Whenever nanomaterials are used to address health science-related problems, the qualities of biocompatibility and biodegradability must be kept in mind, since they ensure that the materials circulating through the carrier biological system used to diagnose or treat a disease are properly eliminated once their function is fulfilled, interfering as little as possible with the normal functioning of the carrier. For this reason, carbon and synthesizing carbon nanostructures (CNPs) were selected to develop diagnostic probes in tumor tissues. Glucose (1 M sol. in deionized water) was taken as a carbon precursor, and hydrolysis was performed in an acidic (36% HCl) and basic (1 M NaOH) medium, using ultrasound for 4 h continuous, at 25 kHz, 30 degrees C, and a power of 90%. The sonicated solutions are neutralized at pH 7 and centrifuged to separate the supernatant, where the carbon nanostructures are obtained. The NPsC were characterized by UV-Vis spectrometry, and the photoluminescent properties by spectrofluorometer and IVIS Lumina, transmission electron microscopy (TEM), and FTIR. The results obtained were that the NPsC have a high emission wavelength (650 nm) when excited at 450 nm, for the NPsC in a basic medium maintain a higher intensity than the NPsC synthesized in an acidic medium.
This manuscript presents a model for reconstructing the temporal dynamics of a real-time telemetry system, where the reception time T-Rx,T-k and the transmission time TTx,k constitute the transport time T-k, enabling system sizing and optimization. The study is based on a testbed [12] updated with single-board computers and a real-time operating system, designed to obtain precise measurements. Temporal constraints in real-time telemetry are analyzed, and the Kalman filter algorithm is used to reconstruct its behavior. This work argues that telemetry process times in an inter net-based client-server scheme are affected by various factors (infrastructure, network topology, time of day, electromagnetic interference, protocols, etc.), causing variability that impacts temporal constraints and system performance. In conclusion, proper system sizing is ensured to optimize real-time communication and guarantee its optimal operation.
The ECG is a crucial tool for the prevention and diagnosis of cardiovascular diseases. However, manual analysis of large volumes of data is prone to errors and generates false alarms. In this work, we propose the design of a model for detecting anomalies in ECG signals, based on machine learning models (CAE, CAE + RF, CAE + SVM). Three approaches were evaluated using performance metrics such as accuracy, F1-score, recall, precision, MCC, ROC, and AUC. According to the results obtained, the model that shows the highest performance and robustness was CAE + RF. Additionally, this model underwent a validation stage with two test sets (A and B). In the first set, segments of normal heartbeats were extracted from the MIT-BIH. Subsequently, different types of artifacts were injected, and in set B: a VAE was trained with normal heartbeats, and the same artifacts as in set A were also injected. The CAE + RF model demonstrated robustness, achieving an AUC = 0.9991, precision = 0.9623, and F1-score = 0.9810. These findings demonstrate that the model allows for robust detection, offering a favorable balance in detecting anomalies and reducing false alarms.
Crime prediction using Artificial Intelligence has increased in technical complexity and public relevance; however, the evidence remains fragmented regarding implementation practices, publication quality, and scientific collaboration. This paper synthesizes and characterizes the literature on Artificial Intelligence applied to Crime Prediction, with emphasis on development languages, quartile levels, co-authorship networks, and thematic categories. A systematic review was conducted following Kitchenham and PRISMA 2020 guidelines, with searches in ProQuest, IEEE Xplore, Scopus, ScienceDirect, and SpringerLink through July 22, 2025. After applying inclusion and exclusion criteria and conducting quality assessment with a threshold of at least 24, 57 studies out of 70 (81.4%) were selected for technical and bibliometric analysis. The results show the dominance of Python as the implementation language (60%), with secondary presence of R (20%) and limited adoption of Java, C++, and Julia at approximately 6.7% each. Production is concentrated in Q1 journals (33/57), followed by Q2 (16/57) and Q3 (8/57), suggesting high visibility, although with heterogeneous impact across sources. Co-authorship analysis reveals core and bridge authors connecting subcommunities, while the thematic map highlights specialized lines such as AI Crime Analytics and AI Policing, as well as marginal topics that are still consolidating. In conclusion, the field is progressing toward more integrated approaches but requires comparable standards, reproducible traceability, and governance frameworks for responsible deployment.
The rapid expansion of biomedical literature has made it increasingly difficult to intersect findings and uncover novel hypotheses, particularly within neurodegenerative disease research. This work presents a domain-adapted approach for hypothesis extraction and synthesis using compact large language models (Qwen3 0.6B-8B parameters). Through parameter-efficient fine-tuning on a curated corpus of open-access biomedical papers, these models work over identifying, relate, and visualize research hypotheses as interconnected knowledge graphs. Their performance is evaluated against a larger, general-purpose models in zero-shot conditions, demonstrating that smaller, specialized models can achieve comparable or superior interpretability and relevance. The study highlights the potential of lightweight, domain-focused LLMs as practical tools for accelerating discovery and improving transparency in biomedical research.
Computing topological indices for molecular graphs is a key element in computational chemistry, particularly when analyzing the structural and functional properties of chemical compounds. Among these indices, the Merrifield-Simmons (M-S) index, defined as the total number of independent vertex sets in a molecular graph, provides valuable information on the compartmentalization, stability, and connectivity of a molecular graph. These properties are essential when dendrimer structures are used in nanotechnology, drug delivery systems, and advanced material design. Dendrimers, especially polyphenylene dendrimers (PPDs), are highly branched macromolecules known for their robustness, shape persistence, and ability to encapsulate and release therapeutic agents in a controlled manner. This paper recognizes the patterns for extreme topologies associated with the M-S index on dendrimer graph molecules, analyzing its structural features that maximize or minimize this topological invariant. This study shows how graph theory and recurrence relations can be used to design efficient counting algorithms, benefiting both the pattern recognition area and practical applications in molecular design, drug delivery, and nanomaterials engineering.
This article presents a study on technological development, research, and current techniques in artificial intelligence and robotics using rule-based systems focused on medical applications in various medical areas. It analyzes various control techniques, image and signal analysis, and vision control as tools in medical applications by describing knowledge-based systems described by human experts or data that go beyond traditional Boolean logic. These are fundamental in artificial intelligence, especially in complex and high-risk areas such as medicine, for managing uncertainty and modeling expert knowledge. Applications ranging from aspects of preventive, non-invasive, and invasive medicine and healthcare will be reviewed, as well as the methods used in artificial intelligence, such as data analysis models, recognition, monitoring, and medical robotics, which improve the accuracy and safety of procedures. In order to observe the current trend in publications on type-1 and type-2 fuzzy logic controllers used in various medical fields, we have included a summary of the findings from the Medline database.
Pleural effusion is a frequent and clinically relevant finding on chest radiographs in acute and critical care. An automated detection pipeline was developed using transfer learning with DenseNet121 on a strictly balanced NIH ChestX-ray14 cohort (3,955 Pleural Effusion / 3,955 No Finding; 7,910 images) partitioned patient-wise into 70% training (5,933), 10% validation (665), and 20% test (1,312). Training used restrained augmentation (RandomZoom 0.1, RandomContrast 0.1); validation and test images were not augmented. After feature-extraction pretraining, targeted fine-tuning was applied to the upper DenseNet121 blocks. Predictions were obtained from softmax class scores and metrics were computed from the resulting confusion matrices. On the test set, results were accuracy 0.824, precision 0.877, recall 0.755, specificity 0.893, and F1 0.811 (TP 497, FN 161, FP 70, TN 584). On the validation set, results were accuracy 0.824, precision 0.853, recall 0.758, specificity 0.883, and F1 0.803 (TP 238, FN 76, FP 41, TN 310). Grad-CAM analyses highlighted saliency over costophrenic recesses, basal opacities, and meniscus-like contours in correctly classified positives. These findings indicate stable discrimination with an operating profile that favors specificity, establishing a transparent and reproducible baseline for pleural-effusion detection on chest radiographs.
Children with Autism Spectrum Disorder (ASD) benefit from personalized visual and cognitive stimulation. We present Deep Generative Visual Therapy (DGVT), an interactive system using Generative Adversarial Networks (GANs) to create tailored visual content for stimulating children with ASD. Our method features a custom GAN architecture trained with symbolic and concrete images to generate suitable stimuli for therapy targets such as attention enhancement, visual sequencing, and pattern matching. The system, built on TensorFlow and Keras, is accessible via Google Colab for real-time control and customization by therapists and educators. A series of cognitive games using generated images supports attention, visual discrimination, and memory. Initial assessments with therapists and pilot users showed positive engagement, indicating GAN-generated stimuli can complement traditional cognitive therapy for ASD. This effort connects generative deep learning with neurodevelopmental treatment to apply adversarial image synthesis in practical sensory and human-centered applications.
Hope speech, which is defined by statements of optimism and encouragement, has a positive impact on social media conversation and is critical in boosting individual well-being, particularly among users who are facing adversity, stress, worry, or illness. As a result, the automatic identification of hopeful content has arisen as an important research topic. However, natural language processing (NLP) systems continue to confront substantial hurdles in accurately detecting hope, which can range from grounded optimism to extreme wishfulness or even scathing sarcasm. To address these issues, we propose SarcHope, a framework for sarcasm detection in hope speech across English and Spanish. Our approach involves re-labeling the IbertLEF-2025 models. Evaluation on an independent test set shows that data, and S-mmBERT yields a top F1-score of 0.8327 on Spanish data. The results advance the field of natural language processing and provide a valuable baseline for subsequent studies.
An Optical Spatial filtering is a optomechanical system widely used in advanced optics and has broad applicability in diverse fields. This system is used to archive optimal beam quality in applications like high-power laser beams or optical interferometry removing high-frequency noise and improve beam profiles. The efficacy of the system relies on a correct alignment process, which is a human guided or hand made process. In particular, the alignment of the Airy disk and the pinhole. Despite decades of research into alignment process, advancements are focused in the hand made alignment process and the automatization remains as a significant challenge. Since automating the system depends on finding the exact center of the diffraction pattern, this work is dedicated to solving that specific problem. While superpixel segmentation algorithms achieved lower error rates, two of the evaluated algorithms exhibit reduced processing times compared to the latter. The algorithms applicability is aiming the real time execution in controllers running in low resource hardware. A hybrid approach is under discussion.
Accurate detection of COVID-19 remains a significant challenge, particularly due to the inherent limitations in conventional diagnostic methods such as RT-PCR testing. This study introduces two deep learning-based classification strategies using the COVIDGR chest X-ray data set. The first approach leverages the ResNeXt50 architecture, enhanced through training hyperparameter optimization via the Differential Evolution algorithm. The second strategy also employs DE to design custom CNN architectures based on the Multi-scale Feature Learning model, optimizing structural hyperparameters. Experimental results show that both methods outperform their non-optimized counterparts as well as several state-of-the-art approaches, achieving an overall performance of 90.32%. These findings highlight the potential of DE as a powerful tool for improving automated diagnosis of respiratory diseases.
