The prognostic nutritional index (PNI), calculated from serum albumin and lymphocyte count, reflects a patient’s immune-nutritional status and has been proposed as a prognostic marker in hepatocellular carcinoma (HCC). However, its role in advanced HCC patients treated with atezolizumab plus bevacizumab (Ate/Bev) remains unclear. In this study, we aimed to evaluate the prognostic value of PNI in patients receiving first-line Ate/Bev therapy.

We retrospectively analyzed 362 patients with unresectable HCC who received Ate/Bev between November 2020 and June 2023 across two centers. Based on prior literature, a cutoff of 45 was used to classify patients into low-PNI (<45) and high-PNI (≥45) groups. Propensity score matching was performed to balance baseline characteristics.

After propensity score matching, 130 patients (65 per group) were included in the analysis. The high-PNI group showed a significantly lower incidence of grade ≥ 3 treatment-related adverse events (10.8% vs. 24.6%, p = 0.039), a higher objective response rate (38.4% vs. 20.0%, p = 0.037), and significantly longer overall survival (16.7 vs. 7.9 months, p = 0.009). Although progression-free survival was longer in the high-PNI group (4.8 vs. 3.0 months), the difference was not statistically significant (p = 0.597). Multivariate analysis confirmed that PNI was an independent predictor of overall survival (hazard ratio: 0.574, 95% confidence interval: 0.353–0.933, p = 0.025), after adjusting for vascular invasion, alpha-fetoprotein levels, concurrent therapy, and post-treatment interventions.

PNI is an independent prognostic factor for overall survival in advanced HCC patients treated with Ate/Bev in real-world clinical practice. Incorporating PNI into routine assessments may enhance risk stratification and guide therapeutic decision-making.

The history of screening for cervical cancer is rich with implementing cutting-edge ideas and technologies. From the very first “Pap smear” to the semi-automated and computerized systems of today, the way we screen for cervical cancer has changed dramatically in the past 75 years. With the advent of new techniques and more advanced machine learning algorithms, we sought to understand the current and future applications of artificial intelligence in clinical pathology around cervical cancer screening, diagnosis, and treatment.

A structured narrative review was conducted to examine the historical evolution and contemporary advances in cervical cancer screening, diagnosis, excisional management, and artificial intelligence applications. Peer-reviewed articles, consensus guidelines, and global policy documents published between January 1990 and March 2025 were identified through targeted searches of PubMed and review of reference lists from relevant publications. Search terms included combinations of “cervical cancer screening,” “Papanicolaou test,” “liquid-based cytology,” “HPV testing,” “colposcopy,” “loop electrosurgical excision procedure,” “digital pathology,” “deep learning,” and “artificial intelligence.” Emphasis was placed on multi-center validation studies, systematic reviews, regulatory and implementation guidance, and global health frameworks. Publications lacking methodological transparency or direct relevance to clinical or translational practice were excluded.

Through a review of the literature, we describe how innovations in conventional and liquid-based cytology, human papillomavirus testing, and organized screening programs established the current prevention framework. Building on this foundation, recent studies demonstrate promising performance of deep learning algorithms applied to conventionally prepared cervical cytology slides, with systems capable of binary normal versus abnormal classification as well as more granular diagnostic categorization. Artificial intelligence-assisted colposcopy and computer-vision approaches have also shown improved diagnostic concordance, workflow efficiency, and potential to expand screening capacity in resource-limited environments.

There has been much work done in the past several years surrounding the implementation of deep learning algorithms in regard to cervical cancer screening. The work in this field shows promise in enhancing diagnostic accuracy, streamlining diagnostic workflow, and decreasing turnaround times from specimen collection to rendering a diagnosis. However, there are still many technical, legal, and ethical questions that must be answered prior to widespread adoption of these algorithms for patient care.

Colorectal cancer histopathological grading relies on the accurate segmentation of glandular structures. Current deep learning–based methods depend heavily on large-scale pixel-level annotations that are labor-intensive and not amenable to clinical practice. Weakly supervised semantic segmentation offers a promising alternative; yet, existing class activation map–based weakly supervised semantic segmentation approaches often produce incomplete, low-quality pseudo-masks that overemphasize discriminative regions and fail to provide reliable supervision for unannotated glandular structures, limiting their suitability for dense histopathology segmentation under sparse supervision. We propose a novel weakly supervised teacher–student framework that leverages sparse pathologists’ annotations and an Exponential Moving Average–stabilized teacher network to generate refined pseudo-masks.

Our framework integrates confidence-based filtering, adaptive fusion of teacher predictions with limited ground truth, and curriculum-guided refinement, enabling the student network to progressively delineate and accurately segment unannotated glandular regions. We validated our framework on an institutional colorectal cancer cohort from The Ohio State University Wexner Medical Center, consisting of 60 hematoxylin and eosin-stained whole-slide images from independent patients with varying degrees of gland differentiation, as well as on public benchmarks including the Gland Segmentation dataset (derived from stage T3–T4 colorectal adenocarcinomas), TCGA-COAD, TCGA-READ, and SPIDER.

The proposed framework achieved strong performance on the institutional dataset despite limited annotations. On the Gland Segmentation dataset, it demonstrated competitive performance compared to both weakly and fully supervised approaches, achieving a mean Intersection over Union of 80.10% ± 1.52 and a mean Dice coefficient of 89.10% ± 2.10. Moreover, cross-cohort evaluations showed robust generalization on TCGA-COAD and TCGA-READ without requiring additional annotations, while reduced performance on SPIDER reflected pronounced domain shift.

Our framework provides an annotation-efficient and generalizable paradigm for accurate gland segmentation in colorectal histopathology, offering a practical pathway toward significantly reducing annotation burdens while preserving high segmentation fidelity.

Cervical cancer, driven mainly by persistent high-risk human papillomavirus infection, remains a major public health problem in Bangladesh, with 9,640 new cases and 5,826 deaths in 2022. Early detection of pre-cancerous cervical lesions (PCL) is essential, yet limited evidence exists on factors associated with PCL among Bangladeshi women. This study aimed to identify factors associated with PCL among women attending cervical cancer screening centers at selected tertiary hospitals.

An age-matched (±5 years) case-control study was conducted in two tertiary hospitals. Cases were women who tested colposcopy-positive for PCL, and controls were visual inspection with acetic acid-negative women attending the same screening centers. A total of 38 cases and 76 controls were included. Data were collected through face-to-face interviews using a structured questionnaire. Multivariable logistic regression identified factors associated with PCL, with significance set at p < 0.05.

A history of sexually transmitted infections (adjusted odds ratio (AOR) = 36.73; 95% confidence interval (CI): 3.25–414.83), pelvic infections (AOR = 6.48; 95% CI: 1.24–33.85), not living with a husband (AOR = 4.48; 95% CI: 1.06–18.90), and overweight/obesity (AOR = 3.58; 95% CI: 1.14–11.22) were significantly associated with higher odds of PCL. Menstrual irregularity, genital ulcer history, and number of lifetime sexual partners showed no significant association.

Sexually transmitted infections, pelvic infections, overweight/obesity, and not living with husband were identified as factors associated with PCL. Strengthened infection prevention, lifestyle counseling, and targeted health education may support ongoing cervical cancer prevention efforts in Bangladesh.

Infections are frequent and lethal complications of acute-on-chronic liver failure (ACLF). Reliable biomarkers to distinguish fungal from bacterial infections remain limited. Given the central role of immune dysfunction in ACLF, we aimed to evaluate the diagnostic value of serum cytokines in differentiating invasive pulmonary aspergillosis (IPA) from bacterial pneumonia (BP) in HBV-associated ACLF.

This retrospective case-control study enrolled ACLF patients admitted to the Tongji Hospital, between 2018 and 2022. Patients were categorized into IPA, BP, and non-infection groups. The BP and non-infection groups were propensity score-matched to the IPA cases. Serum cytokines levels (IL-1β, sIL-2R, IL-6, IL-8, IL-10, TNF-α) and clinical data were collected, with the diagnostic performance of these cytokines as biomarkers assessed via ROC curves.

A total of 32 IPA, 96 BP, and 96 non-infection patients were enrolled, with balanced baseline characteristics. Compared with the non-infection group, the IPA group had higher sIL-2R (1,606.00 vs. 1,211.50 U/mL, P = 0.019) and IL-6 (69.03 vs. 15.98 pg/mL, P < 0.001) levels, but lower IL-8 levels (62.20 vs. 132.00 pg/mL, P = 0.025). The BP group showed elevated sIL-2R (1,792.00 U/mL), IL-6 (49.42 pg/mL), IL-10 (13.40 pg/mL) levels compared to the non-infection group (all P < 0.001). Also, IL-8 was lower in the IPA group than in the BP group (62.20 vs. 176.00 pg/mL, P < 0.001) and its assessment could best distinguish IPA from BP (AUC = 0.743, cut-off = 76.60 pg/mL; sensitivity = 66.7%, specificity = 82.1%).

Serum IL-8 exhibited superior diagnostic value for IPA in patients with HBV-ACLF and could effectively discriminate Aspergillus infections from bacterial infections.