Joining forces for pathology diagnostics with AI assistance: The EMPAIA initiative.

Over the past decade, artificial intelligence (AI) methods in pathology have advanced substantially. However, integration into routine clinical practice has been slow due to numerous challenges, including technical and regulatory hurdles in translating research results into clinical diagnostic products and the lack of standardized interfaces. The open and vendor-neutral EMPAIA initiative addresses these challenges. Here, we provide an overview of EMPAIA's achievements and lessons learned. EMPAIA integrates various stakeholders of the pathology AI ecosystem, i.e., pathologists, computer scientists, and industry. In close collaboration, we developed technical interoperability standards, recommendations for AI testing and product development, and explainability methods. We implemented the modular and open-source EMPAIA Platform and successfully integrated 14 AI-based image analysis apps from eight different vendors, demonstrating how different apps can use a single standardized interface. We prioritized requirements and evaluated the use of AI in real clinical settings with 14 different pathology laboratories in Europe and Asia. In addition to technical developments, we created a forum for all stakeholders to share information and experiences on digital pathology and AI. Commercial, clinical, and academic stakeholders can now adopt EMPAIA's common open-source interfaces, providing a unique opportunity for large-scale standardization and streamlining of processes. Further efforts are needed to effectively and broadly establish AI assistance in routine laboratory use. To this end, a sustainable infrastructure, the non-profit association EMPAIA International, has been established to continue standardization and support broad implementation and advocacy for an AI-assisted digital pathology future.

Digitization of Pathology Labs: A Review of Lessons Learned.

Pathology laboratories are increasingly using digital workflows. This has the potential of increasing laboratory efficiency, but the digitization process also involves major challenges. Several reports have been published describing the individual experiences of specific laboratories with the digitization process. However, a comprehensive overview of the lessons learned is still lacking. We provide an overview of the lessons learned for different aspects of the digitization process, including digital case management, digital slide reading, and computer-aided slide reading. We also cover metrics used for monitoring performance and pitfalls and corresponding values observed in practice. The overview is intended to help pathologists, information technology decision makers, and administrators to benefit from the experiences of others and to implement the digitization process in an optimal way to make their own laboratory future-proof.

Periportal steatosis in mice affects distinct parameters of pericentral drug metabolism.

Little is known about the impact of morphological disorders in distinct zones on metabolic zonation. It was described recently that periportal fibrosis did affect the expression of CYP proteins, a set of pericentrally located drug-metabolizing enzymes. Here, we investigated whether periportal steatosis might have a similar effect. Periportal steatosis was induced in C57BL6/J mice by feeding a high-fat diet with low methionine/choline content for either two or four weeks. Steatosis severity was quantified using image analysis. Triglycerides and CYP activity were quantified in photometric or fluorometric assay. The distribution of CYP3A4, CYP1A2, CYP2D6, and CYP2E1 was visualized by immunohistochemistry. Pharmacokinetic parameters of test drugs were determined after injecting a drug cocktail (caffeine, codeine, and midazolam). The dietary model resulted in moderate to severe mixed steatosis confined to periportal and midzonal areas. Periportal steatosis did not affect the zonal distribution of CYP expression but the activity of selected CYPs was associated with steatosis severity. Caffeine elimination was accelerated by microvesicular steatosis, whereas midazolam elimination was delayed in macrovesicular steatosis. In summary, periportal steatosis affected parameters of pericentrally located drug metabolism. This observation calls for further investigations of the highly complex interrelationship between steatosis and drug metabolism and underlying signaling mechanisms.

Recommendations on compiling test datasets for evaluating artificial intelligence solutions in pathology.

Artificial intelligence (AI) solutions that automatically extract information from digital histology images have shown great promise for improving pathological diagnosis. Prior to routine use, it is important to evaluate their predictive performance and obtain regulatory approval. This assessment requires appropriate test datasets. However, compiling such datasets is challenging and specific recommendations are missing. A committee of various stakeholders, including commercial AI developers, pathologists, and researchers, discussed key aspects and conducted extensive literature reviews on test datasets in pathology. Here, we summarize the results and derive general recommendations on compiling test datasets. We address several questions: Which and how many images are needed? How to deal with low-prevalence subsets? How can potential bias be detected? How should datasets be reported? What are the regulatory requirements in different countries? The recommendations are intended to help AI developers demonstrate the utility of their products and to help pathologists and regulatory agencies verify reported performance measures. Further research is needed to formulate criteria for sufficiently representative test datasets so that AI solutions can operate with less user intervention and better support diagnostic workflows in the future.

Automated Detection of Portal Fields and Central Veins in Whole-Slide Images of Liver Tissue.

Many physiological processes and pathological phenomena in the liver tissue are spatially heterogeneous. At a local scale, biomarkers can be quantified along the axis of the blood flow, from portal fields (PFs) to central veins (CVs), i.e., in zonated form. This requires detecting PFs and CVs. However, manually annotating these structures in multiple whole-slide images is a tedious task. We describe and evaluate a fully automated method, based on a convolutional neural network, for simultaneously detecting PFs and CVs in a single stained section. Trained on scans of hematoxylin and eosin-stained liver tissue, the detector performed well with an F1 score of 0.81 compared to annotation by a human expert. It does, however, not generalize well to previously unseen scans of steatotic liver tissue with an F1 score of 0.59. Automated PF and CV detection eliminates the bottleneck of manual annotation for subsequent automated analyses, as illustrated by two proof-of-concept applications: We computed lobulus sizes based on the detected PF and CV positions, where results agreed with published lobulus sizes. Moreover, we demonstrate the feasibility of zonated quantification of biomarkers detected in different stainings based on lobuli and zones obtained from the detected PF and CV positions. A negative control (hematoxylin and eosin) showed the expected homogeneity, a positive control (glutamine synthetase) was quantified to be strictly pericentral, and a plausible zonation for a heterogeneous F4/80 staining was obtained. Automated detection of PFs and CVs is one building block for automatically quantifying physiologically relevant heterogeneity of liver tissue biomarkers. Perspectively, a more robust and automated assessment of zonation from whole-slide images will be valuable for parameterizing spatially resolved models of liver metabolism and to provide diagnostic information.

Artificial Intelligence in Pathology: From Prototype to Product.

Modern image analysis techniques based on artificial intelligence (AI) have great potential to improve the quality and efficiency of diagnostic procedures in pathology and to detect novel biomarkers. Despite thousands of published research papers on applications of AI in pathology, hardly any research implementations have matured into commercial products for routine use. Bringing an AI solution for pathology to market poses significant technological, business, and regulatory challenges. In this paper, we provide a comprehensive overview and advice on how to meet these challenges. We outline how research prototypes can be turned into a product-ready state and integrated into the IT infrastructure of clinical laboratories. We also discuss business models for profitable AI solutions and reimbursement options for computer assistance in pathology. Moreover, we explain how to obtain regulatory approval so that AI solutions can be launched as in vitro diagnostic medical devices. Thus, this paper offers computer scientists, software companies, and pathologists a road map for transforming prototypes of AI solutions into commercial products.

Evaluation of a numerical simulation for cryoablation - comparison with bench data, clinical kidney and lung cases.

PURPOSE: The accuracy of a numerical simulation of cryoablation ice balls was evaluated in gel phantom data as well as clinical kidney and lung cases. MATERIALS AND METHODS: To evaluate the accuracy, 64 experimental single-needle cryoablations and 12 multi-needle cryoablations in gel phantoms were re-simulated with the corresponding freeze-thaw-freeze cycles. The simulated temperatures were compared over time with the measurements of thermocouples. For single needles, temperature values were compared at each thermocouple location. For multiple needles, Euclidean distances between simulated and measured isotherms (10 °C, 0 °C, -20 °C, -40 °C) were computed. Furthermore, surface and volume of simulated 0 °C isotherms were compared to cryoablation-induced ice balls in 14 kidney and 13 lung patients. For this purpose, needle positions and relevant anatomical structures defining material parameters (kidney/lung, tumor) were reconstructed from pre-ablation CT images and fused with postablation CT images (from which ice balls were extracted by manual delineation). RESULTS: The single-needle gel phantom cases showed less than 5 °C prediction error on average. Over all multiple needle experiments in gel, the mean and maximum isotherm distance were less than 2.3 mm and 4.1 mm, respectively. Average Dice coefficients of 0.82/0.63 (kidney/lung) and mean surface distances of 2.59/3.12 mm quantify the prediction performance of the numerical simulation. However, maximum surface distances of 10.57/10.8 mm indicate that locally larger errors have to be expected. CONCLUSION: A very good agreement of the numerical simulations for gel experiments was measured and a satisfactory agreement of the numerical simulations with measured ice balls in patient data was shown.

Ten quick tips for getting the most scientific value out of numerical data.

Most studies in the life sciences and other disciplines involve generating and analyzing numerical data of some type as the foundation for scientific findings. Working with numerical data involves multiple challenges. These include reproducible data acquisition, appropriate data storage, computationally correct data analysis, appropriate reporting and presentation of the results, and suitable data interpretation. Finding and correcting mistakes when analyzing and interpreting data can be frustrating and time-consuming. Presenting or publishing incorrect results is embarrassing but not uncommon. Particular sources of errors are inappropriate use of statistical methods and incorrect interpretation of data by software. To detect mistakes as early as possible, one should frequently check intermediate and final results for plausibility. Clearly documenting how quantities and results were obtained facilitates correcting mistakes. Properly understanding data is indispensable for reaching well-founded conclusions from experimental results. Units are needed to make sense of numbers, and uncertainty should be estimated to know how meaningful results are. Descriptive statistics and significance testing are useful tools for interpreting numerical results if applied correctly. However, blindly trusting in computed numbers can also be misleading, so it is worth thinking about how data should be summarized quantitatively to properly answer the question at hand. Finally, a suitable form of presentation is needed so that the data can properly support the interpretation and findings. By additionally sharing the relevant data, others can access, understand, and ultimately make use of the results. These quick tips are intended to provide guidelines for correctly interpreting, efficiently analyzing, and presenting numerical data in a useful way.

Focused scores enable reliable discrimination of small differences in steatosis.

BACKGROUND: Automated image analysis enables quantitative measurement of steatosis in histological images. However, spatial heterogeneity of steatosis can make quantitative steatosis scores unreliable. To improve the reliability, we have developed novel scores that are "focused" on steatotic tissue areas. METHODS: Focused scores use concepts of tile-based hotspot analysis in order to compute statistics about steatotic tissue areas in an objective way. We evaluated focused scores on three data sets of images of rodent liver sections exhibiting different amounts of dietary-induced steatosis. The same evaluation was conducted with the standard steatosis score computed by most image analysis methods. RESULTS: The standard score reliably discriminated large differences in steatosis (intraclass correlation coefficient ICC = 0.86), but failed to discriminate small (ICC = 0.54) and very small (ICC = 0.14) differences. With an appropriate tile size, mean-based focused scores reliably discriminated large (ICC = 0.92), small (ICC = 0.86) and very small (ICC = 0.83) differences. Focused scores based on high percentiles showed promise in further improving the discrimination of very small differences (ICC = 0.93). CONCLUSIONS: Focused scores enable reliable discrimination of small differences in steatosis in histological images. They are conceptually simple and straightforward to use in research studies.

Data-Driven Discovery of Immune Contexture Biomarkers.

Background: Features characterizing the immune contexture (IC) in the tumor microenvironment can be prognostic and predictive biomarkers. Identifying novel biomarkers can be challenging due to complex interactions between immune and tumor cells and the abundance of possible features. Methods: We describe an approach for the data-driven identification of IC biomarkers. For this purpose, we provide mathematical definitions of different feature classes, based on cell densities, cell-to-cell distances, and spatial heterogeneity thereof. Candidate biomarkers are ranked according to their potential for the predictive stratification of patients. Results: We evaluated the approach on a dataset of colorectal cancer patients with variable amounts of microsatellite instability. The most promising features that can be explored as biomarkers were based on cell-to-cell distances and spatial heterogeneity. Both the tumor and non-tumor compartments yielded features that were potentially predictive for therapy response and point in direction of further exploration. Conclusion: The data-driven approach simplifies the identification of promising IC biomarker candidates. Researchers can take guidance from the described approach to accelerate their biomarker research.

Physiologically-based modelling in mice suggests an aggravated loss of clearance capacity after toxic liver damage.

Diseases and toxins may lead to death of active liver tissue, resulting in a loss of total clearance capacity at the whole-body level. However, it remains difficult to study, whether the loss of metabolizing tissue is sufficient to explain loss of metabolic capacity of the liver or whether the surviving tissue undergoes an adaptive response to compensate the loss. To understand the cellular impact of toxic liver damage in an in vivo situation, we here used physiologically-based pharmacokinetic modelling to investigate pharmacokinetics of a specifically designed drug cocktail at three different sampling sites of the body in healthy mice and mice treated with carbon tetrachloride (CCl4). Liver zonation was explicitly quantified in the models through immunostaining of cytochrome P450s enzymes. Comparative analyses between the simulated decrease in clearance capacity and the experimentally measured loss in tissue volume indicated that CCl4-induced impairment of metabolic functions goes beyond the mere loss of metabolically active tissue. The here established integrative modelling strategy hence provides mechanistic insights into functional consequences of toxic liver damage in an in vivo situation, which would not have been accessible by conventional methods.

Visualization of Vascular and Parenchymal Regeneration after 70% Partial Hepatectomy in Normal Mice.

A modified silicone injection procedure was used for visualization of the hepatic vascular tree. This procedure consisted of in-vivo injection of the silicone compound, via a 26 G catheter, into the portal or hepatic vein. After silicone injection, organs were explanted and prepared for ex-vivo micro-CT (µCT) scanning. The silicone injection procedure is technically challenging. Achieving a successful outcome requires extensive microsurgical experience from the surgeon. One of the challenges of this procedure involves determining the adequate perfusion rate for the silicone compound. The perfusion rate for the silicone compound needs to be defined based on the hemodynamic of the vascular system of interest. Inappropriate perfusion rate can lead to an incomplete perfusion, artificial dilation and rupturing of vascular trees. The 3D reconstruction of the vascular system was based on CT scans and was achieved using preclinical software such as HepaVision. The quality of the reconstructed vascular tree was directly related to the quality of silicone perfusion. Subsequently computed vascular parameters indicative of vascular growth, such as total vascular volume, were calculated based on the vascular reconstructions. Contrasting the vascular tree with silicone allowed for subsequent histological work-up of the specimen after µCT scanning. The specimen can be subjected to serial sectioning, histological analysis and whole slide scanning, and thereafter to 3D reconstruction of the vascular trees based on histological images. This is the prerequisite for the detection of molecular events and their distribution with respect to the vascular tree. This modified silicone injection procedure can also be used to visualize and reconstruct the vascular systems of other organs. This technique has the potential to be extensively applied to studies concerning vascular anatomy and growth in various animal and disease models.

Quantification of Hepatic Vascular and Parenchymal Regeneration in Mice.

BACKGROUND: Liver regeneration consists of cellular proliferation leading to parenchymal and vascular growth. This study complements previous studies on cellular proliferation and weight recovery by (1) quantitatively describing parenchymal and vascular regeneration, and (2) determining their relationship. Both together are needed to (3) characterize the underlying growth pattern. METHODS: Specimens were created by injecting a polymerizing contrast agent in either portal or hepatic vein in normal or regenerating livers after 70% partial hepatectomy. 3D image data were obtained through micro-CT scanning. Parenchymal growth was assessed by determining weight and volume of the regenerating liver. Vascular growth was described by manually determined circumscribed parameters (maximal vessel length and radius of right inferior portal/hepatic vein), automatically determined cumulative parameters (total edge length and total vascular volume), and parameters describing vascular density (total edge length/volume, vascular volume fraction). The growth pattern was explored by comparing the relative increase of these parameters to the increase expected in case of isotropic expansion. RESULTS: Liver volume recovery paralleled weight recovery and reached 90% of the original liver volume within 7 days. Comparing radius-related vascular parameters immediately after surgical resection and after virtual resection in-silico revealed a slight increase, possibly reflecting the effect of resection-induced portal hyperperfusion. Comparing length-related parameters between post-operative day 7 and after virtual resection showed similar vascular growth in both vascular systems investigated. In contrast, radius-related parameters increased slightly more in the portal vein. Despite the seemingly homogeneous 3D growth, the observed vascular parameters were not compatible with the hypothesis of isotropic expansion of liver parenchyma and vascular structures. CONCLUSION: We present an approach for the quantitative analysis of the vascular systems of regenerating mouse livers. We applied this technique for assessing the hepatic growth pattern. Prospectively, this approach can be used to investigate hepatic vascular regeneration under different conditions.

Zonated quantification of steatosis in an entire mouse liver.

Many physiological processes and pathological conditions in livers are spatially heterogeneous, forming patterns at the lobular length scale or varying across the organ. Steatosis, a common liver disease characterized by lipids accumulating in hepatocytes, exhibits heterogeneity at both these spatial scales. The main goal of the present study was to provide a method for zonated quantification of the steatosis patterns found in an entire mouse liver. As an example application, the results were employed in a pharmacokinetics simulation. For the analysis, an automatic detection of the lipid vacuoles was used in multiple slides of histological serial sections covering an entire mouse liver. Lobuli were determined semi-automatically and zones were defined within the lobuli. Subsequently, the lipid content of each zone was computed. The steatosis patterns were found to be predominantly periportal, with a notable organ-scale heterogeneity. The analysis provides a quantitative description of the extent of steatosis in unprecedented detail. The resulting steatosis patterns were successfully used as a perturbation to the liver as part of an exemplary whole-body pharmacokinetics simulation for the antitussive drug dextromethorphan. The zonated quantification is also applicable to other pathological conditions that can be detected in histological images. Besides being a descriptive research tool, this quantification could perspectively complement diagnosis based on visual assessment of histological images.

Intrahepatic Vascular Anatomy in Rats and Mice--Variations and Surgical Implications.

INTRODUCTION: The intra-hepatic vascular anatomy in rodents, its variations and corresponding supplying and draining territories in respect to the lobar structure of the liver have not been described. We performed a detailed anatomical imaging study in rats and mice to allow for further refinement of experimental surgical approaches. METHODS: LEWIS-Rats and C57Bl/6N-Mice were subjected to ex-vivo imaging using µCT. The image data were used for semi-automated segmentation to extract the hepatic vascular tree as prerequisite for 3D visualization. The underlying vascular anatomy was reconstructed, analysed and used for determining hepatic vascular territories. RESULTS: The four major liver lobes have their own lobar portal supply and hepatic drainage territories. In contrast, the paracaval liver is supplied by various small branches from right and caudate portal veins and drains directly into the vena cava. Variations in hepatic vascular anatomy were observed in terms of branching pattern and distance of branches to each other. The portal vein anatomy is more variable than the hepatic vein anatomy. Surgically relevant variations were primarily observed in portal venous supply. CONCLUSIONS: For the first time the key variations of intrahepatic vascular anatomy in mice and rats and their surgical implications were described. We showed that lobar borders of the liver do not always match vascular territorial borders. These findings are of importance for the design of new surgical procedures and for understanding eventual complications following hepatic surgery.

Representative Sinusoids for Hepatic Four-Scale Pharmacokinetics Simulations.

The mammalian liver plays a key role for metabolism and detoxification of xenobiotics in the body. The corresponding biochemical processes are typically subject to spatial variations at different length scales. Zonal enzyme expression along sinusoids leads to zonated metabolization already in the healthy state. Pathological states of the liver may involve liver cells affected in a zonated manner or heterogeneously across the whole organ. This spatial heterogeneity, however, cannot be described by most computational models which usually consider the liver as a homogeneous, well-stirred organ. The goal of this article is to present a methodology to extend whole-body pharmacokinetics models by a detailed liver model, combining different modeling approaches from the literature. This approach results in an integrated four-scale model, from single cells via sinusoids and the organ to the whole organism, capable of mechanistically representing metabolization inhomogeneity in livers at different spatial scales. Moreover, the model shows circulatory mixing effects due to a delayed recirculation through the surrounding organism. To show that this approach is generally applicable for different physiological processes, we show three applications as proofs of concept, covering a range of species, compounds, and diseased states: clearance of midazolam in steatotic human livers, clearance of caffeine in mouse livers regenerating from necrosis, and a parameter study on the impact of different cell entities on insulin uptake in mouse livers. The examples illustrate how variations only discernible at the local scale influence substance distribution in the plasma at the whole-body level. In particular, our results show that simultaneously considering variations at all relevant spatial scales may be necessary to understand their impact on observations at the organism scale.

Algorithmically generated rodent hepatic vascular trees in arbitrary detail.

Physiologically realistic geometric models of the vasculature in the liver are indispensable for modelling hepatic blood flow, the main connection between the liver and the organism. Current in vivo imaging techniques do not provide sufficiently detailed vascular trees for many simulation applications, so it is necessary to use algorithmic refinement methods. The method of Constrained Constructive Optimization (CCO) (Schreiner et al., 2006) is well suited for this purpose. Its results after calibration have been previously compared to experimentally acquired human vascular trees (Schwen and Preusser, 2012). The goal of this paper is to extend this calibration to the case of rodents (mice and rats), the most commonly used animal models in liver research. Based on in vivo and ex vivo micro-CT scans of rodent livers and their vasculature, we performed an analysis of various geometric features of the vascular trees. Starting from pruned versions of the original vascular trees, we applied the CCO procedure and compared these algorithmic results to the original vascular trees using a suitable similarity measure. The calibration of the postprocessing improved the algorithmic results compared to those obtained using standard CCO. In terms of angular features, the average similarity increased from 0.27 to 0.61, improving the total similarity from 0.28 to 0.40. Finally, we applied the calibrated algorithm to refine measured vascular trees to the (higher) level of detail desired for specific applications. Having successfully adapted the CCO algorithm to the rodent model organism, the resulting individual-specific refined hepatic vascular trees can now be used for advanced modeling involving, e.g., detailed blood flow simulations.