Use of a single-scan Lung-RADS for pulmonary nodule assessment in resource-limited clinical settings.

Early identification of pulmonary lesions is a critical factor in enhancing patient prognosis and survival rates. This study systematically evaluates the diagnostic performance of five deep learning architectures, ConvNeXt Base, ResNet50, EfficientNetV2 Small, InceptionV4, and Xception, for the three-class categorization of Computed Tomography (CT) scans into Benign, Malignant, and Normal categories. Utilizing the public IQ OTH NCCD dataset, we applied a transfer learning approach with ImageNet weights, complemented by a robust training pipeline incorporating dynamic data augmentation and early stopping to mitigate overfitting and ensure model generalization. Model efficacy was rigorously assessed on an independent test set using accuracy, precision, recall, and F1-score metrics. Experimental results indicate that InceptionV4 emerged as the most reliable architecture, achieving an overall accuracy of 0.988 and a macro-averaged F1-score of 0.976. Notably, this model demonstrated perfect sensitivity for the pathologically critical malignant class, achieving a recall rate of 1.00, thereby prioritizing clinical safety. These findings confirm that advanced neural networks can serve as dependable secondary opinion systems for clinicians. Given its superior sensitivity and balanced diagnostic profile, InceptionV4 represents a promising candidate for integration into automated lung cancer screening workflows to improve diagnostic precision.

The growing adoption of Internet of Things (IoT) devices has heightened the risk of network intrusions, demanding robust security mechanisms. This research proposes a Convolutional Neural Network (CNN) based Intrusion Detection System (IDS) optimized with the Multi-Objective Hybrid Non-Dominated Sorting Genetic Algorithm II (MOHNSGA-II) to boost the accuracy and efficacy of intrusion recognition in IoT environments. The IDS framework uses Principal Component Analysis (PCA) to collect charactersitic and the Fisher score algorithm to choose which features to use. This reduces the number of dimensions in the data and makes detection more accurate. The CNN model, designed to classify intrusions into benign, Mirai botnet, and GAFGYT botnet categories, is optimized using MOHNSGA-II to fine-tune hyperparameters and achieve superior performance. The proposed model does much better than traditional classifiers like Random Forest (RF) and Support Vector Machine (SVM), as shown by its accuracy of 98% on the N-BaIoT dataset, with precision, recall, and F1 scores of 97%, 98%, and 99%, respectively. This research discourses critical encounters in IoT security by providing a scalable, efficient, and reliable IDS framework capable of real-time intrusion detection. Future work would focus on enhancing generalizability across diverse datasets and reducing computational complexity to support resource-constrained IoT environments.

Introduction: Non thyroidal neck lesions represent a heterogeneous group of pathologies arising from lymph nodes, salivary glands, softtissues, and other cervical structures. Fine Needle Aspiration Cytology (FNAC) is a minimally invasive and cost-effective diagnostic tool, but its accuracy in differentiating benign from malignant lesions requires continuous evaluation against histopathology, the gold standard. Aim: To evaluate the spectrum of non thyroidal neck lesions using FNAC, categorise them cytologically, and assess the diagnostic accuracy of FNAC by correlating it with histopathology. Materials and Methods: This prospective, cross-sectional study was conducted in the Department of Pathology, Government Medical College, Krishnagiri, Tamil Nadu, India from April 2024 to March 2025. A total of 210 consecutive cases of non thyroidal neck lesions underwent FNAC, of which 103 cases had subsequent histopathological evaluation. Cytological diagnoses were classified into inflammatory, benign, and malignant categories. Histopathological specimens were processed using routine paraffin embedding and Haematoxylin and Eosin (H&E) staining. Concordance between FNAC and histopathology was assessed, and diagnostic indices—sensitivity, specificity, Positive Predictive Value (PPV), Negative Predictive Value (NPV), and overall diagnostic accuracy—were calculated. Statistical analysis was performed using Pearson’s Chi-square test (χ²) with IBM Statistical Package of Social Sciences (SPSS) Statistics Version 25.0. Results: Of the 210 cases, 115 (54.76%) were inflammatory, 51 (24.29%) benign, and 44 (20.95%) malignant. Histopathological correlation in 103 cases revealed 21 inflammatory (20.4%), 51 benign neoplastic (49.5%), and 31 malignant neoplastic (30.1%) lesions. FNAC–histopathology concordance was 92.2% (95/103), with discordance in 7.8% (8/103). Diagnostic performance metrics were: sensitivity 92.68%, specificity 90.48%, PPV 97.43%, NPV 76.00%, and overall diagnostic accuracy 92.23%. Conclusion: FNAC is a highly sensitive and specific preliminary diagnostic modality for non thyroidal neck lesions, showing excellent concordance with histopathology. While it can reliably guide initial clinical management, histopathological confirmation remains essential in inconclusive or suspicious cases.

Abstract The term “atypical spindle cell/pleomorphic lipomatous tumor” was introduced in the WHO Classification of Soft Tissue Tumors in 2020. This tumor is an adipocytic neoplasm of benign or low-grade category, is clinically indolent, has poorly circumscribed margins, and composed of mature adipocytes, lipoblasts, atypical spindle-shaped cells, and multinucleated cells. A 75-year-old male presented with a paratesticular mass. On microscopic examination, the tumor showed a mixture of two components—adipose and fibrous tissue components—which blended with each other along with scattered atypical giant cells. Immunohistochemically, smooth muscle actin showed diffuse positivity in the spindle cells. S100 was negative in the spindle cells and giant cells but was positive in the nuclei of some adipocytes. The Ki-67 index was only 5%. CD34 and desmin were positive in the blood vessel walls—in endothelial cells and muscle layer respectively—and negative in the giant cells. Immunohistochemistry (IHC) with MDM2 (murine double minute 2) and Rb (retinoblastoma) was negative, while CDK4 (cyclin-dependent kinase 4) was variably positive in the nuclei of the spindle cells. The spectrum of adipocytic tumors that show overlapping morphologic features may pose diagnostic difficulty. Precise diagnosis of ASPLT is important, as it can be misdiagnosed as an intermediate grade or malignant lipomatous tumor. A tumor can be diagnosed as ASPLT when it is composed of a heterogeneous mixture of adipocytes, spindle cells with focal atypia, and multinucleated cells. Lipoblasts are not mandatory for diagnosis. IHC with MDM2, Rb1, Ki67, and molecular testing is helpful in differentiating benign ASPLT from other entities and in ensuring a better prognosis. CDK4 is not found to be useful.

Breast cancer remains a significant global health concern, emphasizing the need for advanced and accurate diagnostic tools. This research paper focuses on the application of a Transfer Learning model for the detection of breast cancer in mammography images. Leveraging the power of deep learning, Transfer Learning enables the utilization of pre-trained models on large datasets, optimizing performance even with limited data availability. The study employs a diverse dataset comprising mammography images from various sources, ensuring a comprehensive representation of breast cancer cases. A Convolutional Neural Network (CNN) architecture, pre-trained on a vast dataset, is fine-tuned using the mammography dataset to harness its feature extraction capabilities for breast cancer detection. This approach allows the model to learn intricate patterns and abnormalities indicative of malignancies. Key steps involve the pre-processing of mammography images to enhance the quality and extraction of relevant features through the Transfer Learning model. The research investigates the model’s efficacy in distinguishing between benign and malignant cases, evaluating its accuracy, sensitivity, specificity, and precision. The research proposed EfficientNet B3 with a DenseNet transfer learning model for classifying mammography images into benign or malignant categories. Also, the impact of varying architectures and hyperparameters on the model’s performance is explored for optimization. Results demonstrate promising outcomes, with the Transfer Learning model exhibiting a high degree of accuracy in breast cancer detection. The model’s ability to generalize across diverse datasets underscores its robustness and potential for real-world clinical applications. Visions gained from this research contribute to the ongoing discourse on the integration of advanced technologies in breast cancer diagnostics. This research signifies a crucial step towards enhancing the accuracy and efficiency of breast cancer detection, emphasizing the potential of Transfer Learning in revolutionizing mammography-based diagnostic approaches. Findings show that the EfficientNet-B3 and DenseNet have loss values ranging between 0.9, while VGG16’s loss values are significantly higher, ranging from 7. The findings hold implications for improving early detection, facilitating timely interventions, and ultimately advancing outcomes for individuals at risk of breast cancer.

The American College of Radiology Thyroid Imaging Reporting and Data System (ACR TI-RADS) incorporates conventional grayscale ultrasonography (US) as the only imaging technique without considering the clinical and demographic characteristics of patients. This study assessed whether the addition of demographic information, color Doppler US (CDUS), and strain elastography could enhance malignancy risk stratification beyond the current ACR TI-RADS criteria. This prospective study enrolled 556 adult patients with thyroid nodules ≥10 mm who were referred for fine-needle aspiration (FNA) according to the ACR TI-RADS recommendations. All nodules underwent standardized US evaluations and vascularity assessments using CDUS and strain elastography, with cytological analysis performed according to the Bethesda system. Surgical pathology was the gold standard for malignancy when available. Applying elastography ratio (ER) thresholds (>1.60, >0.44, and >0.54 for ACR TI-RADS categories 3, 4, and 5, respectively) as an additional criterion for FNA reduced the number of procedures from 501 to 260, without missing any malignant cases. Notably, elastography demonstrated an excellent discriminative performance in ACR TI-RADS 3 nodules (Youden index 0.994, AUC 0.994), supporting its value in improving risk stratification in this challenging, predominantly benign category. Integrating elastography into the ACR TI-RADS framework can optimize FNA utilization in the management of thyroid nodules by reducing the number of unnecessary aspiration biopsies.

Respiratory cytology plays a vital role in the early diagnosis of pulmonary diseases, particularly lung cancer, which remains a leading cause of cancer-related mortality worldwide. To enhance standardization and reduce inter-observer variability in lung cytology reporting, the World Health Organization (WHO), in collaboration with the IAC and IARC, introduced a new five-tiered reporting system for lung cytopathology in 2023. This study aims to retrospectively apply the WHO system to respiratory samples at a tertiary care center in India and evaluate diagnostic performance and risk of malignancy (ROM) for each category. A retrospective observational study was conducted from June 2022 to May 2024, analyzing 736 respiratory cytology samples, including BAL, EBUS-TBNA, bronchial wash, bronchial brush, and sputum. Samples were reclassified using the WHO system, and 193 cases had histopathological correlation. ROM was calculated for each category, and diagnostic performance metrics sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and diagnostic accuracy (DA) were assessed using histopathology as the gold standard. Given that histologic correlation was available in only 26.2% of cases, diagnostic performance estimates were interpreted with caution and may have been influenced by verification bias. Of the total cases, 10.2% (75/736) were insufficient/inadequate, 80.1% (589/736) benign, 3.0% (22/736) atypical, 2.0% (15/736) suspicious, and 4.7% (35/736) malignant. The ROM was 43.7% for insufficient, 39.7% for benign, 50.0% for atypical, 75.0% for suspicious, and 100% for malignant. Bronchial brush demonstrated the highest DA (81.3%), while BAL had low sensitivity (10.0%) despite high specificity (98.5%). EBUS-TBNA showed moderate sensitivity (50.0%) and lower specificity (55.6%). The WHO Reporting System provides an effective framework for stratifying respiratory cytology samples. While categories such as malignant and suspicious showed high diagnostic reliability, false negatives in benign and inadequate categories highlight the need for integrated clinical and radiological assessment. Bronchial brush and wash emerged as the most diagnostically accurate specimen types. Further prospective studies are warranted to validate these findings across broader clinical settings.

Breast tumor classification from mammogram images plays an essential role in supporting clinical decision-making, particularly because manual interpretation is often challenged by variations in breast tissue density and suboptimal image quality. This study develops a three-class classification model for normal, benign and malignant categories using the ResNet50 architecture with a transfer learning strategy on the mini-MIAS dataset, which contains 322 images with an imbalanced class distribution. Three optimizers are compared, namely Adam, RMSProp and SGD. Adam represents an adaptive moment-based optimization approach. RMSProp emphasizes stable updates under fluctuating gradients. SGD with momentum serves as a conventional baseline relying on direct gradient updates. The model is trained using a 60 percent training and 40 percent validation split with class weighting and evaluated through accuracy, AUC and F1-score metrics. Experimental results show that Adam achieves the highest performance with 68.27 percent accuracy, 88.58 percent AUC and an F1-score of 0.68. RMSProp attains 58.63 percent accuracy, 76.05 percent AUC and an F1-score of 0.59. SGD yields the lowest performance with 44.18 percent accuracy, 61.33 percent AUC and an F1-score of 0.44. Confusion matrix analysis for the Adam configuration indicates reasonably consistent recognition across all classes, although misclassification remains present. The findings demonstrate that adaptive optimizers are more effective for training ResNet50 on small and imbalanced mammogram datasets. This study provides a foundation for developing more reliable computer-aided diagnostic systems for early breast cancer detection.

To detect breast cancer at an early stage, proper analysis of histopathology images is essential.We propose leveraging Convolutional Neural Network (CNN) technology, specifically DenseNet-201, for diagnosing cancer, aiming to enhance effectiveness and alleviate physician's workload. Our task is to classify cases into benign and malignant categories. Utilizing the BreakHis dataset for both training and evaluation, our method achieves a classification accuracy of 97.46%. Notably, we achieve an F1-score of 0.91, with recall and precision of 0.98 and 0.95, respectively. Through experimental validation and comparison with prior research, we demonstrate the reliability and efficiency of our proposed approach. These results underscore the potential of employing deep learning methods for histopathology image analysis in breast cancer diagnosis, offering promising prospects for improving healthcare outcomes.

The Bethesda System for Reporting Thyroid Cytopathology (TBSRTC) provides a standardised framework for thyroid fine needle aspiration cytology (FNAC). Deep learning approaches, particularly transfer learning, have shown potential for cytology but are rarely applied to TBSRTC categorisation. To evaluate the performance of an ensemble soft voting transfer learning model in categorising thyroid cytology smears according to TBSRTC. A retrospective study of 94 thyroid FNAC cases with 949 representative images was conducted. Six transfer learning models (Xception, ResNet50V2, DenseNet121, MobileNetV2, InceptionV3, EfficientNetB3) were combined using ensemble soft voting with a weighted average. Model performance was assessed using sensitivity, specificity, precision (PPV), negative predictive value (NPV), F1 score and area under the receiver operating characteristic curve (AUCROC). DenseNet121 achieved the highest overall sensitivity (0.91) and specificity (0.98) among all single models. The weighted ensemble model achieved an F1 score of 0.867, marginally below DenseNet121 but with superior precision (0.882) and NPV (0.968). AUCROC was highest in DenseNet 121 (0.99) followed by the weighted ensemble (0.98). This is the first study to apply ensemble transfer learning to TBSRTC categorisation. The approach demonstrated strong predictive accuracy, particularly for benign and malignant categories. Larger datasets, ideally with whole-slide imaging, are needed to further validate these findings.

Fine-needle aspiration cytology (FNAC) remains widely used for the evaluation of palpable breast lesions, particularly in resource-limited settings, though histopathology is the gold standard. The International Academy of Cytology (IAC) Yokohama System provides a standardized five-tier reporting framework with defined risk of malignancy (ROM). This study aimed to evaluate its diagnostic performance, reproducibility, and applicability in a tertiary care setting. A total of 548 breast FNACs from 533 patients were reclassified both prospectively and retrospectively according to the IAC Yokohama categories. Histological correlation was available in 228 cases. Interobserver agreement was assessed among two senior cytopathologists and one junior pathologist using unweighted and weighted kappa statistics. Diagnostic performance was calculated against histology using three definitions of test positivity (Cat-5 only; Cat-4 & 5; Cat-3-5). A total of 548 FNACs from 533 patients were reclassified using the IAC Yokohama system: Cat-1 (36, 6.6%), Cat-2 (318, 58.0%), Cat-3 (14, 2.6%), Cat-4 (25, 4.6%), and Cat-5 (155, 28.3%). Histological correlation was available in 228 cases. The observed ROM was 50% for Cat-1, 7.9% for Cat-2, 45.5% for Cat-3, 93.8% for Cat-4, and 100% for Cat-5. Diagnostic accuracy improved with reclassification: in Group A (Cat-5 positive), sensitivity increased to 78.4% and specificity was 100%; in Group B (Cat-4 & 5 positive), sensitivity was 90.9% and specificity was 97.9%; and in Group C (Cat-3-5 positive), sensitivity reached 94.7% with a specificity of 89.6%. ROC analysis confirmed superior discrimination for the Yokohama system (AUC 0.94 vs. 0.88 for the original). Interobserver reproducibility was excellent, with weighted kappa values of 0.954 (P1 vs. P2), 0.942 (consensus vs. P3), and 0.939 (P2 vs. P3). Agreement was highest in benign and malignant categories and lowest in atypical and suspicious groups. The IAC Yokohama System showed high diagnostic accuracy, excellent reproducibility, and reliable risk stratification. By reducing false negatives and minimizing equivocal reporting, it improved alignment with histology compared with the conventional descriptive system, supporting its routine use in breast cytology practice.

Effective lung cancer detection from CT scans remains critically challenged by class imbalance where benign and normal cases are underrepresented, leading to biased machine learning models with reduced sensitivity for minority classes and potentially missed diagnoses in cancer screening applications. We present a comprehensive comparative analysis of Diffusion Models and Deep Convolutional Generative Adversarial Networks (DCGANs), both incorporating modern architectural enhancements including spectral normalization, self-attention mechanisms, and conditional generation, for addressing class imbalance in lung cancer CT classification. Using the IQ-OTH/NCCD dataset comprising 1,097 CT images across normal, benign, and malignant categories with statistical validation across 10 independent runs, we evaluated both approaches through quantitative image quality metrics (Fréchet Inception Distance, Kullback-Leibler divergence, Kernel Inception Distance, and Inception Score) and downstream classification performance. While Diffusion models consistently outperformed DCGANs across most image quality measures, the clinical significance was confirmed through task-based validation. Both generative approaches successfully addressed class imbalance: DCGAN-augmented datasets achieved overall accuracy of 0.9760 ± 0.0116 with benign recall improvement from 0.833 to 0.933, while Diffusion-augmented datasets reached superior performance of 0.9959 ± 0.0068 with perfect benign recall (1.000 ± 0.000). Critically for cancer screening where false negatives carry severe consequences, Diffusion maintained the highest malignant detection sensitivity (0.997 ± 0.008) with substantially lower performance variance, demonstrating more consistent synthetic data quality. These findings establish that while both modern architectures can mitigate class imbalance, Diffusion models' superior recall performance and lower variability position them as the preferred approach for high-stakes clinical applications, demonstrating that ultimate validation must prioritize downstream clinical task performance over image quality metrics alone.

Deep Learning methods have become a powerful tool in medical imaging, with great potential to improve diagnostic accuracy and support early disease detection. This is especially critical for breast cancer, one of the most common cancers among women, where early detection of abnormal tissue is crucial to improving survival rates. In this paper, we explore the application of Deep Learning techniques to segment and classify breast masses as malignant or benign using ultrasound images, aiming to support breast cancer diagnosis. We propose a modular dual-stage pipeline that first segments suspicious regions and then classifies them into benign or malignant categories. The framework is designed to flexibly integrate different backbone architectures, allowing adaptation to task- or dataset-specific requirements. Experimental results show that, within this pipeline, DeepLabV3+ with a ResNet34 encoder provided the most accurate segmentation, while lightweight classifiers such as MobileNetV3-Small and EfficientNet-B0 yielded the best classification performance. Moreover, an ablation study was conducted to tune parameters and determine their optimal configuration. Finally, our approach was tested on two breast ultrasound datasets, and the results show promising improvements in diagnostic accuracy, demonstrating the potential of our method to enhance early breast cancer detection.

Liver Observations in Chronic Liver Disease on Noncontrast Abbreviated Magnetic Resonance Imaging MRI (AMRI): Proposal of Modified Liver Imaging Reporting and Data System (AMRI-LI-RADS) Categorization.

Disclosure: R. Conard: None. J. Lin: None. B.R. Haugen: Veracyte, Inc. M. Alshalalfa: Veracyte, Inc. H. Yangyang: Veracyte, Inc. J.P. Klopper: Veracyte, Inc.. W.S. Goldner: None.Introduction: Vitamin D receptor (VDR) expression and the relative synthesis and catabolism of vitamin D (VD) appears to play a role in tumor growth regulation. The aim of this study was to characterize VDR, VD regulatory enzymes, and VD pathway target gene expression in thyroid nodules sent for Afirma Genomic Sequencing Classifier (GSC) molecular testing. Methods: mRNA expression of VDR and genes affecting VD synthesis (CYP27B1 (1α-hydroxylase - generation of active 1,25 VD), CYP27A1 (sterol 27-hydroxlase - conversion of VD to 25,VD), and CYP24A1 (24-hydroxylase - catabolism of 1,25 VD )) were analyzed across 47,695 thyroid nodules ((B)ethesda III-VI cytology). Differential expression of these targets was explored across Afirma benign (GSC-B) and suspicious (GSC-S) categories, cytology group, and relative to the BRAFV600E variant. The expression of VD target genes (CD14, ORM1, CALM, CLMN) was assessed and a VD activity score was derived by averaging the expression level of these targets. Results: 30,259 nodules were GSC-Benign (GSC-B), 15,815 were GSC-Suspicious (GSC-S), and 1,621 were cytologically B V or VI. In GSC-S and BV/VI nodules, VDR expression was lower in nodules with BRAFV600E compared to BRAFwt. CYP27A1 expression was not significantly different from GSC-B levels in any other category. CYP24A1 expression was higher in GSC-S compared to GSC-B, and in B V/VI nodules, and higher still in BRAFV600E compared to BRAFwt subgroups. CYP27B1 expression was lower in GSC-S and BV/VI nodules compared to GSC-B nodules and even lower in BRAFV600E mutant GSC-S and BV/VI nodules compared to those with BRAFwt. The same pattern was seen for each VD signaling target gene. The novel VD activity score followed the same pattern, with the lowest levels in samples with BRAFV600E. Samples with RAS like alterations (RAS family and PAX8-PPARy fusion) had a VD activity score lower than GSC-B and higher than those with BRAFV600E. Additionally, the VD activity score was strongly positively correlated with VDR expression (r=0.45) and strongly negatively correlated with ERK activity (r= -0.48). Conclusion: Cytologically indeterminate thyroid nodules with suspicious GSC results and malignant nodules appear to have reduced VD signaling. This VD signaling is further diminished in the presence of BRAFV600E and to a lesser extent, RAS like alterations. Similarly, the novel VD activity score we developed is lower in molecularly suspicious nodules and those with high-risk molecular variants and fusions. The VD activity score is also negatively correlated with ERK signaling, a marker of MAPK signaling and tumor aggressiveness. It is possible the VD activity score could serve as a preoperative marker that may modify prognosis in the absence of variants and fusions correlated with low or high-risk tumors, or in the presence of intermediate risk mutations.Presentation: Monday, July 14, 2025

Introduction: Distinguishing between nondiagnostic (ND) and benign (B) categories in lung cytopathology remains clinically challenging, especially given the significant overlap and the high risk of malignancy (ROM) often reported for ND cases. The 2022 WHO Reporting System for Lung Cytopathology addresses these issues but acknowledges that ND may carry up to a 60% ROM. We conducted a large, multicenter study to clarify the ND-B boundary and evaluate how radiologic findings influence ROM. Methods: From 12 institutions, 363 consecutive lung cytopathology cases categorized as insufficient/inadequate/ND (I/I/ND) or B with histopathological follow-up were analysed. The locally categorized cytopathological cases were subclassified centrally into: ND with insufficient cellularity (IS-C), artefactual/sample preparation error (IS-P), non-representative (NR1: no suspicious lesion; NR2: suspicious lesion); and B (B1: benign cells, no suspicious lesion; B2: benign cells, suspicious lesion). ROM was defined as the percentage of histologically confirmed malignancies in each group. Results: Overall, 60.6% (220/363) of cases were confirmed as malignant on histopathological evaluation. Within the ND category (n = 149), 70.5% (105/149) were malignant, exceeding the malignancy risk range estimated by the WHO system (40–60%). In comparison, the ROM for cases classified as B (n = 214) was 53.7% (115/214), which is consistent with the WHO system reference range. Notably, when ND or B cytopathology coincided with suspicious imaging findings (NR2 [n = 57] or B2 [n = 124]), the ROM exceeded 75% (134/181). These results indicate that subclassification based on imaging findings provides a more refined estimation of malignancy risk. Cases with B cytopathology may still carry a high likelihood of malignancy when imaging features are suspicious, reinforcing the importance of integrated diagnostic evaluation. Conclusions: These findings demonstrate that imaging correlation is critical for accurate risk assessment in the overlap between the ND and B cytopathology categories. Subclassification of ND and B cases based on imaging features and consistent reporting of ROM can help identify patients who may benefit from repeat sampling or further diagnostic evaluation. This approach has the potential to enhance diagnostic accuracy and improve clinical decision-making.

Cybersecurity faces persistent challenges due to the rapid growth and complexity of network-based threats. Conventional rule-based systems and classical machine learning approaches often lack the adaptability required to detect advanced and dynamic attacks in real time. This study introduces a deep Q-learning framework for autonomous threat detection and mitigation within a simulated network environment that reflects realistic traffic, malicious behaviors, and system conditions. The framework incorporates experience replay and target network stabilization to strengthen learning and policy optimization. Evaluation was performed on a synthesized dataset containing benign traffic and multiple attack categories, including distributed denial of service (DDoS), phishing, advanced persistent threats, and malware. The proposed system achieved 96.7% detection accuracy, an F1-score of 0.94, and reduced detection latency to 50 ms. These results surpassed the performance of rule-based methods and traditional classifiers such as support vector machines, random forests, convolutional neural networks, and recurrent neural networks. A central contribution lies in combining dynamic feature selection with reinforcement learning (RL), allowing the agent to adapt to evolving threats and diverse network conditions. The findings demonstrate the potential of deep reinforcement learning (DRL) as a scalable and efficient solution for real-time cybersecurity defense.

Background: Fluid cytology plays a pivotal role in the evaluation of serous effusions and other body fluids. It offers a minimally invasive, cost-effective, and rapid diagnostic tool for identifying a wide spectrum of benign and malignant conditions. Despite being a vital part of diagnostic pathology, the utility and diagnostic spectrum of cytological examination in body fluids remain underexplored in certain regional healthcare settings, such as Bihar. Objective: To assess the diagnostic utility, spectrum of cytological findings, and the distribution of benign versus malignant cases in various body fluids (pleural, peritoneal, and cerebrospinal fluid) over a one-year period at a tertiary care hospital in Bihar. Materials and Methods: A retrospective study was conducted in the Department of Pathology, Darbhanga Medical College and Hospital, Darbhanga, Bihar, over a period of 12 months (August 2024 to July 2025). A total of 125 fluid samples were included, comprising pleural, peritoneal, and cerebrospinal fluids received in the cytology section. All samples were processed using standard centrifugation, smear preparation, and staining techniques (Papanicolaou, Giemsa, and H&E where needed). Cytological findings were categorized into benign, suspicious, and malignant categories. Results: Out of the 125 fluid samples analyzed, 102 (81.6%) were diagnosed as benign/reactive, 13 (10.4%) were malignant, and 10 (8%) were suspicious for malignancy. Among malignant cases, adenocarcinoma was the most frequently identified malignancy, predominantly involving pleural fluid. The diagnostic yield was highest for pleural effusions, followed by ascitic and cerebrospinal fluids. Cytological examination provided significant diagnostic clues in cases of suspected malignancy and infection, thus aiding in clinical decision-making. Conclusion: Fluid cytology is an indispensable diagnostic modality in the assessment of body fluids. Its noninvasive nature, combined with reasonable accuracy, makes it a valuable first-line investigation. In resource-constrained settings like Bihar, it proves to be both practical and cost-effective, especially in early detection of malignancies and infectious diseases.

Precise classification of lung cancer stages based on CT images remains a significant challenge in oncology. This is vitally necessary for determining prognosis and creating practical treatment plans. Traditional methods mainly rely on human interpretation, which can be inconsistent and prone to fluctuation. To overcome these limitations an automated deep learning model based on the EfficientNet-B0 based architecture is proposed. Explainable AI features enhanced through Gradient-weighted Class Activation Mapping (Grad-CAM) help further boost this model. Training of the model was conducted with 1,190 CT scans from the IQ-OTH/NCCD dataset. All the images fell into the benign, malignant, and normal categories. The suggested technique performs remarkably well, reaching 99% accuracy, 99% precision, and recall rates of 96% for benign cases, 99% for malignant cases, and 100% for normal occurrences. Grad-CAM makes the model more interpretable and transparent by providing visual explanations of its results. It identifies the most important regions in the scans that significantly contribute to the classification results. Apart from contributing to the field of medical image analysis, accurate precision and complete explanations also bring automated diagnosis systems credibility and reliability.