Introduction
Giardia lamblia is the most common intestinal protozoan parasite worldwide, though rarely encountered on duodenal biopsy in developed countries. It is typically an incidental finding on biopsies taken to exclude coeliac disease. The histological changes and clinical features can mimic coeliac disease, and the organisms are easily overlooked on Haematoxylin & Eosin (H&E) sections. Untreated giardiasis can cause chronic malabsorption and nutritional deficiencies. We developed a deep learning model to automate Giardia detection in duodenal biopsies.

Purpose
We set out to develop a deep learning-based automated approach to the detection of Giardia lamblia in duodenal biopsies.

Methods
We trained a U-Net segmentation model to identify Giardia in two morphological classes: "sickle-shaped" and "pear-shaped" (reflecting different sectioning orientations of the organism). The training dataset comprised 4 whole slide images (WSIs) from a single centre containing 2,678 annotated parasites from H&E-stained duodenal biopsies. For inference, WSIs were divided into 2048×2048 pixel patches at 40× magnification. Classification as Giardia-positive versus negative was determined by averaging Giardia pixels across the 10 highest-burden patches per WSI. External validation was performed on 75 biopsies (10 Giardia-positive, 65 normal) from three independent centres.

Results
The model achieved 100% sensitivity and 94% specificity (AUC 0.98) on the external validation set, demonstrating robust generalisation across three independent centres despite training on data from a single source.

Conclusions
We present the first machine learning model for automated Giardia detection in histopathology. Given its rarity, standard classification approaches are impractical; our segmentation-based method provides a high-sensitivity screening tool with robust multi-centre generalisation.

Background

T-cell lymphoma (TCL) arises from a clonal expansion of T lymphocytes. TCL is often difficult to distinguish histologically from benign T-cell infiltrates. Patients often require multiple biopsies, meaning diagnosis can be delayed by months or even years. Current PCR‑based T‑cell receptor (TCR) clonality studies have long turnaround times and cannot assess tissue architecture, morphology, or immunophenotype.

Purpose

We previously validated a pair of highly specific antibodies against mutually exclusively expressed T-cell receptor beta constant regions, TCRbeta1 and TCRbeta2. These are amenable to immunohistochemical staining of formalin-fixed, paraffin-embedded (FFPE) tissue sections, providing a novel strategy to determine T-cell monotypia as a surrogate for T-cell clonality, like kappa and lambda for B cells. Here we develop automated cell counting for accurate quantification of the TCRbeta2:TCRbeta1 ratio.

Methods

For 8 benign and 13 TCL-containing samples, three non-overlapping fields of TCRbeta1 or TCRbeta2-immunostained serial sections were photographed and used for quantitative analysis. Automated cell counting was performed in QuPath, an image analysis software, using a pre-trained StarDist model, under human supervision.

Results

Automated quantification using QuPath and StarDist produced accurate cell counts, permitting TCRbeta2:TCRbeta1 ratio calculation. The TCRbeta2:TCRbeta1 cell ratio in benign tonsils and lymph nodes was between 0.49:1 and 1.25:1, consistent with previously published quantitative real-time PCR (Q-PCR) and in situ hybridisation (ISH) data. Most TCL cases had obvious TCRbeta1/2 restriction, rendering counting superfluous. However, automated counting of positively immunostained cells was very helpful in determining the likely clonal status of lymphomas with a significant tumour-infiltrating benign T-cell population.

Conclusions

We further validate our robust, inexpensive, and highly specific TCRbeta1/2 immunohistochemistry for assessing T cell monotypia in FFPE tissue, as a surrogate for T-cell clonality. We show that this method is amenable to automated cell counting permitting accurate calculation of the TCRbeta2:TCRbeta1 ratio, and broadening the potential clinical applications of this test.

Background: Prostate cancer disproportionately affects Sub-Saharan African men, who often present with aggressive, late-stage disease.
Purpose: This systematic review synthesizes contemporary evidence on the immunophenotypes and molecular patterns of prostate cancer in East Africa to inform regional screening and precision oncology strategies.
Methods: Following Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 guidelines, a systematic search was conducted across PubMed, Scopus, and Web of Science for studies published between 2011 and 2025. Eligible studies focused on histologically confirmed cases in East Africa reporting on immunohistochemical markers or genomic alterations. Quality was appraised using Joanna Briggs Institute Critical Appraisal Tools.
Results: Seven studies (five cross-sectional, two case-control) involving over 1,850 participants from Uganda, Sudan, Kenya, and Rwanda were included. Findings reveal a “late-stage presentation paradox,” with mean prostate-specific antigen levels reaching 434.06 ng/mL and high-grade malignancies (Gleason Score ≥ 8) in up to 82.4% of cases. Key biomarkers identified include near-universal Cyclin D1 expression (98.3%), ERG positivity (75.4%), and pathogenic variants at the 8q24 locus (rs72725854). Loss of BRCA1/2 was also noted in high-grade tumors. All included studies demonstrated a low risk of bias, though confounding management remains a recurring methodological weakness.
Conclusions: East African prostate cancer is defined by an aggressive molecular profile and systemic diagnostic delays. The identification of population-specific genetic drivers necessitates a shift from Western-derived risk models toward “decolonized” precision oncology and decentralized diagnostic infrastructure to improve early detection and patient outcomes.

Aim:
We describe four cases of syringotropic melanoma and review their clinicopathologic features in the context of the literature.

Methods:
We retrospectively identified four melanomas from the St Vincent’s University Hospital database fulfilling criteria for syringotropic involvement. Clinical information was provided by the requisition forms and authorized pathology reports. A literature review was conducted in English using the Pubmed and Google Scholar database with the search term “syringotropic melanoma.” Immunohistochemistry was used to confirm syringotropic involvement.

Results:
Three patients were female, aged 81, 83 and 27, with lesions on the face, sole of foot and arm respectively. One patient was male, 75 with a lesion on his scalp. Melanoma subtypes included superficial spreading, lentigo maligna and acral lentiginous melanoma. All cases showed dermal invasion (Clark IV), with Breslow thickness 2mm, 2.5mm, 1.4mm and 2mm respectively. Syringotropic involvement was confirmed using immunohistochemistry for SOX10 and Mel A, recognising eccrine cells positive for SOX10 and negative for Mel A. Conventional adverse prognostic markers include lymphovascular invasion for one case, and absent tumour infiltrating lymphocytes (TILs) for three. In each case, syringotropism resulted in an increased Breslow thickness.

In our literature review, ten papers were found with a total of 48 cases of syringotropism. The breakdown of cases is as follows; 34 had syringotropism without invasion, 13 cases had syringocentric invasion with all but one resulting in an increase in Breslow thickness, and 1 case showed syringocentric invasion which may have contributed to Breslow thickness.

Conclusion:
Syringotropic melanoma is rare and may present across multiple anatomical sites. Recognition of syringotropism is important to avoid underestimating tumour invasion and, in its presence, consideration should be given to extra levels and immunohistochemistry.

**Of note, summary of the core data items for each cases would pre provided

Background
Intracellular signal initiation from receptor tyrosine kinases (RTKs), based on recruitment of downstream effector proteins to proline-rich motifs (PRMs) on the C-termini of the receptors has recently been described and termed ‘Tier 2 signalling’, in contrast to the canonical (Tier 1) signalling which requires up-regulation of RTK kinase activity. Independent of the ‘on/off’ switch provided by receptor post-translational phosphorylation/dephosphorylation, Tier 2 signalling is entirely dependent on the competitive binding of cytosolic proteins with PRM-recognising domains (e.g. SH3 domains), dictated by their relative concentrations in different intracellular conditions, to the C-terminal PRMs of receptors. Tier 2 signalling has been shown to drive oncogenic outcomes. The current focus on Tier 2 signalling associated with RTKs in epithelial cells has not previously been explored in other receptor systems or cell types, e.g. immune cells. Notably, T cells contain membrane-bound receptors with C-terminal PRMs and cytoplasmic PRM-binding proteins.

Purpose
To identify binding partners of two selected PRMs from the C-terminal of receptors on T cells: CD3ε and CD2 to explore the potential presence of Tier 2 signalling in T cells.


Methods
We used an unbiased affinity purification mass spectrometry approach to identify binding partners of these two selected PRMs and validated a selected ‘hit’ using super-resolution microscopy.

Results
We discovered 7 binding partners for the CD3ε PRM (5 putatively novel) and 5 putatively novel binding partners for the CD2 PRM (False Detection Rate=0.05, p-value
Conclusions
Our experimental approach provides for future validation of other novel putative binding partners found of both PRMs. Functional characterisation of any validated hits, explored in different intracellular conditions, would allow further delineation of Tier 2 signalling pathways in T cells and identification of potential therapeutic targets.