Background:
Central nervous system (CNS) lesions, though relatively uncommon, pose significant diagnostic and therapeutic challenges due to their varied pathology and critical location. Intraoperative crush smear cytology offers a rapid, low-cost alternative to frozen sections, especially in resource-limited environments where cryostat use is restricted by high brain tissue water content.
Purpose:
To assess the diagnostic validity of intraoperative crush smear cytology and preoperative radiological findings by comparing them with final histopathological diagnoses, and to identify associated diagnostic challenges and limitations.
Methods:
A descriptive, cross-sectional study was carried out at the National Hospital of Sri Lanka. Data from 548 cases of CNS lesions were analysed. Intraoperative crush smear and radiological diagnoses were compared against definitive histological outcomes.
Results:
Crush smear cytology showed high diagnostic performance for primary CNS tumours: glial tumours (sensitivity 86.5%, specificity 98.3%), meningiomas (95.6%, 90.8%), schwannomas (92.9%, 99.6%), and pituitary lesions (98%, 100%). Metastatic lesions showed lower sensitivity (67.4%). Radiological diagnosis had high specificity (>92%) and sensitivity varied by tumour type, with pituitary and meningioma lesions showing the best concordance.
Conclusion:
Diagnostic accuracy depended on smear quality, cellularity, and background elements such as necrosis, calcification, and specific cell patterns (e.g., rosettes, whorls). Glial tumours were the most common lesions. Crush smear cytology proved useful in intraoperative differentiation of neoplastic and non-neoplastic conditions, guiding surgical decisions and avoiding repeat interventions.
Crush smear cytology is a reliable, efficient intraoperative diagnostic tool, particularly where cryostat facilities are unavailable. Its value is expected to rise with increasing stereotactic biopsy use in CNS lesion management.

Background
Cervical cancer remains a major public health problem in low-resource settings, where access to screening services is limited. Conventional Papanicolaou smear cytology is widely used but is associated with a high rate of unsatisfactory samples due to obscuring blood, mucus, inflammation, and uneven cell distribution. Liquid-based cytology (LBC) improves specimen adequacy and background clarity; however, its use is limited by the cost of proprietary fixatives and automated processing systems (SurePath and ThinPrep).
Purpose
Evaluate specimen adequacy, cytomorphological quality, and diagnostic performance of a modified liquid-based cytology technique compared with conventional Pap smear.
Methods
Comparative analytical study conducted at the University Teaching Hospital of Kigali, Rwanda. Cervical samples from 151 women aged 35–70 years undergoing routine screening were collected using a split-sample method. For each participant, a conventional Pap smear was prepared, followed by immersion of the sampling device into a locally prepared ethanol–acetic acid–saline fixative for liquid-based cytology processing. Slides were independently examined by two pathologists using the Bethesda System. Specimen adequacy, background cleanliness, cellular morphology, and diagnostic categories were compared. Diagnostic agreement was assessed using percentage agreement and Cohen’s kappa.
Results
Specimen adequacy was higher with the modified liquid-based cytology technique than with conventional Pap smears (94.7% vs 83.4%). Unsatisfactory samples were reduced from 16.6% with conventional cytology to 5.3% with the liquid-based method. Liquid-based preparations demonstrated improved background quality, with reduced blood and mucus obscuring in 72.8% of cases compared with 41.1% of conventional smears. Overall diagnostic agreement was 89.4%, with substantial concordance (Cohen’s kappa = 0.80).
Conclusions
Manual Liquid-based cytology using an in-house fixative is a reliable alternative to conventional Pap smear cytology. Its implementation could reduce unsatisfactory sample rates, improve diagnostic quality, and enhance access to cervical cancer screening in resource-limited settings.

Background:
Histological grading of ductal carcinoma in-situ (DCIS) is central to clinical management and risk stratification. However, inter-observer variability remains a recognised limitation, particularly within intermediate grade (IG), which may affect treatment decisions and patient selection for de-escalation trials.
Purpose:
To examine the morphologic determinants of grading agreement at the level of individual DCIS ducts and evaluate whether interpretable morpho-spatial features can improve reproducibility of DCIS grading.
Methods:
Fourteen expert breast pathologists independently graded 1,383 ducts from 141 H&E whole-slide images. Automated feature extraction was performed using our DeepLab region segmentation model and a HoVer-Net-DCIS nuclear segmentation model fine-tuned on 50,000 manually annotated nuclei. This pipeline enabled extraction of 211 nuclear and ductal morpho-spatial features capturing epithelial nuclear spacing, nuclear density, pleomorphism, and duct architecture. A Random Forest classifier trained on these features was evaluated using case-grouped cross-validation. Slide-level grades were derived by aggregating duct-level predictions.
Results:
The segmentation pipeline demonstrated strong performance (DeepLab Dice 0.81;HoVer-Net nuclear classification F1 0.83). Three-tier DCIS grading showed moderate reproducibility (71–73%), with intermediate grade demonstrating the greatest instability and frequent transitions to both low- and high-grade. Nuclear spatial organisation strongly correlated with both grade and diagnostic agreement. Greater nuclear spacing and variability in nuclear size were associated with high-grade ducts and higher rater agreement, whereas nuclear crowding was associated with lower agreement. A Random Forest classifier trained on these features demonstrated strong discrimination of high-grade DCIS (slide-level AURC 0.87) and, in cross-group validation, achieved higher agreement with the majority consensus grade than any individual pathologist.
Conclusions:
DCIS grading uncertainty is concentrated within IG and is largely explainable by quantifiable nuclear spatial architecture. An interpretable morpho-spatial model can robustly standardise the clinically actionable high-grade boundary and provide a reproducible framework for decision support. The algorithm is currently being evaluated on longitudinal cohorts with multi-rater comparison.

Background: Digital autopsy, in which the primary investigation is cross sectional imaging, now accounts for almost 20% of coroners’ autopsies. A key limitation of post-mortem computed tomography (PMCT) is its inability to directly measure heart weight and thus identify pathological cardiac hypertrophy. Standard volumetric analysis of the heart is not possible on PMCT due to the inability to reliably visualise cardiac chambers.

Purpose: The aim of this study was to determine whether cardiac measurements taken from contrast-enhanced PMCT scans can be used to estimate heart weight.

Methods: Twenty-five cases that underwent both PMCT with coronary angiography and subsequent invasive autopsy were retrospectively analysed. Cardiac measurements were taken from archival PMCT scans (transverse heart diameter, transverse chest diameter, ventricular septal thickness, LV wall thickness, coronal short-axis LV diameter). Heart weight measured at invasive autopsy and demographic data (age, sex, height, and body weight) were extracted from post-mortem reports.

Results: Mean heart weight was 482g (range 354–744g). Transverse heart diameter on PMCT correlated with heart weight (R=0.688). A multivariable regression model (n=23) incorporating body weight improved predictive performance (adjusted R²=0.45 for heart diameter alone, R²=0.733 also including body weight), explaining 73% of the variability in heart weight in this cohort. An exploratory geometric model was created to estimate cardiac muscle volume using transverse heart diameter and composite LV wall thickness. Estimated muscle volume demonstrated a strong positive correlation with heart weight (R=0.776). Linear regression analysis formed the equation: Heart weight (g) = 0.000215 × cardiac muscle volume (mm³) + 330.25.

Conclusion: Combining PMCT-derived measurements and body weight provides a feasible method of predicting heart weight. An additional model incorporating LV wall thickness can be used to estimate heart weight independent of body weight. External validation of the models in larger, more diverse cohorts is required before implementation in routine coronial service.

Background: Ubiquitination-deubiquitination pathways regulate proteosome mediated protein degradation and may be dysregulated in oestrogen receptor (ER)-driven breast cancer (BC). Oestrogen and its metabolites can -induce DNA damage, which, if unrepaired, may accelerate mutagenesis, promote aggressive tumour phenotypes and contribute to endocrine therapy resistance. However, The clinicopathological significance of DNA damage response (DDR) regulating ubiquitin ligases (UBLs) and deubiquitinases (DUBs) in ER-positive BC remains largely unknown.
Methods: Expression of UBLs (DDB2, CUL4A, HLTF, HUWE1 and RAD18) and DUBs (USP7, USP1, USP5 and PSMD14) was assessed by immunohistochemistry in 1406 ER-positive early-stage BCs. Transcriptomic analyses were performed using publicly available data sets comprising approximately 15,000 tumours. Pre-clinically, ER-positive BC cell lines [MCF7 (p53 wild type) and T47D (p53-mutated)] were DDB2 depleted by siRNAs and evaluated for ER signalling, proliferation, invasion, tamoxifen sensitivity, cell cycle progression, DNA double-strand-break accumulation and apoptosis.
Results: Low expression of DDB2, CUL4A, HLTF, RAD18, USP7 and USP1 was associated with poorer breast cancer specific survival (BCSS) (all p<0.05). In multivariate analysis, DDB2 remained an independent predictor of poor BCSS (p=0.009) alongside nodal stage (p=0.008), tumour grade (p<0.001) and age (p=0.007). Low DDB2 expression was associated with larger tumour size (p<0.001), high grade tumours (p<0.001), de-differentiation (p=0.003), increased mitotic activity (p<0.001), lymphovascular invasion (p=0.03), lymph node metastases (p=0.02) and high-risk Nottingham Prognostic Index (NPI) (p<0.001). In p53 wild type tumours, low DDB2 expression was linked with adverse survival (p=0.007). Similarly, low DDB2 transcript levels remained associated with poorer survival (p<0.01).
Preclinically, DDB2 depletion in MCF7 and T47D cells increased proliferation and invasion, enhanced ER signalling and upregulated multiple genes involved in ER signalling pathway. Compared with controls, DDB2 depleted cells demonstrated increased resistance to tamoxifen, accelerated G1/S cell cycle progression with upregulation of proteins involved in G1/S transition [MCF7 (upregulation of CDK2-Cyclin E) and T47D (upregulation of CDK4-Cyclin D)], with reduced γ H2AX accumulation and reduced apoptotic cells.
Conclusions: DDB2 deficiency is associated with endocrine therapy resistance in ER-positive breast cancer and may represent a potential biomarker for risk stratification and therapeutic targeting.