Development of HepatIA: A computed tomography annotation platform and database for artificial intelligence training in hepatocellular carcinoma detection at a Brazilian tertiary teaching hospital.

BACKGROUND: Hepatocellular carcinoma (HCC) is a prevalent tumor with high mortality rates. Computed tomography (CT) is crucial in the non-invasive diagnosis of HCC. Recent advancements in artificial intelligence (AI) have shown significant potential in medical imaging analysis. However, developing these AI algorithms is hindered by the scarcity of comprehensive, publicly available liver imaging datasets. OBJECTIVES: This study aims to detail the tools, data organization, and database structuring used in creating HepatIA, a medical imaging annotation platform and database at a Brazilian tertiary teaching hospital. HepatIA supports liver disease AI research at the institution. MATERIAL AND METHODS: The authors collected baseline characteristics and CT scans of 656 patients from 2008 to 2021. The database, designed using PostgreSQL and implemented with Django and Vue.js, includes 692 CT volumes from a four-phase abdominal CT protocol. Radiologists made segmentation annotations using the OHIF medical image viewer, incorporating MONAI Label for pre-annotation segmentation models. The annotation process included detailed descriptions of liver morphology and nodule characteristics. RESULTS: The HepatIA database currently includes healthy individuals and those with liver diseases such as HCC and cirrhosis. The database dashboard facilitates user interaction with intuitive plots and histograms. Key patient demographics include 64% males and an average age of 56.89 years. The database supports various filters for detailed searches, enhancing research capabilities. CONCLUSION: A comprehensive data structure was successfully created and integrated with the IT systems of a teaching hospital, enabling research on deep learning algorithms applied to abdominal CT scans for investigating hepatic lesions such as HCC.

A YOLO-V5 approach for the evaluation of normal fillings and overhanging fillings: an artificial intelligence study.

Dental fillings, frequently used in dentistry to address various dental tissue issues, may pose problems when not aligned with the anatomical contours and physiology of dental and periodontal tissues. Our study aims to detect the prevalence and distribution of normal and overhanging filling restorations using a deep CNN architecture trained through supervised learning, on panoramic radiography images. A total of 10480 fillings and 2491 overhanging fillings were labeled using CranioCatch software from 2473 and 1850 images, respectively. After the data obtaining phase, validation (80%), training 10%), and test-groups (10%) were formed from images for both labelling. The YOLOv5x architecture was used to develop the AI model. The model's performance was assessed through a confusion matrix and sensitivity, precision, and F1 score values of the model were calculated. For filling, sensitivity is 0.95, precision is 0.97, and F1 score is 0.96; for overhanging were determined to be 0.86, 0.89, and 0.87, respectively. The results demonstrate the capacity of the YOLOv5 algorithm to segment dental radiographs efficiently and accurately and demonstrate proficiency in detecting and distinguishing between normal and overhanging filling restorations.

Artificial intelligence in mammography: a systematic review of the external validation.

Objective: To conduct a systematic review of external validation studies on the use of different Artificial Intelligence algorithms in breast cancer screening with mammography. Data source: Our systematic review was conducted and reported following the PRISMA statement, using the PubMed, EMBASE, and Cochrane databases with the search terms "Artificial Intelligence," "Mammography," and their respective MeSH terms. We filtered publications from the past ten years (2014 - 2024) and in English. Study selection: A total of 1,878 articles were found in the databases used in the research. After removing duplicates (373) and excluding those that did not address our PICO question (1,475), 30 studies were included in this work. Data collection: The data from the studies were collected independently by five authors, and it was subsequently synthesized based on sample data, location, year, and their main results in terms of AUC, sensitivity, and specificity. Data synthesis: It was demonstrated that the Area Under the ROC Curve (AUC) and sensitivity were similar to those of radiologists when using independent Artificial Intelligence. When used in conjunction with radiologists, statistically higher accuracy in mammogram evaluation was reported compared to the assessment by radiologists alone. Conclusion: AI algorithms have emerged as a means to complement and enhance the performance and accuracy of radiologists. They also assist less experienced professionals in detecting possible lesions. Furthermore, this tool can be used to complement and improve the analyses conducted by medical professionals.

Comparative analysis of classical artificial intelligence and deep learning models for atrial fibrillation detection in electrocardiograms

BACKGROUND/INTRODUCTION: Early diagnosis of atrial fibrillation (AF) presents a challenging yet critical task for appropriate interventions aimed at reducing disease-related burden. In this context, strategies employing classical artificial intelligence (CAI) and deep learning (DL) have emerged as promising approaches to optimize cardiac disorder screening and detection. PURPOSE: This study aimed to compare a CAI model and a DL model for the detection of AF in patients undergoing electrocardiographic (ECG) examinations in tertiary healthcare centers. METHODS: Between December 2022 and November 2023, a total of 135,476 ECGs were performed, comprising 5,067 with AF and 130,409 without AF. The ECGs were analyzed using both artificial intelligence models. The obtained results were then compared to the gold standard (cardiologist's report). In the CAI model, signals were extracted from ECG images, analyzing five key parameters: cardiac rhythm, atrial depolarization, atrioventricular conduction, ventricular depolarization, and ventricular repolarization (figure 1A). These parameters were benchmarked against the standard values from the Brazilian Society of Cardiology guidelines for detecting cardiac anomalies. Conversely, the DL model utilized a one-dimensional ResNet-based Convolutional Neural Network (CNN). This model was trained using ADAM optimization and binary cross-entropy loss, enabling the learning of complex patterns in the data (figure 1B). RESULTS: The mean age was 54.6 years (71.9 years with AF and 53.9 without AF). In the AF population, 52.2% were male (46% were male in the overall sample). In the analysis conducted, the CAI model showed a sensitivity and specificity of 90% and 62%, respectively, while the DL model had 90% and 69%, respectively. ROC curves were generated for both models, demonstrating the superior performance of the DL model (figure 2A). CONCLUSIONS: Although the sensitivity remained similar between the models, the DL model distinguished itself with higher specificity. These results suggest that artificial intelligence, particularly the deep learning approach, holds promise as a supportive resource in AF diagnosis. However, further studies are needed to evaluate the models more thoroughly and determine their clinical applicability in a broader context.

A Real-Time Embedded System for Driver Drowsiness Detection Based on Visual Analysis of the Eyes and Mouth Using Convolutional Neural Network and Mouth Aspect Ratio.

Currently, the number of vehicles in circulation continues to increase steadily, leading to a parallel increase in vehicular accidents. Among the many causes of these accidents, human factors such as driver drowsiness play a fundamental role. In this context, one solution to address the challenge of drowsiness detection is to anticipate drowsiness by alerting drivers in a timely and effective manner. Thus, this paper presents a Convolutional Neural Network (CNN)-based approach for drowsiness detection by analyzing the eye region and Mouth Aspect Ratio (MAR) for yawning detection. As part of this approach, endpoint delineation is optimized for extraction of the region of interest (ROI) around the eyes. An NVIDIA Jetson Nano-based device and near-infrared (NIR) camera are used for real-time applications. A Driver Drowsiness Artificial Intelligence (DD-AI) architecture is proposed for the eye state detection procedure. In a performance analysis, the results of the proposed approach were compared with architectures based on InceptionV3, VGG16, and ResNet50V2. Night-Time Yawning-Microsleep-Eyeblink-Driver Distraction (NITYMED) was used for training, validation, and testing of the architectures. The proposed DD-AI network achieved an accuracy of 99.88% with the NITYMED test data, proving superior to the other networks. In the hardware implementation, tests were conducted in a real environment, resulting in 96.55% and 14 fps on average for the DD-AI network, thereby confirming its superior performance.

The artificial intelligence revolution...in unethical publishing: Will AI worsen our dysfunctional publishing system?

Scholarly publishing has been shaped by the pressure of a liquid economy to become an exercise in branding more than a vehicle for the advancement of science. The current revolution in artificial intelligence (AI) is poised to make matters worse. The new generation of large language models (LLMs) have shown impressive capabilities in text generation and are already being used to write papers, grants, peer review reports, code for analyses, and even perform literature reviews. Although these models can be used in positive ways, the metrics and pressures of academia, along with our dysfunctional publishing system, stimulate their indiscriminate and uncritical use to speed up research outputs. Thus, LLMs are likely to amplify the worst incentives of academia, greatly increasing the volume of scientific literature while diluting its quality. At present, no effective solutions are evident to overcome this grim scenario, and nothing short of a cultural revolution within academia will be needed to realign the practice of science with its traditional ideal of a rigorous search for truth.

Assistentes Virtuais Inteligentes e saúde mental: debates regulatórios no Brasil / Intelligent Virtual Assistants and mental health: regulatory debates in Brazil / Asistentes Virtuales Inteligentes y salud mental: debates regulatorios en Brasil

A inserção dos Assistentes Virtuais Inteligentes na vida cotidiana representa um marco na história da comunicação entre humanos e máquinas. Devido às suas características interativas, estes estão sendo cada vez mais apropriados e desenvolvidos para fins de cuidado, especialmente no âmbito da saúde mental. Este artigo visa compreender se e como o debate regulatório brasileiro oferece instrumentos para lidar com os desafios e as preocupações desses sistemas de Inteligência Artificial em relação à saúde mental. A partir de uma análise documental, mapeamos exemplos de aplicação dos Assistentes Virtuais Inteligentes em saúde mental, a fim de identificar riscos a direitos dos usuários e avaliar, na legislação brasileira vigente e em discussão, se há proteção suficiente para lidar com eles. Por meio de uma abordagem crítica, salientamos a insuficiência da legislação brasileira atual e a necessidade de ampliação do debate sobre como equilibrar possíveis riscos e benefícios dessas tecnologias.

The integration of Intelligent Virtual Assistants into everyday life marks a milestone in the history of human-machine communication. Due to their interactive characteristics, they are increasingly being appropriated and developed for caregiving purposes, especially in the field of mental health. This article aims to understand whether and how the Brazilian regulatory debate provides tools to address the challenges and concerns of these Artificial Intelligence systems concerning mental health. Through a document analysis, we map examples of Intelligent Virtual Assistants's applications to mental health to identify risks to users' rights and evaluate whether the current and the proposed Brazilian legislation offer sufficient protection to address these risks. Through a critical approach, we highlight the inadequacy of current Brazilian legislation and the need to expand the debate on how to balance the potential risks and benefits of these technologies.

La inserción de los Asistentes Virtuales Inteligentes en la vida cotidiana representa un hito en la historia de la comunicación entre humanos y máquinas. Debido a sus características interactivas, cada vez son más apropiados y desarrollados para fines de cuidado. Este artículo tiene como objetivo comprender si y cómo el debate regulatorio brasileño ofrece instrumentos para abordar los desafíos y preocupaciones de estos sistemas de Inteligencia Artificial en relación con la salud mental. A partir de un análisis documental, mapeamos ejemplos de la aplicación de los Asistentes Virtuales Inteligentes a la salud mental, con el fin de identificar riesgos para los derechos de los usuarios y evaluar, en la legislación brasileña vigente y en discusión, si hay protección suficiente para abordarlos. Destacamos la insuficiencia de la legislación brasileña actual y la necesidad de ampliar el debate sobre cómo equilibrar los posibles riesgos y beneficios de estas tecnologías.

Questões conjunturais sobre a regulação da IA / Conjunctural issues in AI regulation / Cuestiones coyunturales sobre la regulación de la IA

O texto parte da disputa em torno da regulação da inteligência artificial (IA) realmente existente, dividida entre dois polos, o da proteção da inovação e o da necessidade de mitigar riscos sociais e ambientais. Avança para apontar a colossal concentração de poder computacional e de dados nas mãos dos oligopólios digitais, as big techs. Levanta preocupações sobre o impacto dessa concentração na soberania digital e na capacidade do país em controlar suas infraestruturas e dados. Argumenta como a falta de transparência e explicabilidade nos sistemas de IA agrava os riscos de discriminação e exclusão social. Ainda, defende que a regulação da IA deve garantir direitos, salvaguardar a soberania e promover um ecossistema digital autônomo e inclusivo.

The text begins by addressing the debate surrounding the regulation of existing artificial intelligence (AI), divided between two poles: the protection of innovation and the need to mitigate social and environmental risks. It progresses to highlight the colossal concentration of computational power and data in the hands of digital oligopolies, the big tech companies. The text raises concerns about the impact of this concentration on digital sovereignty and the country's ability to control its infrastructures and data. It argues that the lack of transparency and explainability in AI systems exacerbates the risks of discrimination and social exclusion. The text advocates that AI regulation should ensure rights, safeguard sovereignty, and promote an autonomous and inclusive digital ecosystem.

El texto parte de la disputa en torno a la regulación de la inteligencia artificial (IA) existente, dividida entre dos polos: la protección de la innovación y la necesidad de mitigar riesgos sociales y ambientales. Avanza para señalar la colosal concentración de poder computacional y de datos en manos de los oligopolios digitales, las big tech. Plantea preocupaciones sobre el impacto de esta concentración en la soberanía digital y en la capacidad del país para controlar sus infraestructuras y datos. Argumenta que la falta de transparencia y explicabilidad en los sistemas de IA agrava los riesgos de discriminación y exclusión social. El texto defiende que la regulación de la IA debe garantizar derechos, salvaguardar la soberanía y promover un ecosistema digital autónomo e inclusivo.

Comparison of the refractive prediction errors of artificial intelligence formula with 3 conventional formulas and 11 combination methods in cataract surgery on eyes with short axial length.

PURPOSE: To compare the refractive prediction error of Hill-radial basis function 3.0 with those of 3 conventional formulas and 11 combination methods in eyes with short axial lengths. METHODS: The refractive prediction error was calculated using 4 formulas (Hoffer Q, SRK-T, Haigis, and Hill-RBF) and 11 combination methods (average of two or more methods). The absolute error was determined, and the proportion of eyes within 0.25-diopter (D) increments of absolute error was analyzed. Furthermore, the intraclass correlation coefficients of each method were computed to evaluate the agreement between target refractive error and postoperative spherical equivalent. RESULTS: This study included 87 eyes. Based on the refractive prediction error findings, Hoffer Q formula exhibited the highest myopic errors, followed by SRK-T, Hill-RBF, and Haigis. Among all the methods, the Haigis and Hill-RBF combination yielded a mean refractive prediction error closest to zero. The SRK-T and Hill-RBF combination showed the lowest mean absolute error, whereas the Hoffer Q, SRK-T, and Haigis combination had the lowest median absolute error. Hill-radial basis function exhibited the highest intraclass correlation coefficient, whereas SRK-T showed the lowest. Haigis and Hill-RBF, as well as the combination of both, demonstrated the lowest proportion of refractive surprises (absolute error >1.00 D). Among the individual formulas, Hill-RBF had the highest success rate (absolute error ≤0.50 D). Moreover, among all the methods, the SRK-T and Hill-RBF combination exhibited the highest success rate. CONCLUSIONS: Hill-radial basis function showed accuracy comparable to or surpassing that of conventional formulas in eyes with short axial lengths. The use and integration of various formulas in cataract surgery for eyes with short axial lengths may help reduce the incidence of refractive surprises.

Clinical value of artificial intelligence 3D echocardiography in evaluating left atrial volume and pulmonary vein structure in patients with atrial fibrillation.

OBJECTIVE: To explore the clinical value of 3D Echocardiography (3DE) in evaluating the changes of left atrial volume and pulmonary vein structure in patients with Atrial Fibrillation (AF). METHODS: Clinical data were collected from 54 AF patients. Left Atrial Anteroposterior Diameter (LADap), Left Atrial left and right Diameter (LADml), and Left Atrial upper and lower Diameter (LADsi) were measured; the maximum Left Atrial Volume (LAVmax), minimum Left Atrial Volume (LAVmin), left atrial presystolic volume (LAVpre), and Cross-Sectional Area (CSA) of each pulmonary vein were analyzed. Passive Ejection Fraction (LAPEF) was calculated. The differences in left atrial volume and pulmonary vein structure between patients with AF and healthy people were compared, and the correlation between the indexes was analyzed. The diagnostic value of the above indicators for AF patients was analyzed. RESULTS: LADap, LADml, LADsi, LAVmax, LAVmin, LAVpre, LAPEF, LSPV CSA, LIPV CSA, RSPV CSA, and RIPV CSA of AF patients were significantly higher. There was a significant positive correlation between left atrial diameter and pulmonary vein structure. There was a significant positive correlation between left atrial volume and pulmonary vein structure. There was a negative correlation between LAPEF and pulmonary vein structure. LADap, LADml, LADsi, LAVmax, LAVmin, LAVpre, LAPEF, LSPV CSA, LIPV CSA, RSPV CSA, and RIPV CSA had a diagnostic value for AF patients. CONCLUSION: 3DE is applicable for evaluating left atrial volume and pulmonary vein structure in patients with AF.

Detecting pediatric appendicular fractures using artificial intelligence.

OBJECTIVE: The primary objective was to assess the diagnostic accuracy of a deep learning-based artificial intelligence model for the detection of acute appendicular fractures in pediatric patients presenting with a recent history of trauma to the emergency department. The secondary goal was to examine the effect of assistive support on the emergency doctor's ability to detect fractures. METHODS: The dataset was 5,150 radiographs of which 850 showed fractures, while 4,300 radiographs did not show any fractures. The process utilized 4,532 (88%) radiographs, inclusive of both fractured and non-fractured radiographs, in the training phase. Subsequently, 412 (8%) radiographs were appraised during validation, and 206 (4%) were set apart for the testing phase. With and without artificial intelligence assistance, the emergency doctor reviewed another set of 2,000 radiographs (400 fractures and 600 non-fractures each) for labeling in the second test. RESULTS: The artificial intelligence model showed a mean average precision 50 of 89%, a specificity of 92%, a sensitivity of 90%, and an F1 score of 90%. The confusion matrix revealed that the model trained with artificial intelligence achieved accuracies of 93 and 95% in detecting fractures, respectively. Artificial intelligence assistance improved the reading sensitivity from 93.7% (without assistance) to 97.0% (with assistance) and the reading accuracy from 88% (without assistance) to 94.9% (with assistance). CONCLUSION: A deep learning-based artificial intelligence model has proven to be highly effective in detecting fractures in pediatric patients, enhancing the diagnostic capabilities of emergency doctors through assistive support.

Investigating artificial intelligence models for predicting joint pain from serum biochemistry.

OBJECTIVE: The study used machine learning models to predict the clinical outcome with various attributes or when the models chose features based on their algorithms. METHODS: Patients who presented to an orthopedic outpatient department with joint swelling or myalgia were included in the study. A proforma collected clinical information on age, gender, uric acid, C-reactive protein, and complete blood count/liver function test/renal function test parameters. Machine learning decision models (Random Forest and Gradient Boosted) were evaluated with the selected features/attributes. To categorize input data into outputs of indications of joint discomfort, multilayer perceptron and radial basis function-neural networks were used. RESULTS: The random forest decision model outperformed with 97% accuracy and minimum errors to anticipate joint pain from input attributes. For predicted classifications, the multilayer perceptron fared better with an accuracy of 98% as compared to the radial basis function. Multilayer perceptron achieved the following normalized relevance: 100% (uric acid), 10.3% (creatinine), 9.8% (AST), 5.4% (lymphocytes), and 5% (C-reactive protein) for having joint pain. Uric acid has the highest normalized relevance for predicting joint pain. CONCLUSION: The earliest artificial intelligence-based detection of joint pain will aid in the prevention of more serious orthopedic complications.

Comparison of Faster R-CNN, YOLO, and SSD for Third Molar Angle Detection in Dental Panoramic X-rays.

The use of artificial intelligence algorithms (AI) has gained importance for dental applications in recent years. Analyzing AI information from different sensor data such as images or panoramic radiographs (panoramic X-rays) can help to improve medical decisions and achieve early diagnosis of different dental pathologies. In particular, the use of deep learning (DL) techniques based on convolutional neural networks (CNNs) has obtained promising results in dental applications based on images, in which approaches based on classification, detection, and segmentation are being studied with growing interest. However, there are still several challenges to be tackled, such as the data quality and quantity, the variability among categories, and the analysis of the possible bias and variance associated with each dataset distribution. This study aims to compare the performance of three deep learning object detection models-Faster R-CNN, YOLO V2, and SSD-using different ResNet architectures (ResNet-18, ResNet-50, and ResNet-101) as feature extractors for detecting and classifying third molar angles in panoramic X-rays according to Winter's classification criterion. Each object detection architecture was trained, calibrated, validated, and tested with three different feature extraction CNNs which are ResNet-18, ResNet-50, and ResNet-101, which were the networks that best fit our dataset distribution. Based on such detection networks, we detect four different categories of angles in third molars using panoramic X-rays by using Winter's classification criterion. This criterion characterizes the third molar's position relative to the second molar's longitudinal axis. The detected categories for the third molars are distoangular, vertical, mesioangular, and horizontal. For training, we used a total of 644 panoramic X-rays. The results obtained in the testing dataset reached up to 99% mean average accuracy performance, demonstrating the YOLOV2 obtained higher effectiveness in solving the third molar angle detection problem. These results demonstrate that the use of CNNs for object detection in panoramic radiographs represents a promising solution in dental applications.

Salivary detection of Chikungunya virus infection using a portable and sustainable biophotonic platform coupled with artificial intelligence algorithms.

The current detection method for Chikungunya Virus (CHIKV) involves an invasive and costly molecular biology procedure as the gold standard diagnostic method. Consequently, the search for a non-invasive, more cost-effective, reagent-free, and sustainable method for the detection of CHIKV infection is imperative for public health. The portable Fourier-transform infrared coupled with Attenuated Total Reflection (ATR-FTIR) platform was applied to discriminate systemic diseases using saliva, however, the salivary diagnostic application in viral diseases is less explored. The study aimed to identify unique vibrational modes of salivary infrared profiles to detect CHIKV infection using chemometrics and artificial intelligence algorithms. Thus, we intradermally challenged interferon-gamma gene knockout C57/BL6 mice with CHIKV (20 µl, 1 X 105 PFU/ml, n = 6) or vehicle (20 µl, n = 7). Saliva and serum samples were collected on day 3 (due to the peak of viremia). CHIKV infection was confirmed by Real-time PCR in the serum of CHIKV-infected mice. The best pattern classification showed a sensitivity of 83%, specificity of 86%, and accuracy of 85% using support vector machine (SVM) algorithms. Our results suggest that the salivary ATR-FTIR platform can discriminate CHIKV infection with the potential to be applied as a non-invasive, sustainable, and cost-effective detection tool for this emerging disease.