The study, published in Acta Parasitologica by a large team of researchers led by Fuzhong Fang and Zhiyi Lin of the First Affiliated Hospital of Shihezi University in Xinjiang, China, set out to answer a deceptively simple question: does hepatic alveolar echinococcosis, commonly abbreviated HAE, truly exhibit the invasive growth behavior of hepatocellular carcinoma, the most common form of primary liver cancer? To find out, the investigators collected lesional tissue samples from patients diagnosed with each condition, prepared histological slides, and systematically examined the specimens under light microscopy, comparing the two diseases across every clinically meaningful interface between lesion and host liver.

The first major finding concerns the blood vessels within the lesions, and it cuts to the heart of why HAE has been compared to cancer in the first place. Malignant tumors are master angiogenic machines: they recruit and construct dense clusters of brand-new microvessels to feed their expanding mass, a process that also provides a highway for metastasis. When the researchers examined the vascular structures embedded in HAE lesions, however, they found something fundamentally different. The vessels inside the parasitic lesions were predominantly residual vessels, pre-existing hepatic vessels that had been encased and progressively compressed by the advancing lesion rather than newly forged by an angiogenic program. In hepatocellular carcinoma samples, by contrast, clustered neovascularization predominated, confirming the aggressive, self-supplying character of the cancer.

The same logic applied to the bile ducts. Within HAE lesions, the bile ducts the team observed represented residual normal bile duct structures that had been swallowed up and enveloped by the growing parasite mass, not new ducts generated by the lesion. In hepatocellular carcinoma, only a small percentage of cases showed subtle proliferation of small bile ducts within the tumor stroma. Intriguingly, the picture was more symmetrical when it came to nerves. Perineural encasement, in which the lesion wraps around nerve trunks without breaching them, was observed in 12.6 percent of HAE cases versus 8.4 percent of HCC cases, a difference the authors report as statistically non-significant with a P value of 0.319. True neural invasion, where the process actually penetrates the nerve, occurred in 3.9 percent of HAE cases and 1.9 percent of HCC cases, also not significantly different at P equals 0.644. In this narrow sense, the parasite and the cancer do share a behavioral quirk.

The most decisive evidence emerged at the interface between lesional and hepatic tissue, the true battleground where malignant tumors declare their identity. In the HAE specimens, the fibrous capsule surrounding the lesion remained intact throughout, with no evidence of infiltration through this biological barrier. The lesion expanded outward like a slowly inflating balloon, pushing structures aside rather than chewing through them. The hepatocellular carcinoma specimens told a completely different story: cancer cell infiltration across the interface was observed in fully 80.4 percent of cases, a difference the authors report as statistically significant at P less than 0.05. HAE likewise showed no direct invasion of the hepatic parenchyma, consistent with the preservation of the liver’s fibrous septa, whereas septal disruption with direct parenchymal invasion along the sinusoidal channels was identified in 17.8 percent of the HCC cases, again a significant difference.

Vascular behavior at the interface sharpened the contrast further. The incidence of vascular compression, in which the growing lesion simply squeezes a vessel, was comparable between the two groups, occurring in 20.4 percent of HAE cases and 16.8 percent of HCC cases with a P value of 0.507. But when the researchers looked for genuine vascular invasion, malignant cells or larval tissue actually breaching the vessel wall and entering the lumen, the two diseases diverged dramatically. Vascular invasion was rare in HAE, detected in just 1.9 percent of cases, while it occurred in 31.8 percent of hepatocellular carcinoma cases, a highly significant difference. The bile ducts followed a similar pattern of benign equivalence: bile duct compression rates were not significantly different between the conditions, at 16.5 percent for HAE and 10.3 percent for HCC, and bile duct invasion was rare in both, at 1.9 percent versus 3.7 percent with a P value of 0.714.

At the outermost boundary, where lesions reached the liver’s surface, the pattern held firm. HAE lesions adjacent to the hepatic capsule were consistently covered by dense fibrous tissue, with no evidence of invasion through the capsule itself. Hepatocellular carcinoma, true to its nature, showed direct capsular invasion in 17.8 percent of cases, another statistically significant difference. Taken together, the authors argue, these findings demonstrate that from a pathological perspective, the growth pattern of hepatic alveolar echinococcosis differs markedly from the invasive growth of hepatocellular carcinoma. The parasite grows expansively, always encased within a fibrous capsule, displacing anatomy rather than destroying it. The cancer grows infiltratively, breaching every natural barrier the liver can erect and seeding its own spread through vessels, septa, and capsules.

The clinical implications of this distinction could be substantial. If HAE lesions reliably maintain an intact fibrous capsule as they expand, then the boundary between parasite and host tissue represents a genuine surgical plane, a clean theoretical line along which a scalpel can safely travel. The study’s authors explicitly state that this expansive, capsule-confined growth pattern provides a theoretical foundation for the complete surgical resection of the lesion, and it may bolster confidence in organ-preserving and capsule-based techniques such as intracapsular enucleation, approaches designed to remove the lesion while sparing as much healthy liver as possible. For a disease in which radical resection can demand removal of the majority of the liver, and in which transplantation is sometimes the only remaining option, the anatomical reassurance that the parasite respects its capsule could help more patients qualify for limited resections and better functional outcomes.

The findings also carry conceptual weight for how the medical community frames a neglected zoonotic disease. Alveolar echinococcosis remains a serious public health problem across large swaths of Central Asia, northwestern China, Siberia, and parts of Europe, with fatality rates that were historically dismal and a global burden estimated in the hundreds of thousands of disability-adjusted life years. Calling HAE a parasitic cancer has always been a useful shorthand for conveying urgency, but the new data suggest the metaphor should be used with precision: the disease behaves like a cancer in its relentless, lethal potential and its capacity for local destruction, yet it does not obey the histological grammar of malignancy. Understanding that difference, the researchers conclude, offers a new theoretical basis for optimizing clinical management strategies, and a reminder that even the most intimidating imitations in pathology can be unmasked by careful examination of the tissue itself.

Subject of Research: Histopathological differences in growth patterns between hepatic alveolar echinococcosis and hepatocellular carcinoma

Article Title: Pathological Study on Differences in Growth Patterns Between Hepatic Alveolar Echinococcosis and Hepatocellular Carcinoma

Article References: Pathological Study on Differences in Growth Patterns Between Hepatic Alveolar Echinococcosis and Hepatocellular Carcinoma. (n.d.). https://doi.org/10.1007/s11686-026-01387-x

Image Credits: AI Generated

DOI: 10.1007/s11686-026-01387-x

Keywords: hepatic alveolar echinococcosis, hepatocellular carcinoma, growth patterns, fibrous capsule, vascular invasion, neovascularization, pathology, parasitic liver disease, Echinococcus multilocularis, expansive growth, infiltrative growth, surgical resection

Meningioma, a tumor arising from the meninges—the protective membranes enveloping the brain and spinal cord—is the most frequently diagnosed intracranial neoplasm in dogs, accounting for roughly half of all canine primary brain tumors. In humans, meningiomas represent approximately 40 percent of primary brain tumors. Despite this shared prominence, diagnostic criteria for canine meningioma have remained poorly codified, and until now no study had subjected a large cohort of canine cases to standardized review by a panel of multiple pathologists working alongside physician neuropathologists.

The CBTC board, which brought together veterinary pathologists and human neuropathologists from institutions across the United States and Europe, undertook a comprehensive evaluation of 190 cases of canine meningioma. Tumors were collected from the diagnostic archives of ten veterinary schools, including Auburn, Colorado State, Cornell, Mississippi State, Ohio State, Texas A&M, UC Davis, Illinois, Penn, and Virginia Tech. Samples came exclusively from treatment-naïve patients, prioritized when collected within the previous five years and accompanied by clinical follow-up data such as tumor location, therapy, progression, and cause of death. Histology slides were cut at four micrometers, stained with hematoxylin and eosin, and digitally scanned at 40X resolution, allowing every reviewer to examine identical digital images.

To enable direct comparison with human medicine, the board applied the 2016 World Health Organization classification and grading guidelines for human meningioma to each canine tumor. Reviewers assessed histologic subtype, mitotic count within a fixed region of 2.37 square millimeters, invasion into surrounding nervous tissue, geographic coagulative necrosis, and the atypical features used in human grading: sheeting architecture, small-cell formation, hypercellularity, and macronucleoli. The process began with a virtual meeting in October 2020 to familiarize the physician neuropathologists with canine histology, followed by additional consensus sessions to resolve difficult diagnostic calls before a final independent assessment of all cases by October 2021.

The results demonstrated that canine meningiomas mirror the morphologic diversity of the human disease. The overwhelming majority of tumors—140 of 190, or roughly 74 percent—were classified as meningothelial, with transitional, microcystic, psammomatous, rhabdoid, fibrous, metaplastic, clear cell, and chordoid subtypes making up the remainder. Most cases, 109 tumors or 57 percent, were assigned grade 2 under the human WHO scheme, while 79 were grade 1 and only two reached grade 3. Roughly 60 percent of tumors with assessable adjacent brain or spinal cord tissue showed invasion into the nervous system, and features such as necrosis, sheeting, and hypercellularity were commonly observed, closely paralleling the histologic picture of atypical human meningioma.

Yet the clinical correlation told a more complicated story. Among the subset of patients with detailed progression and cause-of-death information, and again in a larger cohort restricted to live/dead status, there was no statistically significant difference in outcome between grade 1 and grade 2 canine tumors. Relative risk ratios for death at one, two, and three years after diagnosis—0.75, 0.93, and 1.02 respectively—were all statistically insignificant. This challenges the assumption that the human WHO grading framework, which strongly predicts recurrence and progression in people, can be transplanted wholesale to veterinary patients. Several features that signal higher grade in humans, notably macronucleoli and small-cell formation, appeared frequently even in grade 1 canine tumors, and medical pathologists on the board repeatedly observed that macronucleoli occur more often in canine tumors regardless of grade.

One feature did emerge as potentially meaningful: necrosis. Although many progressing tumors lacked necrosis, every case in the outcome-linked subset that did exhibit necrosis went on to show clinical progression, compared with 57 percent of cases lacking it. The authors suggest that necrosis, when present, may be a significant indicator of more aggressive behavior and a critical feature for future prognostic studies. Brain tumors were also significantly more likely than spinal tumors to show necrosis. Notably, the study echoes a shifting paradigm in human medicine, where tumors showing brain invasion without other atypical features—so-called brain-invasive, otherwise benign tumors—now benefit from additional molecular testing, suggesting that molecular characterization could likewise improve prognostication in dogs.

The review also shed light on the reproducibility of diagnosis across specialists. Interobserver agreement, measured by intraclass correlation coefficients, was generally low for most histologic features, with the highest agreement—0.44 for both—achieved for necrosis and CNS invasion, the same features that command the highest concordance in human meningioma review. Concordance exceeded 70 percent for most histologic subtypes, especially those with distinctive appearances such as clear cell and microcystic tumors, while mitotic counting proved the least reliable. Interestingly, veterinary pathologists agreed with one another significantly more often than physician neuropathologists on histologic subtype, hypercellularity, and macronucleoli, suggesting that species-specific experience is critical for diagnostic consistency.

Clinically, the cohort reflected established patterns of canine disease. The median patient age was ten years, and mixed-breed dogs, Golden Retrievers, Labrador Retrievers, Boxers, and West Highland Terriers were the most commonly represented breeds. Most tumors arose over the brain, particularly the olfactory and frontal lobes, with a smaller proportion in the cervical spinal cord. Surgery remained the mainstay of treatment, with a median survival of 437 days for dogs treated surgically, while dogs receiving surgery combined with radiation or other adjuvant therapies survived longer on average, at 28 to 55 months, consistent with previously reported ranges.

The CBTC board concluded that while canine and human meningioma share important histologic features, a revised grading framework specific to dogs is needed, distinct from the human WHO scheme. The consortium recommends that future studies include review of all slides by at least three pathologists, systematic evaluation of location, subtype, mitotic count, invasion, necrosis, and other atypical features, and enrollment limited to cases with complete clinical follow-up. The authors also call for the kind of large-scale, shared cancer databases in veterinary medicine that exist in human oncology, and note that molecular characterization, transcriptional profiling, and immune landscape studies of related canine tumor collections are already underway. If successful, these efforts could position the canine patient as a genuinely translational model—accelerating biomarker discovery and therapeutic development for both dogs and the humans who love them.

Subject of Research: Comparative pathology of canine and human meningioma and the validation of dogs as a naturally occurring model for human meningioma research

Article Title: A comparative evaluation of canine meningioma supporting the canine patient as a naturally occurring animal model for human meningioma: a report from the NCI Comparative Brain Tumor Consortium (CBTC) meningioma pathology board

Article References: Church, M. E., Rissi, D. R., Koehler, J. W., Miller, A. D., Beck, J. A., Belluco, S., Bitar, M., Chkheidze, R., Corps, K. N., Matiasek, K., Phillips, J. J., Rajan, S., Stemmer-Rachamimov, A., Yip, S., Shih, J. H., Mazcko, C., & LeBlanc, A. (2026). A comparative evaluation of canine meningioma supporting the canine patient as a naturally occurring animal model for human meningioma: a report from the NCI Comparative Brain Tumor Consortium (CBTC) meningioma pathology board. Veterinary Oncology, 3(1), Article 16. https://doi.org/10.1186/s44356-026-00068-1

Image Credits: AI Generated

DOI: 10.1186/s44356-026-00068-1

Keywords: meningioma, canine, comparative oncology, brain tumor, WHO grading, pathology, NCI Comparative Brain Tumor Consortium, veterinary oncology, CNS invasion, tumor necrosis, animal model, One Health

Segmentation, the pixel-by-pixel outlining of objects of interest within an image, is the bedrock of quantitative pathology. Once a tumor region, a gland, a nucleus, or an inflammatory cell cluster has been precisely delineated in a digitized tissue slide, everything else follows: automated cell counting, measurement of invasive fronts, quantification of stromal composition, and the downstream prediction of diagnosis and prognosis. Yet the traditional route to such measurements has demanded a dedicated deep learning model for each structure of interest, each trained on painstakingly annotated data, or else repeated manual prompting in which a user must click points or drag boxes around every single object in every single image. Neither approach scales gracefully to the enormous structural diversity of human histology.

PathSegmentor takes a different route. Instead of relying on spatial prompts, the model accepts a textual description of the target structure, such as a named anatomical region, histological structure, or object type, and produces the corresponding segmentation mask directly. The language interface means that a researcher or clinician does not need to identify and localize every object before the model can work; a single descriptive phrase suffices. In practice, the team reports, these text prompts substantially reduced the need for object-by-object point or box annotations while remaining robust to natural variation in wording, so that slightly different phrasings of the same request still yield reliable segmentations.

The foundation for this flexibility is PathSeg, an aggregated resource assembled from 21 publicly available pathology image segmentation datasets. Together these contribute 275,200 image–mask–label triples, each pairing a pathology image with its expert ground-truth mask and a categorical label. Crucially, the labels are not a flat, idiosyncratic jumble. The team organized them into a three-level hierarchy that moves from anatomical region, through histological structure, down to the specific object type. This structure allows a single naming scheme to span everything from broad tissue classes to individual epithelial cells, smooth muscle cells, leukocytes, red blood cells, lymphocytes, healthy and cancerous prostate glands, and tumor epithelium, across organs including the breast, colon, lung, and prostate.

Trained on this resource, a single PathSegmentor model achieved the highest overall segmentation performance across 16 internal test datasets, encompassing roughly 45,000 evaluated image–mask pairs. The comparisons were rigorous and broad: the authors benchmarked against task-specific state-of-the-art architectures such as nnU-Net and DeepLabV3+, against specialized pathology adaptations like SAM-Path, and against spatial-prompted segmentation foundation models including MedSAM and SAM-Med2D, as well as the multi-modal biomedical model BiomedParse. Statistical comparisons used one-sided Student’s t-tests with Holm-adjusted p-values and bootstrap-estimated 95 percent confidence intervals, and the model’s advantage was consistent across the hierarchy of anatomical regions, histological structures, and object types, with particularly notable gains on intricate, morphologically complex objects where competing approaches struggled.

Internal performance alone proves little in medical imaging, where domain shift, differences in scanners, staining protocols, patient populations, and disease prevalence, can silently degrade a model that looked flawless in development. The researchers therefore validated PathSegmentor on external public datasets and, importantly, on clinical cohorts contributed by collaborating hospitals, including two clinical datasets whose redistribution is restricted by institutional rules. The model generalized across these external settings and across object categories, suggesting that its text-driven, hierarchy-aware design captures representations of histological structure that transfer beyond the specific datasets it was trained on.

Perhaps the most clinically consequential application demonstrated in the study is interpretability. Deep learning classifiers for breast cancer, including models trained on large whole-slide image collections, are famously opaque: they produce a diagnosis or a biomarker call without a transparent account of which image features drove the decision. The team exploited PathSegmentor’s predicted structures as a lens onto these classification models. By systematically perturbing individual segmented objects and by mapping class activation within the boundaries of each predicted structure, they could attribute a breast cancer classifier’s behavior to specific, nameable histological entities, distinguishing the contributions of tumor regions, healthy glands, and other recognized structures in a way that raw saliency maps cannot.

This object-level explanation capability connects to a broader movement in medical artificial intelligence toward models whose reasoning can be audited by the clinicians who must act on their outputs. Instead of explaining a black box with another approximation of a black box, the approach grounds explanations in concrete anatomical and histological units that pathologists already understand. For cancer diagnosis in particular, where lobular versus ductal morphology, lymphocytic infiltration, and gland architecture all carry distinct clinical meaning, the ability to ask not just what the model predicted but which structures it relied upon could accelerate trust, regulatory review, and clinical adoption of computational pathology tools.

The authors note that their results establish a unified framework for flexible pathology segmentation with potential utility for clinically interpretable image analysis, and the practical infrastructure reflects that ambition. The PathSegmentor source code and analysis scripts are publicly released under an MIT License, with the exact release archived on Zenodo, and PathSeg’s dataset sources and harmonization procedures are documented in the paper’s supplementary materials. Remaining challenges are candidly acknowledged in the authors’ own failure analysis: the model can still struggle with complex or ambiguous boundaries and with densely clustered cells, and the clinical datasets underpinning external validation are available only by request through the contributing hospitals. Even so, the combination of a quarter-million annotated triples, a single model spanning organs and scales, and a natural-language interface represents a significant step toward pathology image analysis in which the question you ask, in words, determines what the machine sees.

The lineage behind PathSegmentor traces back to the Segment Anything Model, the general-purpose vision system introduced in 2023 that demonstrated how a single segmentation model, trained on more than a billion masks, could generalize to unfamiliar images when guided by interactive prompts. That breakthrough quickly inspired medical adaptations, most prominently MedSAM, which showed that the same prompt-driven paradigm could be retuned for radiology, microscopy, and other clinical modalities. What these systems lacked, however, was a way to specify targets semantically. A point or box tells a model where to look, but not what it is looking at, which is precisely the gap that language-prompted approaches such as BiomedParse and PathSegmentor now aim to close in biomedical imaging.

The three-level label hierarchy in PathSeg also reflects a practical truth about pathology annotation: structures in tissue slides are inherently nested. A breast tissue section belongs to an anatomical region, contains histological structures such as glands or stroma, and those structures in turn harbor object types like lymphocytes or tumor epithelium. Encoding this nesting into the training labels gives the model a form of structured supervision that flat label spaces cannot provide, and it mirrors how pathologists themselves are trained to reason about tissue, from organ context down to cellular detail.

The interpretability experiments build on a well-established tension in medical machine learning. Saliency-based explanation techniques, including class activation mapping and randomized input sampling methods, highlight pixels that influence a classifier but produce heatmaps that resist clinical interpretation. Object-level perturbation offers an alternative: by removing or altering a named structure, such as a segmented gland or an infiltrating lymphocyte population, and observing how the classifier’s output changes, the analysis yields attributions expressed in vocabulary that pathologists use daily. This aligns with arguments in the field that high-stakes decisions demand explanations grounded in human-meaningful concepts rather than pixel-level artifacts.

The connection to breast cancer classification is particularly apt given the field’s history. Large molecular studies of breast tumors, including the landmark Cancer Genome Atlas characterization, established that morphological features visible in routine histology correlate with molecular phenotypes, and subsequent work has shown that image-derived features can predict diverse molecular properties of tumors. A segmentation model that reliably delineates the relevant structures provides the missing bridge between such slide-level predictions and the concrete histological evidence behind them.

It is also worth situating the computational approach within the broader methodology of computational pathology. Because whole-slide images are gigapixel-scale and diagnosis often depends on subtle findings scattered across a slide, many classification pipelines rely on multiple-instance learning, in which a slide is treated as a bag of patches with only slide-level labels available. Object-level explanations from a segmentation foundation model could complement these pipelines by revealing which instances within the bag carry diagnostic weight, addressing a long-standing limitation of weakly supervised approaches.

The release of code, an archived version, and documented dataset harmonization procedures follows growing expectations for reproducibility in machine learning for healthcare, allowing independent groups to verify the reported benchmark comparisons and to extend the framework to new organs, stains, and scanning systems as annotated data continue to accumulate.

Subject of Research: Segment anything in pathology images with natural language

Article Title: Segment anything in pathology images with natural language

Article References: Chen, Z., Hou, J., Lin, L., Wang, Y., Bie, Y., Wang, X., Zhou, Y., Li, D., Tan, H., Liang, L., Chan, R. C. K., & Chen, H. (2026). Segment anything in pathology images with natural language. Nature Computational Science. https://doi.org/10.1038/s43588-026-01042-5

Image Credits: AI Generated

DOI: 10.1038/s43588-026-01042-5

Keywords: Segment, anything, pathology, images, natural, language, scientific research