Artificial Intelligence Prediction Research Articles

Machine Learning-Based Suicide Risk Prediction Model for Suicidal Trajectory on Social Media Following Suicidal Mentions: Independent Algorithm Validation.

Previous efforts to apply machine learning-based natural language processing to longitudinally collected social media data have shown promise in predicting suicide risk. Our primary objective was to externally validate our previous machine learning algorithm, the Suicide Artificial Intelligence Prediction Heuristic (SAIPH), against external survey data in 2 independent cohorts. A second objective was to evaluate the efficacy of SAIPH as an indicator of changing suicidal ideation (SI) over time. The tertiary objective was to use SAIPH to evaluate factors important for improving or worsening suicidal trajectory on social media following suicidal mention. Twitter (subsequently rebranded as X) timeline data from a student survey cohort and COVID-19 survey cohort were scored using SAIPH and compared to SI questions on the Beck Depression Inventory and the Self-Report version of the Quick Inventory of Depressive Symptomatology in 159 and 307 individuals, respectively. SAIPH was used to evaluate changing SI trajectory following suicidal mentions in 2 cohorts collected using the Twitter application programming interface. An interaction of the mean SAIPH score derived from 12 days of Twitter data before survey completion and the average number of posts per day was associated with quantitative SI metrics in each cohort (student survey cohort interaction β=.038, SD 0.014; F4,94=3.3, P=.01; and COVID-19 survey cohort interaction β=.0035, SD 0.0016; F4,493=2.9, P=.03). The slope of average daily SAIPH scores was associated with the change in SI scores within longitudinally followed individuals when evaluating periods of 2 weeks or less (ρ=0.27, P=.04). Using SAIPH as an indicator of changing SI, we evaluated SI trajectory in 2 cohorts with suicidal mentions, which identified that those with responses within 72 hours exhibit a significant negative association of the SAIPH score with time in the 3 weeks following suicidal mention (ρ=-0.52, P=.02). Taken together, our results not only validate the association of SAIPH with perceived stress, SI, and changing SI over time but also generate novel methods to evaluate the effects of social media interactions on changing suicidal trajectory.

Artificial intelligence-assisted fitting method using corneal topography outcomes enhances success rate in orthokeratology lens fitting.

Based on ideal outcomes of corneal topography following orthokeratology (OK), an innovative machine learning algorithm for corneal refractive therapy (CRT) was developed to investigate the precision of artificial intelligence (AI)-assisted OK lens fitting. A total of 797 eyes that had been fitted with CRT lenses and demonstrated good lens centration with plus power ring intact in their topography were retrospectively included. A comprehensive AI model included spherical refraction, keratometry readings, eccentricity, corneal astigmatism, horizontal visible iris diameter, inferior-superior index, surface asymmetry index, surface regularity index and 8-mm chordal corneal height difference. A simplified AI model omitted the latter four parameters. Correlation and disparity in predicted lens parameters between the AI prediction and manufacturer's conventional lens fitting method were compared. There was overall no significant difference between AI predicted parameters and the final ordered parameters (p>0.05). The horizontal return zone depth (RZD1, p=0.022) and vertical return zone depth (RZD2, p<0.001) values suggested by the conventional method were significantly lower, while the horizontal landing zone angle (LZA1) was significantly larger (p=0.002) than those of the final ordered lens. The AI predicted parameters were significantly correlated to those of the final ordered lens (p<0.01), with the correlation coefficients of base curve radius (BCR), RZD1, RZD2, LZA1, vertical LZA (LZA2) and total lens diameter (TD) being 0.958, 0.708, 0.773, 0.697, 0.654 and 0.730, respectively, for the comprehensive AI model. The correlation coefficients were higher in RZD2, LZA1 and TD with the AI model as compared to conventional method. Compared with the conventional method, AI predicted lens parameters exhibit less disparity and improved accuracy, with a potential to facilitate more efficient and precise CRT OK lens fitting.

Guest Editorial: Patience in AI Development Is a Virtue: Why 99% Correct Is 100% Wrong

In heavy-asset industries such as oil and gas, precision is crucial. The saying "99% accurate is 100% wrong" reflects this reality. Despite the excitement of new technology, even minor errors can have significant consequences. For example, an artificial intelligence (AI) system inaccurately predicting a critical machinery component's lifespan could lead to unexpected failures, causing costly downtime and safety hazards. Kongsberg Digital, an early adopter of Microsoft's large language models (LLMs), has witnessed AI's transformative power firsthand. Over the past 2 years, these LLMs have driven a resurgence in AI interest. Generative AI's growth presents unprecedented opportunities for the heavy-asset industry, transforming interactions with complex systems. Implementing AI can optimize maintenance schedules, predict failures before they happen, and streamline operations. The rise of generative AI has also spotlighted the more traditional areas of analytics, classification, prediction, and physics-based simulations. This renewed interest has led the oil and gas industry to look for ways to utilize AI to enhance operational efficiency, reduce costs, and improve safety standards. However, AI inaccuracies or “hallucinations” are deal-breakers. AI-generated misinformation can mislead decision-makers, potentially resulting in disastrous outcomes. In some cases, using AI is unnecessary and adds little value. Heavy-asset industries prioritize safety and have historically been conservative in adopting new technologies. While post-ChatGPT developments are significant, the industry lags due to its zero tolerance for failure. This caution is justified; in environments where lives and substantial investments are at stake, even minor errors can be catastrophic. Moreover, precision in AI ensures not only safety and efficiency but also drives sustainability. Accurate AI predictions minimize waste and reduce energy consumption. Inaccurate decisions can lead to excessive energy use and waste, counteracting sustainability goals. AI can lower carbon emissions and reduce the environmental footprint by optimizing energy usage and ensuring regulatory compliance. Building trust in AI requires responsibility in development and implementation, ensuring security without compromise. We must be transparent about data lineage—a principle that guides our transformative journey. For AI to fully integrate into heavy-asset industries, it must be accurate but also secure and trustworthy. This transparency builds confidence among stakeholders, from engineers to executives.

Application of artificial intelligence in prediction of future land use / land cover for cities in transition: a comparative analysis

Application of artificial intelligence in prediction of future land use / land cover for cities in transition: a comparative analysis

Spatial-temporal characteristics and the importance of environmental factors in relation to algal blooms in coastal seas

In the prediction of algal blooms with the artificial intelligence (AI) technology, the proper selection of the indices is crucial to the quality of the results. Although eutrophication and climate change have been identified as the main contributors to algal blooms, limited knowledge of the relationships among environmental factors hinders the accurate and correct AI prediction. To explore the importance of environmental factors in the occurrence of algal blooms in coastal seas and facilitate the selection of inputs for the prediction model, this study examines observed water quality data and analyzes their relationship using machine learning methods. It was found that nutrients, water temperature, and salinity play important roles in controlling algal blooms in coastal seas, and their influence varies depending on locations. The high-risk cases of algal bloom mainly occur in environments with medium salinity (22–28 psu), particularly in open estuaries with high nutrient levels. According to the feature importance analysis, water temperature presented the most significant impact on algal bloom risk compared to other environmental factors. The variations of water temperature and Chl-a concentration are synchronous when the water temperature is below 32 °C. Additionally, the relationships between inorganic nutrients (NO3−-NO2− and Ortho-P) and Chl-a exhibit asynchronous across multi-temporal scales. The uptake of inorganic nutrients occurs in large quantities within the 24 h prior to the boom in Chl-a concentration. This demonstrates that variations in concentration of inorganic nutrients can be considered as crucial factors in algal bloom predictions, and their significant decrease can be served as an important signal of an impending algal bloom. This study enhances our understanding of the mechanisms behind algal blooms and facilitates their management and prediction strategies in coastal seas.

Development of the design and synthesis of metal-organic frameworks (MOFs) - from large scale attempts, functional oriented modifications, to artificial intelligence (AI) predictions.

Owing to the exceptional porous properties of metal-organic frameworks (MOFs), there has recently been a surge of interest, evidenced by a plethora of research into their design, synthesis, properties, and applications. This expanding research landscape has driven significant advancements in the precise regulation of MOF design and synthesis. Initially dominated by large-scale synthesis approaches, this field has evolved towards more targeted functional modifications. Recently, the integration of computational science, particularly through artificial intelligence predictions, has ushered in a new era of innovation, enabling more precise and efficient MOF design and synthesis methodologies. The objective of this review is to provide readers with an extensive overview of the development process of MOF design and synthesis, and to present visions for future developments.

A self-driven ESN-DSS approach for effective COVID-19 time series prediction and modelling.

Since the outbreak of the COVID-19 epidemic, it has posed a great crisis to the health and economy of the world. The objective is to provide a simple deep-learning approach for predicting, modelling, and evaluating the time evolutions of the COVID-19 epidemic. The Dove Swarm Search (DSS) algorithm is integrated with the echo state network (ESN) to optimize the weight. The ESN-DSS model is constructed to predict the evolution of the COVID-19 time series. Specifically, the self-driven ESN-DSS is created to form a closed feedback loop by replacing the input with the output. The prediction results, which involve COVID-19 temporal evolutions of multiple countries worldwide, indicate the excellent prediction performances of our model compared with several artificial intelligence prediction methods from the literature (e.g., recurrent neural network, long short-term memory, gated recurrent units, variational auto encoder) at the same time scale. Moreover, the model parameters of the self-driven ESN-DSS are determined which acts as a significant impact on the prediction performance. As a result, the network parameters are adjusted to improve the prediction accuracy. The prediction results can be used as proposals to help governments and medical institutions formulate pertinent precautionary measures to prevent further spread. In addition, this study is not only limited to COVID-19 time series forecasting but also applicable to other nonlinear time series prediction problems.

AI predictive models and advancements in microdissection testicular sperm extraction for non-obstructive azoospermia: a systematic scoping review

Abstract STUDY QUESTION How accurately can artificial intelligence (AI) models predict sperm retrieval in non-obstructive azoospermia (NOA) patients undergoing micro-testicular sperm extraction (m-TESE) surgery? SUMMARY ANSWER AI predictive models hold significant promise in predicting successful sperm retrieval in NOA patients undergoing m-TESE, although limitations regarding variability of study designs, small sample sizes and a lack of validation studies restrict the overall generalizability of studies in this area. WHAT IS KNOWN ALREADY Previous studies have explored various predictors of successful sperm retrieval in m-TESE, including clinical and hormonal factors. However, no consistent predictive model has yet been established. STUDY DESIGN, SIZE, DURATION A comprehensive literature search was conducted following PRISMA-ScR guidelines, covering PubMed and Scopus databases from 2013 to May 15th, 2024. Relevant English-language studies were identified using Medical Subject Headings (MeSH) terms. We also used PubMed’s “similar articles” and “cited by” features for thorough bibliographic screening to ensure comprehensive coverage of relevant literature. PARTICIPANTS/MATERIALS, SETTING, METHODS The review included studies on patients with NOA where AI-based models were used for predicting m-TESE outcomes, by incorporating clinical data, hormonal levels, histopathological evaluations and genetic parameters. Various machine learning and deep learning techniques, including logistic regression, were employed. The Prediction Model Risk of Bias Assessment Tool (PROBAST) evaluated the bias in the studies, and their quality was assessed using the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) guidelines, ensuring robust reporting standards and methodological rigour. MAIN RESULTS AND THE ROLE OF CHANCE Out of 427 screened articles, 45 met the inclusion criteria, with most using logistic regression and machine learning to predict m-TESE outcomes. AI-based models demonstrated strong potential by integrating clinical, hormonal and biological factors. However, limitations of the studies included small sample sizes, legal barriers and challenges in generalizability and validation. While some studies featured larger, multi-centre designs, many were constrained by sample size. Most studies had a low risk of bias in participant selection and outcome determination, and two-thirds were rated as low risk for predictor assessment, but the analysis methods varied. LIMITATIONS, REASONS FOR CAUTION The limitations of this review include the heterogeneity of the included research, potential publication bias and reliance on only two databases (PUBMED and SCOPUS), which may limit the scope of findings. Additionally, the absence of a meta-analysis prevents quantitative assessment of the consistency of models. Despite this, the review offers valuable insights into AI predictive models for m-TESE in NOA. WIDER IMPLICATIONS OF THE FINDINGS The review highlights the potential of advanced AI techniques in predicting successful sperm retrieval for NOA patients undergoing m-TESE. By integrating clinical, hormonal, histopathological and genetic factors, AI models can enhance decision-making and improve patient outcomes, reducing the number of unsuccessful procedures. However, to further enhance the precision and reliability of AI predictions in reproductive medicine, future studies should address current limitations by incorporating larger sample sizes and conducting prospective validation trials. This continued research and development is crucial for strengthening the applicability of AI models and ensuring broader clinical adoption. STUDY FUNDING/COMPETING INTEREST(S) The authors would like to acknowledge Mashhad University of Medical Sciences, Mashhad, Iran, for financial support (Grant ID : 4020802). The authors declare no competing interests. REGISTRATION NUMBER N/A

Artificial Intelligence Prediction Model of Occurrence of Cerebral Vasospasms Based on Machine Learning.

Symptomatic cerebral vasospasms are deleterious complication of the rupture of a cerebral aneurysm and potentially lethal. The existing scales used to classify the initial presentation of a subarachnoid hemorrhage (SAH) offer a blink of the outcome and the possibility of occurrence of symptomatic cerebral vasospasms. Altogether, neither are they sufficient to predict outcome or occurrence of events reliably nor do they offer a united front. This study tests the common grading scales and factors that otherwise affect the outcome, in an artificial intelligence (AI) based algorithm to create a reliable prediction model for the occurrence of cerebral vasospasms. Applying the R environment, an easy-to-operate command line was programmed to prognosticate the occurrence of vasospasms. Eighty-seven patients with aneurysmal SAH during a 24-month period of time were included for study purposes. The holdout and cross-validation methods were used to evaluate the algorithm (65 patients constituted the validation set and 22 patients constituted the test set). The Support Vector Machines (ksvm) classification method provided a high accuracy. The medical dataset included demographic data, the Hunt and Hess scale (H&H), Fisher grade, Barrow Neurological Institute (BNI) scale, length of intervention for aneurysmal repair, etc. RESULTS: Our prediction model based on the AI algorithm demonstrated an accuracy of 61 to 86% for the event of symptomatic vasospasms. For subgroup analysis, 28.8% (n = 13) patients in the surgical cohort developed symptomatic vasospasm. Of these, 50% (n = 7) were admitted with Fisher scale grade 4, 37.5% (n = 5) with H&H 5, and 28.5% (n = 4) with BNI 5. In the endovascular cohort, vasospasms occurred in 31.8% (n = 14) patients. Of these, 69% (n = 9) patients were admitted with Fisher grade 4, 23% (n = 3) patients with H&H 5, and 7% (n = 1) patients with BNI 5. From our data, we may believe that the algorithm presented can help in identifying patients with SAH who are at "high" or "low" risk of developing symptomatic vasospasms. This risk balancing might further allow the treating physician to go for an earlier intervention trying to prevent permanent sequelae. Certainly, accuracy will improve with a higher caseload and more statistical coefficients.

A Quantitative Comparison Between Human and Artificial Intelligence in the Detection of Focal Cortical Dysplasia.

Artificial intelligence (AI) is thought to improve lesion detection. However, a lack of knowledge about human performance prevents a comparative evaluation of AI and an accurate assessment of its impact on clinical decision-making. The objective of this work is to quantitatively evaluate the ability of humans to detect focal cortical dysplasia (FCD), compare it to state-of-the-art AI, and determine how it may aid diagnostics. We prospectively recorded the performance of readers in detecting FCDs using single points and 3-dimensional bounding boxes. We acquired predictions of 3 AI models for the same dataset and compared these to readers. Finally, we analyzed pairwise combinations of readers and models. Twenty-eight readers, including 20 nonexpert and 5 expert physicians, reviewed 180 cases: 146 subjects with FCD (median age: 25, interquartile range: 18) and 34 healthy control subjects (median age: 43, interquartile range: 19). Nonexpert readers detected 47% (95% confidence interval [CI]: 46, 49) of FCDs, whereas experts detected 68% (95% CI: 65, 71). The 3 AI models detected 32%, 51%, and 72% of FCDs, respectively. The latter, however, also predicted more than 13 false-positive clusters per subject on average. Human performance was improved in the presence of a transmantle sign ( P < 0.001) and cortical thickening ( P < 0.001). In contrast, AI models were sensitive to abnormal gyration ( P < 0.01) or gray-white matter blurring ( P < 0.01). Compared with single experts, expert-expert pairs detected 13% (95% CI: 9, 18) more FCDs ( P < 0.001). All AI models increased expert detection rates by up to 19% (95% CI: 15, 24) ( P < 0.001). Nonexpert+AI pairs could still outperform single experts by up to 13% (95% CI: 10, 17). This study pioneers the comparative evaluation of humans and AI for FCD lesion detection. It shows that AI and human predictions differ, especially for certain MRI features of FCD, and, thus, how AI may complement the diagnostic workup.

Abstract 4142793: Development of an Angiography-Based Multimodal AI Model for Predicting the risk of In-Stent Restenosis

Background: Despite advances in Second-generation drug-eluting stents (DES), 5-10% of patients still experience in-stent restenosis (ISR) after percutaneous coronary intervention (PCI), which generates significant financial burden and elevates the risk of acute coronary syndrome (ACS) and rehospitalization. Thus, early identification of patients at high risk for ISR is crucial for guiding clinical stratification and early intervention. Aims: To develop and validate a multimodal artificial intelligence (AI) model based on coronary angiography images for predicting ISR risk in patients post-DES implantation. Methods: To establish an accurate predictive model, our approach begins with the pre-training on 100,000 angiographic images to enhance the model’s capability in recognizing image features. Subsequently, we employ the DenseNet architecture as the primary deep learning model, incorporating angiographic images from 2,000 cases of DES-treated de novo lesions—1,000 from patients who did not experience ISR within two years and 1,000 from those who did. A multivariate logistic regression analysis, including radiomic features, clinical baselines, and functional information, constructs the predictive model. Additionally, a separate prospective cohort of 300 cases was assembled for validation to simulate real-world application and to verify the model's reliability and accuracy. Results: Our study successfully developed an AI prediction model for ISR, utilizing a large cohort of coronary angiography images, which effectively predicts ISR with high accuracy. Leveraging the DenseNet architecture and finely tuned machine learning algorithms, the model achieved a sensitivity and specificity of 90% in the validation cohort. The ROC curve from the test phase demonstrated an AUC above 0.90, underscoring the model’s exceptional diagnostic capabilities. Furthermore, the implementation of this model in a prospective cohort confirmed its reliability and practical utility in real-world clinical settings. Conclusions: This study introduces the first multimodal AI model using angiographic imaging to predict ISR. By demonstrating high diagnostic accuracy and reliability in real-world settings, this model serves as an essential tool for early ISR detection and intervention, ultimately helping to reduce the incidence of major adverse cardiac events (MACEs) and mortality.

Abstract 4114276: Identification of the heterogeneity in the risk of heart failure in those with left bundle branch block pattern using artificial intelligence

Introduction: The association between left bundle branch block (LBBB) and heart failure is well-established. However, the lack of a practical risk assessment tool for patients with LBBB poses a clinical challenge to the management despite the heterogeneity of the LBBB population. Hypothesis: Artificial intelligence (AI) can extract features to stratify the risk of heart failure (HF) within the heterogeneity of LBBB. Methods: All 12-lead ECGs diagnosed with LBBB recorded from 1/1/2015 to 12/31/2019 were identified in our institution. ECG data were analyzed as 2D matrices with a shape of 12×2500. The Uniform Manifold Approximation and Projection (UMAP) was used to visualize the heterogeneity of LBBB ECG. Additionally, a 2-dimensional convolutional neural network model was trained to detect LBBB ECGs associated with past HF diagnosis. The model was then applied to ECGs from an external cohort of patients with LBBB but without a history of HF. Cumulative incidences of HF admission were compared by stratifying according to tertiles of the model output (low, intermediate and high AI score groups) Results: We identified 15,124 LBBB ECGs in our institution (training dataset) and 2,463 individuals with LBBB ECGs in the external cohort (external test dataset). The UMAP projection of the ECGs revealed distinct heterogeneity in the data ( Fig A ). This heterogeneity was not fully explained by patient demographics, ECG features, or institutions. When the external cohort was divided into 3 groups according to the tertile of AI prediction, patients with high AI scores had a higher risk of HF admission (high vs low AI score group: hazard ratio (HR), 2.10; 95% confidence interval (CI), 1.66-2.65; intermediate vs low AI score group: HR, 1.39; 95%CI, 1.08-1.78: log-rank p &lt;0.0001; Fig 1B ). This finding was unchanged after adjusting for multiple clinical characteristics and ECG parameters, including QRS duration. Conclusion: We observed significant heterogeneity in ECGs presenting LBBB, and the AI model excellently stratified the risk of heart failure admission in patients with LBBB from a single recording of 12-lead ECG, suggesting great potential for clinical utility in managing patients with LBBB.

Correlation between oxygenation function and laboratory indicators in COVID-19 patients based on non-enhanced chest CT images and construction of an artificial intelligence prediction model.

By extracting early chest CT radiomic features of COVID-19 patients, we explored their correlation with laboratory indicators and oxygenation index (PaO2/FiO2), thereby developed an Artificial Intelligence (AI) model based on radiomic features to predict the deterioration of oxygenation function in COVID-19 patients. This retrospective study included 384 patients with COVID-19, whose baseline information, laboratory indicators, oxygenation-related parameters, and non-enhanced chest CT images were collected. Utilizing the PaO2/FiO2 stratification proposed by the Berlin criteria, patients were divided into 4 groups, and differences in laboratory indicators among these groups were compared. Radiomic features were extracted, and their correlations with laboratory indicators and the PaO2/FiO2 were analyzed, respectively. Finally, an AI model was developed using the PaO2/FiO2 threshold of less than 200 mmHg as the label, and the model's performance was assessed using the area under the receiver operating characteristic curve (AUC), sensitivity and specificity. Group datas comparison was analyzed using SPSS software, and radiomic features were extracted using Python-based Pyradiomics. There were no statistically significant differences in baseline characteristics among the groups. Radiomic features showed differences in all 4 groups, while the differences in laboratory indicators were inconsistent, with some PaO2/FiO2 groups showed differences and others not. Regardless of whether laboratory indicators demonstrated differences across different PaO2/FiO2 groups, they could all be captured by radiomic features. Consequently, we chose radiomic features as variables to establish an AI model based on chest CT radiomic features. On the training set, the model achieved an AUC of 0.8137 (95% CI [0.7631-0.8612]), accuracy of 0.7249, sensitivity of 0.6626 and specificity of 0.8208. On the validation set, the model achieved an AUC of 0.8273 (95% CI [0.7475-0.9005]), accuracy of 0.7739, sensitivity of 0.7429 and specificity of 0.8222. This study found that the early chest CT radiomic features of COVID-19 patients are strongly associated not only with early laboratory indicators but also with the lowest PaO2/FiO2. Consequently, we developed an AI model based on CT radiomic features to predict deterioration in oxygenation function, which can provide a reliable basis for further clinical management and treatment.

Bridging Human and Machine Intelligence: Reverse-Engineering Radiologist Intentions for Clinical Trust and Adoption

In the rapidly evolving landscape of medical imaging, the integration of artificial intelligence (AI) with clinical expertise offers unprecedented opportunities to enhance diagnostic precision and accuracy. Yet, the "black box" nature of AI models often limits their integration into clinical practice, where transparency and interpretability are important. This paper presents a novel system leveraging the Large Multimodal Model (LMM) to bridge the gap between AI predictions and the cognitive processes of radiologists. This system consists of two core modules, Temporally Grounded Intention Detection (TGID) and Region Extraction (RE). The TGID module predicts the radiologist's intentions by analyzing eye gaze fixation heatmap videos and corresponding radiology reports. Additionally, the RE module extracts regions of interest that align with these intentions, mirroring the radiologist’s diagnostic focus. This approach introduces a new task, radiologist intention detection, and is the first application of Dense Video Captioning (DVC) in the medical domain. By making AI systems more interpretable and aligned with radiologist’s cognitive processes, this proposed system aims to enhance trust, improve diagnostic accuracy, and support medical education. Additionally, it holds the potential for automated error correction, guiding junior radiologists, and fostering more effective training and feedback mechanisms. This work sets a precedent for future research in AI-driven healthcare, offering a pathway towards transparent, trustworthy, and human-centered AI systems. We evaluated this model using NLG(Natural Language Generation), time-related, and vision-based metrics, demonstrating superior performance in generating temporally grounded intentions on REFLACX and EGD-CXR datasets. This model also demonstrated strong predictive accuracy in overlap scores for medical abnormalities and effective region extraction with high IoU(Intersection over Union), especially in complex cases like cardiomegaly and edema. These results highlight the system's potential to enhance diagnostic accuracy and support continuous learning in radiology. We are also releasing the source code for our project, available here.

Artificial intelligence prediction of the mechanical properties of banana peel-ash and bagasse blended geopolymer concrete

This research explores the application of Artificial Intelligence (AI) techniques to assess the mechanical properties of geopolymer concrete made from a blend of Banana Peel-Ash (BPA) and Sugarcane Bagasse Ash (SCBA), using a sodium silicate (Na2SiO3) to sodium hydroxide (NaOH) ratio ranging from 1.5 to 3. Utilizing three AI methodologies—Artificial Neural Networks (ANN), Adaptive Neuro-Fuzzy Inference System (ANFIS), and Gene Expression Programming (GEP)—the study aims to enhance prediction accuracy for the mechanical properties of geopolymer concrete based on 104 datasets. By optimizing mix designs through varying proportions of BPA and SCBA, alkaline activator molarity, and aggregate-to-binder ratios, the research identified combinations that significantly enhance mechanical properties, demonstrating notable international relevance as it contributes to global efforts in sustainable construction by effectively utilizing industrial by-products. The experimental results demonstrated that increasing the molarity of the alkaline activator from 4 to 10 M significantly enhanced both the compressive and flexural strengths of the geopolymer concrete. Specifically, a mixture containing 52.5% SCBA and 47.5% BPA at a 10 M molarity achieved a maximum compressive strength of 33.17 MPa after 20 h of curing. In contrast, a mixture composed of 95% SCBA and 5% BPA at a 4 M molarity exhibited a substantially lower compressive strength of only 21.27 MPa. Additionally, the highest recorded flexural strength of 9.95 MPa (77.25% SCBA and 22.5 BPA) was observed at the 10 M molarity, while the flexural strength at 4 M was lowest, at 4.12 MPa (95% SCBA and 5% BPA). Microstructural analysis through Scanning Electron Microscopy with Energy-Dispersive X-ray Spectroscopy (ED-SEM) revealed insights into the pore structure and elemental composition of the concrete, while Thermogravimetric Analysis (TGA) provided data on the material’s thermal stability and decomposition characteristics. Performance analysis of the AI models showed that the ANN model had an average MSE of 1.338, RMSE of 1.157, MAE of 3.104, and R2 of 0.989, while the ANFIS model outperformed with an MSE of 0.345, RMSE of 0.587, MAE of 1.409, and R2 of 0.998. The GEP model demonstrated an MSE of 1.233, RMSE of 1.110, MAE of 1.828, and R2 of 0.992, confirming that ANFIS is the most accurate model for predicting the mechanical and rheological properties of geopolymer concrete. This study highlights the potential of integrating AI with experimental data to optimize the formulation and performance of geopolymer concrete, advancing sustainable construction practices by effectively utilizing industrial by-products.

Efficiency or Equity? How Public Values Shape Bureaucrats' Willingness to Use Artificial Intelligence to Reduce Administrative Burdens

Generative and predictive Artificial Intelligence (AI) has the potential to revolutionize many features of policy implementation, but only if bureaucrats understand and decide to utilize AI. While existing research has examined the public's demand for AI in policy implementation, very little scholarship has examined the views of bureaucrats. In this article, we implement two survey experiments—one on predictive AI and another on generative AI—among government employees to reveal how information about the effects of AI on public values (equity and efficiency) affect bureaucrats' decisions on whether to utilize AI to ease administrative burdens on clients. We find that, regardless of AI type (i.e., predictive vs. generative) and randomly assigned policy contexts with different socially constructed target populations, bureaucrats are less likely to adopt AI when they are presented with information about the potential biases embedded in algorithms and the negative impacts on social equity. On the other hand, we find that highlighting the efficiency gains of utilizing AI has no impact on support for using AI. These findings suggest that concerns over equity may be a key barrier to bureaucrats' decisions on adopting AI in policy implementation.

Characterizing the progression of subclinical cardiac amyloidosis through artificial intelligence applied to electrocardiographic images and echocardiograms

Abstract Background Transthyretin amyloid cardiomyopathy (ATTR-CM) follows an indolent course and may remain undiagnosed for many years before resulting in clinical disease. Purpose To assess the utility of scalable artificial intelligence (AI) methods applied to electrocardiographic (ECG) images and echocardiograms in detecting longitudinal changes that precede the clinical diagnosis of ATTR-CM. Methods Across 5 hospitals of a large U.S.-based hospital system, we identified a total of 234 patients (age 79 [IQR:70-85] years, n=56 (23.9%) women) with ATTR-CM, defined by the presence of a positive bone scintigraphy scan (or pharmacotherapy with an approved transthyretin stabilizer) between 2015 and 2023. Using 10:1 age- and sex-matched controls, we independently trained convolutional neural network algorithms to discriminate ATTR-CM from controls using i) ECG images (AUROC 0.915) or ii) echocardiographic studies by incorporating all available views (AUROC 0.93). In a leave-out testing set of n=82 individuals with serial ECGs and n=49 individuals with serial echocardiograms we characterized longitudinal patterns in the evolution of the AI ECG and echo-based scores, and estimated the percentage of individuals that would have screened positive across distinct time intervals. Results In the leave-out testing set 52.4% and 34.7% of all individuals who went on to receive a diagnosis of ATTR-CM would have screened positive by AI deployed to ECGs and echocardiograms, respectively, 1 to 3 years before their confirmatory testing, with the percentage increasing to 75.6% and 69.4% in the 12 months before their eventual diagnosis (A). Across both modalities, there was a longitudinal increase in AI predictions when comparing the median of all studies performed more than a year before diagnosis versus the median of any studies performed within 12 months of diagnosis (AI-Echo: 0.20 [IQR: 0.04-0.35] to 0.49 [0.24-0.59] &amp; AI-ECG: 0.07 [0.02-0.20] to 0.32 [0.11-0.49], p&amp;lt;0.001 for both) (B), consistent with evolution in the ECG and echocardiographic signature of ATTR-CM before its clinical diagnosis. Conclusions We demonstrate that AI applied directly to ECG images and echocardiography may enable scalable monitoring of subclinical ATTR-CM progression.

AI in terrorism sentencing: evaluating predictive accuracy and ethical implications

ABSTRACT This study investigates the potential of artificial intelligence (AI) to predict sentencing outcomes in terrorism-related attempted murder cases. By comparing AI model predictions with actual sentences from Israeli civilian and military courts, the research evaluates AI’s accuracy and consistency. Findings indicate that AI predictions align closely with human judgments yet also reveal significant challenges in capturing case-specific nuances. This study contributes to the debate on AI’s role in the legal system, providing insights into developing responsible and effective AI integration in judicial decision-making.

Artificial Intelligence-Assisted Daily Quality Control System for the Histologic Diagnosis of Gastrointestinal Endoscopic Biopsies: A 1-Year Experience.

Seegene Medical Foundation, one of the major clinical laboratories in South Korea, developed SeeDP, an artificial intelligence (AI)-based postanalytic daily quality control (QC) system that reassesses all gastrointestinal (GI) endoscopic biopsy (EB) slides for incorrect diagnoses. To review the operational records and clinical impact of SeeDP since its launch in March 2022. Operational records of SeeDP were retrieved for the period of March 1, 2022, to February 28, 2023. Among cases scanned during 40 working days (March 10, 2022, to May 4, 2022), all discordant cases encountered by 2 pathologists were reviewed. Cases of SeeDP-assisted revised diagnoses were collected and compared with cases recognized using conventional methods. Occasional scanner failures and various types of aberrant errors compromised QC coverage, resulting in the scanning of only 67.7% (572 254 of 844 906) of all EB slides submitted and 0.8% of the scanned slides being further excluded from the AI analysis. The AI predictions differed from the pathologists' diagnoses in 42 760 of the 557 672 gastrointestinal EB slides (7.7%) successfully assessed by the AI models; however, a detailed review of discordant slides revealed that true misdiagnosis accounted for only 5.5% (25 of 454) of the disagreements. Compared with conventional error recognition methods, SeeDP detected more misdiagnoses (7 versus 14) within a significantly shorter time (average, 3.6 versus 38.7 days; P < .001), including 1 signet ring cell carcinoma initially diagnosed as gastritis. AI-based daily QC systems are plausible solutions to guarantee high-quality pathologic diagnosis by enabling rapid detection and correction of misdiagnosis.

Credit Risk Assessment and Financial Decision Support Using Explainable Artificial Intelligence

The greatest technological transformation the world has ever seen was brought about by artificial intelligence (AI). It presents significant opportunities for the financial sector to enhance risk management, democratize financial services, ensure consumer protection, and improve customer experience. Modern machine learning models are more accessible than ever, but it has been challenging to create and implement systems that support real-world financial applications, primarily due to their lack of transparency and explainability—both of which are essential for building trustworthy technology. The novelty of this study lies in the development of an explainable AI (XAI) model that not only addresses these transparency concerns but also serves as a tool for policy development in credit risk management. By offering a clear understanding of the underlying factors influencing AI predictions, the proposed model can assist regulators and financial institutions in shaping data-driven policies, ensuring fairness, and enhancing trust. This study proposes an explainable AI model for credit risk management, specifically aimed at quantifying the risks associated with credit borrowing through peer-to-peer lending platforms. The model leverages Shapley values to generate AI predictions based on key explanatory variables. The decision tree and random forest models achieved the highest accuracy levels of 0.89 and 0.93, respectively. The model’s performance was further tested using a larger dataset, where it maintained stable accuracy levels, with the decision tree and random forest models reaching accuracies of 0.90 and 0.93, respectively. To ensure reliable explainable AI (XAI) modeling, these models were chosen due to the binary classification nature of the problem. LIME and SHAP were employed to present the XAI models as both local and global surrogates.