausibility rating on the four-point scale.

\downarrow

= lower is better;

\uparrow

= higher is better.

Virtual Staining

Virtual staining uses AI to predict histochemical or immunohistochemical appearances from unstained or differently stained tissue images.

See all 18 papers

How Far Has AI Come in Liver Fibrosis Staging? A Large-Scale Real-World Dataset and Benchmark
May 25, 2026Yuanye Liu, Nannan Shi, Zhejia Zhang +20Medical Image Classification Medical Image Benchmarks
School of Data Science, Fudan University, Shanghai, China · Department of Radiology, Shanghai Public Health Clinical Center, Fudan University, Shanghai, China · Department of Electrical and Computer Engineering, Northwestern University, Evanston, USA +12
UNIStainNet: Foundation-Model-Guided Virtual Staining of H&E to IHC
Mar 13, 2026Jillur Rahman Saurav, Thuong Le Hoai Pham, Pritam Mukherjee +3Virtual Staining Computational Pathology
Department of Computer Science, University of Texas at Arlington, Arlington, Texas, USA · Department of Radiology, St. Jude Children’s Research Hospital, Memphis, Tennessee, USA · Department of Pathology, St. Jude Children’s Research Hospital, Memphis, Tennessee, USA
Semi-Supervised Virtual Staining via Morphology Preservation and Histopathological Realism Constraints
Sep 1, 2026Baoshun Wang, Weiping Lin, Linwu Wang +3Computational Pathology Image-to-Image Translation
Department of Computer Science, School of Informatics, Xiamen University, Xiamen 361005, China · EuroMov Digital Health in Motion, Univ Montpellier, IMT Mines Ales, Ales, France · Service de M´edecine Nucl´eaire, Centre Hospitalier Universitaire de Nˆımes, Universit´e de Montpellier, Nˆımes, France

May 25, 2026J/K move · Enter open · S save

How Far Has AI Come in Liver Fibrosis Staging? A Large-Scale Real-World Dataset and Benchmark

Yuanye Liu, Nannan Shi, Zhejia Zhang, Hanxiao Zhang+19

School of Data Science, Fudan University, Shanghai, China · Department of Radiology, Shanghai Public Health Clinical Center, Fudan University, Shanghai, China · Department of Electrical and Computer Engineering, Northwestern University, Evanston, USA+12

Medical Image Classification Medical Image Benchmarks

Open paper AskPDF arXiv

Despite years of methodological progress, how far AI has come in liver fibrosis staging has never been systematically evaluated under the heterogeneous, multi-center conditions that define clinical practice. To address this gap, we introduce LiFS, a large-scale dataset and benchmark derived from the MICCAI 2025 CARE-Liver challenge, comprising 610 patients across multiple centers and scanners with multi-sequence MRI. To the best of our knowledge, LiFS is the first benchmark providing complete gadoxetic acid-enhanced sequences with histopathology-confirmed annotations from diverse real-world scanners. Through systematic evaluation of 9 independently developed methods selected from 96 registered teams against in-cohort radiologist reference results, our findings address how far current AI has progressed toward clinical-level liver fibrosis staging from three complementary perspectives. First, against radiologists, the best AI methods were broadly comparable to the senior radiologist and significantly exceeded the junior radiologist in selected settings, while median AI performance generally approached junior-radiologist levels. Second, from a data perspective, cross-center heterogeneity, label imbalance, and contrast-enhanced sequence variability emerge as the dominant challenges for AI methods. Third, from a technical perspective, methodological design choices, including spatial registration, input dimensionality, multi-modal fusion strategy, and backbone architecture, appear to modulate cross-center robustness, although no single choice alone closes the gap. Overall, LiFS provides a rigorous real-world benchmark for positioning the current state of AI in liver fibrosis staging and for enabling future research on the key challenges that limit clinically reliable deployment.

Abstract

Virtual Staining

Explore similar work

How Far Has AI Come in Liver Fibrosis Staging? A Large-Scale Real-World Dataset and Benchmark

UNIStainNet: Foundation-Model-Guided Virtual Staining of H&E to IHC

Semi-Supervised Virtual Staining via Morphology Preservation and Histopathological Realism Constraints