Misclassification Rate Research Articles

Risk Management in Using Artificial Neural Networks

The article examines risks faced by banks during their lending processes and the mechanisms for managing these risks, utilizing modern statistical methods. Specifically, the study focused on the artificial neural network model as a technique of artificial intelligence that has successfully applied various classifications and discrimination tasks among institutions. A random sample of 46 institutions obtained loans from the branches of the National Bank of Algeria (BNA), Local Development Bank (BDL), Popular Credit of Algeria (CPA), and Agricultural and Rural Development Bank (BADR) in El Bayadh province, Algeria. Each of these institutions was characterized by 14 measurable variables with numerical values derived from the financial statements (balance sheets and income statements), as well as 3 qualitative non-accounting variables extracted from the loan applicants’ files (age of the institution, sector of activity (services/productive), institution status (viable/struggling). The sample of these 46 institutions was initially divided into two groups: 64% comprised financially stable institutions, and the other 36% were struggling institutions. The research checks whether the risk assessment of each of these 46 institutions using artificial neural networks will identify their institution status (viable/struggling) in the same way as it was in the base sample. The training phase recorded a prediction error rate of 0%, and the network testing phase misclassification rate was 5.6%. The overall correct classification rate for the multilayer artificial neural network was 92.9%, with a total error rate of 7.1%. The contribution rate of the non-accounting variable “sector of activity” was 100%, and the variable “age of the institution” was 94.4%. Other variables had minor percentages, underscoring the importance of qualitative variables in the classification process. Thus, the study proved that artificial neural network model is an effective model for distinguishing between viable and struggling institutions, significantly contributing to banking risk management.

LIVERSTAT for risk stratification for patients with metabolic dysfunction-associated fatty liver disease.

LIVERSTAT is an artificial intelligence-based noninvasive test devised to screen for and provide risk stratification for metabolic dysfunction-associated fatty liver disease (MAFLD) by using simple blood biomarkers and anthropometric measurements. We aimed to study LIVERSTAT in patients with MAFLD and to explore its role for the diagnosis of advanced fibrosis. This is a retrospective study of data from MAFLD patients who underwent a liver biopsy. Patients with type 2 diabetes who underwent transient elastography and had liver stiffness measurement (LSM)<5kPa were included as patients with no fibrosis. Among these patients, controlled attenuation parameter <248dB/m was considered as no steatosis. The LIVERSTAT results were generated based on a proprietary algorithm, blinded to the histological and LSM data. The data for 350 patients were analyzed (mean age 53years, 45% male, advanced fibrosis 22%). The sensitivity, specificity, positive predictive value, negative predictive value, and misclassification rate of LIVERSTAT to diagnose advanced fibrosis were 90%, 50%, 30%, 95%, and 42%, respectively. The corresponding rates for Fibrosis-4 score (FIB4) were 56%, 83%, 44%, 89%, and 22%, respectively. When LSM was used as a second test, the corresponding rates for LIVERSTAT were 60%, 97%, 76%, 94%, and 8%, respectively, while the corresponding rates for FIB4 were 38%, 99%, 83%, 89%, and 11%, respectively. LIVERSTAT had a higher negative predictive value compared with FIB4 and a lower misclassification rate compared with FIB4 when used in a two-step approach in combination with LSM for the diagnosis of advanced fibrosis.

Validation of the Subaxial Cervical Spine Injury Classification score in children: a single-institution experience at a level 1 pediatric trauma center.

The Subaxial Cervical Spine Injury Classification (SLIC) score has not been previously validated for a pediatric population. The authors compared the SLIC treatment recommendations for pediatric subaxial cervical spine trauma with real-world pediatric spine surgery practice. A retrospective cohort study at a pediatric level 1 trauma center was conducted in patients < 18 years of age evaluated for trauma from 2012 to 2021. An SLIC score was calculated for each patient, and the subsequent recommendations were compared with actual treatment delivered. Percentage misclassification, sensitivity, specificity, positive (PPV) and negative predictive value (NPV), and area under the receiver operating characteristic (ROC) curve (AUC) were calculated. Two hundred forty-three pediatric patients with trauma were included. Twenty-five patients (10.3%) underwent surgery and 218 were managed conservatively. The median SLIC score was 2 (interquartile range = 2). Sixteen patients (6.6%) had an SLIC score of 4, for which either conservative or surgical treatment is recommended; 27 children had an SLIC score ≥ 5, indicating a recommendation for surgical treatment; and 200 children had an SLIC score ≤ 3, indicating a recommendation for conservative treatment. Of the 243 patients, 227 received treatment consistent with SLIC score recommendations (p < 0.001). SLIC sensitivity in determining surgically treated patients was 79.2% and the specificity for accurately determining who underwent conservative treatment was 96.1%. The PPV was 70.3% and the NPV was 97.5%. There was a 5.7% misclassification rate (n = 13) using SLIC. Among patients for whom surgical treatment would be recommended by the SLIC, 29.6% (n = 8) did not undergo surgery; similarly, 2.5% (n = 5) of patients for whom conservative management would be recommended by the SLIC had surgery. The ROC curve for determining treatment received demonstrated excellent discriminative ability, with an AUC of 0.96 (OR 3.12, p < 0.001). Sensitivity decreased when the cohort was split by age (< 10 and ≥ 10 years old) to 0.5 and 0.82, respectively; specificity remained high at 0.98 and 0.94. The SLIC scoring system recommended similar treatment when compared with the actual treatment delivered for traumatic subaxial cervical spine injuries in children, with a low misclassification rate and a specificity of 96%. These findings demonstrate that the SLIC can be useful in guiding treatment for pediatric patients with subaxial cervical spine injuries. Further investigation into the score in young children (< 10 years) using a multicenter cohort is warranted.

The burden of diagnostic error in dentistry: A study on periodontal disease misclassification

BackgroundPeriodontal disease constitutes a widely prevalent category of non-communicable diseases and ranks among the top 10 causes of disability worldwide. Little however is known about diagnostic errors in dentistry. In this work, by retrospectively deploying an electronic health record (EHR)-based trigger tool, followed by gold standard manual review, we provide epidemiological estimates on the rate of diagnostic misclassification in dentistry through a periodontal use case. MethodsAn EHR-based trigger tool (a retrospective record review instrument that uses a list of triggers (or clues), i.e., data elements within the health record, to alert reviewers to the potential presence of a wrong diagnosis) was developed, tested and run against the EHR at the two participating sites to flag all cases having a potential misdiagnosis. All cases flagged as potentially misdiagnosed underwent extensive manual reviews by two calibrated domain experts. A subset of the non-flagged cases was also manually reviewed. ResultsA total of 2,262 patient charts met the study's inclusion criteria. Of these, the algorithm flagged 1,124 cases as potentially misclassified and 1,138 cases as potentially correctly diagnosed. When the algorithm identified a case as potentially misclassified, compared to the diagnosis assigned by the gold standard, the kappa statistic was 0.01. However, for cases the algorithm marked as potentially correctly diagnosed, the review against the gold standard showed a kappa statistic of 0.9, indicating near perfect agreement. The observed proportion of diagnostic misclassification was 32 %. There was no significant difference by clinic or provider characteristics. ConclusionOur work revealed that about a third of periodontal cases are misclassified. Diagnostic errors have been reported to happen more frequently than other types of errors, and to be more preventable. Benchmarking diagnostic quality is a first step. Subsequent research endeavor will delve into comprehending the factors that contribute to diagnostic errors in dentistry and instituting measures to prevent them. Clinical SignificanceThis study sheds light on the significance of diagnostic excellence in the delivery of dental care, and highlights the potential role of technology in aiding diagnostic decision-making at the point of care.

Achieving Better Category Separability for Hyperspectral Image Classification: A Spatial-Spectral Approach.

The task of hyperspectral image (HSI) classification has attracted extensive attention. The rich spectral information in HSIs not only provides more detailed information but also brings a lot of redundant information. Redundant information makes spectral curves of different categories have similar trends, which leads to poor category separability. In this article, we achieve better category separability from the perspective of increasing the difference between categories and reducing the variation within category, thus improving the classification accuracy. Specifically, we propose the template spectrum-based processing module from spectral perspective, which can effectively expose the unique characteristics of different categories and reduce the difficulty of model mining key features. Second, we design an adaptive dual attention network from spatial perspective, where the target pixel can adaptively aggregate high-level features by evaluating the confidence of effective information in different receptive fields. Compared with the single adjacency scheme, the adaptive dual attention mechanism makes the ability of target pixel to combine spatial information to reduce variation more stable. Finally, we designed a dispersion loss from the classifier's perspective. By supervising the learnable parameters of the final classification layer, the loss makes the category standard eigenvectors learned by the model more dispersed, which improves the category separability and reduces the rate of misclassification. Experiments on three common datasets show that our proposed method is superior to the comparison method.

Adequacy of sigmoidoscopy as compared to colonoscopy for assessment of disease activity in patients of ulcerative colitis: a prospective study.

Patients of ulcerative colitis (UC) on follow-up are routinely evaluated by sigmoidoscopy. There is no prospective literature to support this practice. We assessed agreement between sigmoidoscopy and colonoscopy prospectively in patients with disease extent beyond the sigmoid colon. We conducted a prospective observational study at a tertiary care institute for agreement between sigmoidoscopy and colonoscopy. We assessed endoscopic activity using the Mayo Endoscopic Score (MES) and Ulcerative Colitis Endoscopic Index of Severity (UCEIS) and histological activity using the Nancy Index (NI), Robarts Histopathology Index (RHI), and Simplified Geboes Score (SGS). Sigmoidoscopy showed a strong agreement with colonoscopy for MES and UCEIS with a kappa (κ) of 0.96 and 0.94 respectively. The misclassification rate for MES and UCEIS was 3% and 5% respectively. Sigmoidoscopy showed perfect agreement (κ = 1.00) with colonoscopy for assessment of the presence of endoscopic activity in the colon using MES ≥ 1 as activity criteria and strong agreement (κ = 0.93) using MES > 1 as activity criteria. Sigmoidoscopy showed strong agreement with colonoscopy for assessment of the presence of endoscopic activity using UCEIS (κ = 0.92). Strong agreement was observed between sigmoidoscopy and colonoscopy using NI (κ = 0.86), RHI (κ = 1.00), and SGS (κ = 0.92) for the detection of histological activity. The misclassification rate for the detection of histological activity was 2%, 0%, and 1% for NI, RHI, and SGS respectively. Sigmoidoscopy showed strong agreement with colonoscopy for endoscopic and histologic disease activity. Sigmoidoscopy is adequate for assessment of disease activity in patients with UC during follow-up evaluation.

The impact of misclassification on covariate-adaptive randomized clinical trials with generalized linear models

Covariate-adaptive randomization (CAR) is a type of randomization method that uses covariate information to enhance the comparability between different treatment groups. Under such randomization, the covariate is usually well balanced, i.e., the imbalance between the treatment group and placebo group is controlled. In practice, the covariate is sometimes misclassified. The covariate misclassification affects the CAR itself and statistical inferences after the CAR. In this paper, we examine the impact of covariate misclassification on CAR from two aspects. First, we study the balancing properties of CAR with unequal allocation in the presence of covariate misclassification. We show the convergence rate of the imbalance and compare it with that under true covariate. Second, we study the hypothesis test under CAR with misclassified covariates in a generalized linear model (GLM) framework. We consider both the unadjusted and adjusted models. To illustrate the theoretical results, we discuss the validity of test procedures for three commonly-used GLM, i.e., logistic regression, Poisson regression and exponential model. Specifically, we show that the adjusted model is often invalid when the misclassified covariates are adjusted. In this case, we provide a simple correction for the inflated Type-I error. The correction is useful and easy to implement because it does not require misclassification specification and estimation of the misclassification rate. Our study enriches the literature on the impact of covariate misclassification on CAR and provides a practical approach for handling misclassification.

Diagnostic performance of plasma pTau217, pTau181, Aβ1-42 and Aβ1-40 in the LUMIPULSE automated platform for the detection of Alzheimer disease

BackgroundRecently developed blood markers for Alzheimer's disease (AD) detection have high accuracy but usually require ultra-sensitive analytic tools not commonly available in clinical laboratories, and their performance in clinical practice is unknown.MethodsWe analyzed plasma samples from 290 consecutive participants that underwent lumbar puncture in routine clinical practice in a specialized memory clinic (66 cognitively unimpaired, 130 participants with mild cognitive impairment, and 94 with dementia). Participants were classified as amyloid positive (A +) or negative (A-) according to CSF Aβ1–42/Aβ1–40 ratio. Plasma pTau217, pTau181, Aβ1–42 and Aβ1–40 were measured in the fully-automated LUMIPULSE platform. We used linear regression to compare plasma biomarkers concentrations between A + and A- groups, evaluated Spearman’s correlation between plasma and CSF and performed ROC analyses to assess their diagnostic accuracy to detect brain amyloidosis as determined by CSF Aβ1–42/Aβ1–40 ratio. We analyzed the concordance of pTau217 with CSF amyloidosis.ResultsPlasma pTau217 and pTau181 concentration were higher in A + than A- while the plasma Aβ1–42/Aβ1–40 ratio was lower in A + compared to A-. pTau181 and the Aβ1–42/Aβ1–40 ratio showed moderate correlation between plasma and CSF (Rho = 0.66 and 0.69, respectively). The areas under the ROC curve to discriminate A + from A- participants were 0.94 (95% CI 0.92–0.97) for pTau217, and 0.88 (95% CI 0.84–0.92) for both pTau181 and Aβ1–42/Aβ1–40. Chronic kidney disease (CKD) was related to increased plasma biomarker concentrations, but ratios were less affected. Plasma pTau217 had the highest fold change (× 3.2) and showed high predictive capability in discriminating A + from A-, having 4–7% misclassification rate. The global accuracy of plasma pTau217 using a two-threshold approach was robust in symptomatic groups, exceeding 90%.ConclusionThe evaluation of blood biomarkers on an automated platform exhibited high diagnostic accuracy for AD pathophysiology, and pTau217 showed excellent diagnostic accuracy to identify participants with AD in a consecutive sample representing the routine clinical practice in a specialized memory unit.

Swarm optimized differential evolution and probabilistic extreme learning based intrusion detection in MANET

MANETs are an attracting mechanism foSr several applications, to name a few being rescue functioning, environmental surveillance and so on due to the reason that they allow users to communicate without the utilization of persistent framework. This pliability, although creates additional security proneness. As a consequence of its advantages and growing insistence, MANETs have fascinated a lot of attentiveness from the scientific research community. In spite of that, exposed or unprotected to be more susceptible to several attacks that perpetrate destruction on their performance than any network. Conventional cryptography mechanisms cannot completely safeguard MANETs in terms of susceptibility owing to dispersed nature however these issues can be addressed by using optimization and deep learning techniques-based Intrusion Detection System (IDS). In this work, we develop a Binary Swarm Optimized Differential and Method of Moments Probabilistic Extreme Learning (BSOD-MMPEL) node behavior-based IDS for intrusion detection in MANET with minimum training time, misclassification and high precision, accuracy is proposed. Initially, network samples are collected from KDD-CUP-1999 dataset. To ensure robust and significant balance between exploration and exploitation with high probability of convergence to local sub-optima Gudermannian Activation Binary Swarm Optimized Differential Evolution-based Feature Selection is applied. Next, with the selected features, early detection of intrusion employing Method of Moments Probabilistic Extreme Learning Node Behavior-based IDS is designed. Simulations are performed to validate the result. The performance evaluation in terms of precision, accuracy, sensitivity, training time and misclassification rate show that the proposed method outperforms existing IDS in MANET. The proposed BSOD-MMPEL method is achieved to improve precision of 23 % and accuracy of 27 % and reduce training time of 62 % and misclassification rate of 68 % than existing conventional methods.

Exploring the variation in associations between socioeconomic indicators and non-communicable diseases in the Tromsø Study: an algorithmic approach

Aims: We contribute to the methodological literature on the assessment of health inequalities by applying an algorithmic approach to evaluate the capabilities of socioeconomic variables in predicting the prevalence of non-communicable diseases in a Norwegian health survey. Methods: We use data from the seventh survey of the population based Tromsø Study (2015–2016), including 11,074 women and 10,009 men aged 40 years and above. We apply the random forest algorithm to predict four non-communicable disease outcomes (heart attack, cancer, diabetes and stroke) based on information on a number of social root causes and health behaviours. We evaluate our results using the classification error, the mean decrease in accuracy, partial dependence statistics. Results: Results suggest that education, household income and occupation to a variable extent contribute to predicting non-communicable disease outcomes. Prediction misclassification ranges between 25.1% and 35.4% depending on the non-communicable diseases under study. Partial dependences reveal mostly expected health gradients, with some examples of complex functional relationships. Out-of-sample model validation shows that predictions translate to new data input. Conclusions: Algorithmic modelling can provide additional empirical detail and metrics for evaluating heterogeneous inequalities in morbidity. The extent to which education, income and occupation contribute to predicting binary non-communicable disease outcomes depends on both non-communicable diseases and socioeconomic indicator. Partial dependences reveal that social gradients in non-communicable disease outcomes vary in shape between combinations of non-communicable disease outcome and socioeconomic status indicator. Misclassification rates highlight the extent of variation within socioeconomic groups, suggesting that future studies may improve predictive accuracy by exploring further subpopulation heterogeneity.

Outcome-based analytical performance specifications: current status and future challenges.

Analytical performance specifications (APS) based on outcomes refer to how 'good' the analytical performance of a test needs to be to do more good than harm to the patient. Analytical performance of a measurand affects its clinical performance. Without first setting clinical performance requirements, it is difficult to define how good analytically the test needs to be to meet medical needs. As testing is indirectly linked to health outcomes through clinical decisions on patient management, often simulation-based studies are used to assess the impact of analytical performance on the probability of clinical outcomes which is then translated to Model 1b APS according to the Milan consensus. This paper discusses the related key definitions, concepts and considerations that should assist in finding the most appropriate methods for deriving Model 1b APS. We review the advantages and limitations of published methods and discuss the criteria for transferability of Model 1b APS to different settings. We consider that the definition of the clinically acceptable misclassification rate is central to Model 1b APS. We provide some examples and guidance on a more systematic approach for first defining the clinical performance requirements for tests and we also highlight a few ideas to tackle the future challenges associated with providing outcome-based APS for laboratory testing.

Meta-Fed IDS: Meta-learning and Federated learning based fog-cloud approach to detect known and zero-day cyber attacks in IoMT networks

The Internet of Medical Things (IoMT) is a transformative fusion of medical sensors, equipment, and the Internet of Things, positioned to transform healthcare. However, security and privacy concerns hinder widespread IoMT adoption, intensified by the scarcity of high-quality datasets for developing effective security solutions. Addressing these challenges, we propose a novel framework for cyberattack detection in dynamic IoMT networks. This framework integrates Federated Learning with Meta-learning, employing a multi-phase architecture for identifying known attacks, and incorporates advanced clustering and biased classifiers to address zero-day attacks. The framework's deployment is adaptable to dynamic and diverse environments, utilizing an Infrastructure-as-a-Service (IaaS) model on the cloud and a Software-as-a-Service (SaaS) model on the fog end. To reflect real-world scenarios, we introduce a specialized IoMT dataset. Our experimental results indicate high accuracy and low misclassification rates, demonstrating the framework's capability in detecting cyber threats in complex IoMT environments. This approach shows significant promise in bolstering cybersecurity in advanced healthcare technologies.

Multivariate functional principal component analysis and k-means clustering to identify kinematic foot types during gait in children with cerebral palsy

BackgroundChildren with neuromuscular disorders, such as cerebral palsy, frequently develop foot deformities, such as equinopronovalgus and equinosupovarus, leading to walking difficulties and discomfort. Traditional assessment methods, including clinical measures and radiographs, often fail to capture the dynamic nature of these deformities, resulting in suboptimal treatment. 3D gait analysis using multisegment foot models offers a more detailed understanding of these deformities. Research questionTo determine whether the combination of multisegment foot models, multivariate functional principal component analysis, and k-means cluster analyses could identify distinct, clinically relevant foot types in a large pediatric cohort with cerebral palsy. MethodsThis was a retrospective analysis of 3D gait data from 197 patients with cerebral palsy collected using a multisegment foot model. Multivariate functional principal component analysis was used to reduce these data prior to using k-means clustering to identify foot posture clusters. Further analyses, including ANOVA and Fisher's Exact tests, were used to evaluate demographic, radiographic, and gait characteristics to explain the clinical relevance of each cluster. ResultsAnalysis of kinematic data from 371 feet revealed six clinically significant clusters, with a low misclassification rate of 2 %. One-factor ANOVAs demonstrated significant differences across clusters for all MPCs, whereas no significant differences were noted in basic anthropometric variables. Significant variations were observed in radiographic and gait function variables, and a strong association between GMFCS levels and cluster categorization was identified. SignificanceThe novel approach of integrating multivariate functional principal component analysis and k-means clustering identified a spectrum of foot deformities in children with CP, ranging from equinosupovarus to marked equinopronovalgus. This methodology provides an objective classification based on kinematic data and can facilitate improved diagnosis and treatment of cerebral palsy-related foot deformities.

Prediction of Student Dropout in Malaysian’s Private Higher Education Institute using Data Mining Application

Student dropout issue is a major concern among the academics and management of the university. The higher rate of student dropout impacted the university reputation such as reducing student enrolment, affecting the revenue of the university, financial losses for the country, and increase the existence of a social problem among the students. In this study, 2 popular classifiers were utilized to predict the student dropout namely decision tree and logistic regression model respectively. Several sets of experimental setting were employed which include three set of data partitioning - along with different types of decision tree and regression model. As for the logistic regression model, different data imputation and transformation method was tested to ensure that the model built is valid. A total of 7706 student data extracted from one of the private universities in Malaysia database (between year 2018-2019) to assess the capability of the classifier. The classifier performance is evaluated using machine learning performance measure of accuracy and misclassification rate. The result indicates that, decision tree - chi-square (2 branches) achieved slightly better classification performance of 89.49% on 80/20 data partitioning. The chosen model also identified the most important variable for accurate prediction of student dropout. Application of this model has the potential to accurately predict at risk student and to reduce student dropout rates.

Spectral co-clustering in multi-layer directed networks

Modern network analysis often involves multi-layer network data in which the nodes are aligned, and the edges on each layer represent one of the multiple relations among the nodes. Current literature on multi-layer network data is mostly limited to undirected relations. However, direct relations are more common and may introduce extra information. This study focuses on community detection (or clustering) in multi-layer directed networks. To take into account the asymmetry, a novel spectral-co-clustering-based algorithm is developed to detect co-clusters, which capture the sending patterns and receiving patterns of nodes, respectively. Specifically, the eigendecomposition of the debiased sum of Gram matrices over the layer-wise adjacency matrices is computed, followed by the k-means, where the sum of Gram matrices is used to avoid possible cancellation of clusters caused by direct summation. Theoretical analysis of the algorithm under the multi-layer stochastic co-block model is provided, where the common assumption that the cluster number is coupled with the rank of the model is relaxed. After a systematic analysis of the eigenvectors of the population version algorithm, the misclassification rates are derived, which show that multi-layers would bring benefits to the clustering performance. The experimental results of simulated data corroborate the theoretical predictions, and the analysis of a real-world trade network dataset provides interpretable results.

Enhancing palliative care triage: decision support system for patient prioritisation

ABSTRACT This study presents an innovative approach to patient triaging in palliative care through a convolutional neural network (CNN). CNN, trained on a dataset annotated by medical experts, is adept at identifying levels of patient urgency from complex clinical data and is tailored to decipher intricate patterns enabling the accurate prioritisation of care needs. Our evaluation involved experimental simulations to test the model’s precision and reliability in triaging. CNN demonstrated exceptional performance, with all major metrics – recall, F1-Score, precision, and accuracy – exceeding 98%. This represents a substantial improvement over current triaging methods and underscores the potential of AI to lead healthcare innovation. The model’s precision in identifying urgent patient cases, coupled with its low misclassification rate, is particularly pertinent to the high-stakes environment of palliative care.These outcomes suggest that AI has the potential to transform palliative care triaging, ensuring that patients receive timely and appropriate interventions is promising in terms of enhancing the responsiveness of care delivery, marking a pivotal step towards a more patient-focused approach.

The great detectives: humans versus AI detectors in catching large language model-generated medical writing

BackgroundThe application of artificial intelligence (AI) in academic writing has raised concerns regarding accuracy, ethics, and scientific rigour. Some AI content detectors may not accurately identify AI-generated texts, especially those that have undergone paraphrasing. Therefore, there is a pressing need for efficacious approaches or guidelines to govern AI usage in specific disciplines.ObjectiveOur study aims to compare the accuracy of mainstream AI content detectors and human reviewers in detecting AI-generated rehabilitation-related articles with or without paraphrasing.Study designThis cross-sectional study purposively chose 50 rehabilitation-related articles from four peer-reviewed journals, and then fabricated another 50 articles using ChatGPT. Specifically, ChatGPT was used to generate the introduction, discussion, and conclusion sections based on the original titles, methods, and results. Wordtune was then used to rephrase the ChatGPT-generated articles. Six common AI content detectors (Originality.ai, Turnitin, ZeroGPT, GPTZero, Content at Scale, and GPT-2 Output Detector) were employed to identify AI content for the original, ChatGPT-generated and AI-rephrased articles. Four human reviewers (two student reviewers and two professorial reviewers) were recruited to differentiate between the original articles and AI-rephrased articles, which were expected to be more difficult to detect. They were instructed to give reasons for their judgements.ResultsOriginality.ai correctly detected 100% of ChatGPT-generated and AI-rephrased texts. ZeroGPT accurately detected 96% of ChatGPT-generated and 88% of AI-rephrased articles. The areas under the receiver operating characteristic curve (AUROC) of ZeroGPT were 0.98 for identifying human-written and AI articles. Turnitin showed a 0% misclassification rate for human-written articles, although it only identified 30% of AI-rephrased articles. Professorial reviewers accurately discriminated at least 96% of AI-rephrased articles, but they misclassified 12% of human-written articles as AI-generated. On average, students only identified 76% of AI-rephrased articles. Reviewers identified AI-rephrased articles based on ‘incoherent content’ (34.36%), followed by ‘grammatical errors’ (20.26%), and ‘insufficient evidence’ (16.15%).Conclusions and relevanceThis study directly compared the accuracy of advanced AI detectors and human reviewers in detecting AI-generated medical writing after paraphrasing. Our findings demonstrate that specific detectors and experienced reviewers can accurately identify articles generated by Large Language Models, even after paraphrasing. The rationale employed by our reviewers in their assessments can inform future evaluation strategies for monitoring AI usage in medical education or publications. AI content detectors may be incorporated as an additional screening tool in the peer-review process of academic journals.

External Validation of a Machine Learning Model for Schizophrenia Classification.

Background and Objective: Excellent generalizability is the precondition for the widespread practical implementation of machine learning models. In our previous study, we developed the schizophrenia classification model (SZ classifier) to identify potential schizophrenia patients in the Japanese population. The SZ classifier has exhibited impressive performance during internal validation. However, ensuring the robustness and generalizability of the SZ classifier requires external validation across independent sample sets. In this study, we aimed to present an external validation of the SZ classifier using outpatient data. Methods: The SZ classifier was trained by using online survey data, which incorporate demographic, health-related, and social comorbidity features. External validation was conducted using an outpatient sample set which is independent from the sample set during the model development phase. The model performance was assessed based on the sensitivity and misclassification rates for schizophrenia, bipolar disorder, and major depression patients. Results: The SZ classifier demonstrated a sensitivity of 0.75 when applied to schizophrenia patients. The misclassification rates were 59% and 55% for bipolar disorder and major depression patients, respectively. Conclusions: The SZ classifier currently encounters challenges in accurately determining the presence or absence of schizophrenia at the individual level. Prior to widespread practical implementation, enhancements are necessary to bolster the accuracy and diminish the misclassification rates. Despite the current limitations of the model, such as poor specificity for certain psychiatric disorders, there is potential for improvement if including multiple types of psychiatric disorders during model development.

Tuning-free sparse clustering via alternating hard-thresholding

Model-based clustering is a commonly-used technique to partition heterogeneous data into homogeneous groups. When the analysis is to be conducted with a large number of features, analysts face simultaneous challenges in model interpretability, clustering accuracy, and computational efficiency. Several Bayesian and penalization methods have been proposed to select important features for model-based clustering. However, the performance of those methods relies on a careful algorithmic tuning, which can be time-consuming for high-dimensional cases. In this paper, we propose a new sparse clustering method based on alternating hard-thresholding. The new method is conceptually simple and tuning-free. With a user-specified sparsity level, it efficiently detects a set of key features by eliminating a large number of features that are less useful for clustering. Based on the selected key features, one can readily obtain an effective clustering of the original high-dimensional data under a general sparse covariance structure. Under mild conditions, we show that the new method leads to clusters with a misclassification rate consistent to the optimal rate as if the underlying true model were used. The promising performance of the new method is supported by both simulated and real data examples.

A multichannel analysis of imbalanced computed tomography data for lung cancer classification

Lung cancer holds the highest fatality rate among cancers, emphasizing the importance of early detection. Computer algorithms have gained prominence across various domains, including lung cancer diagnosis. These algorithms assist specialists, especially in medical imaging, yet current efforts lack comprehensive CT data analysis; especially in handling imbalanced datasets and fully exploiting spatial information. The lack of spatial analysis hinders the ability to identify subtle variations in texture and structure that are crucial for detecting lung cancer early and accurately. Therefore, this study uses a multichannel analysis of computed tomography (CT) images and deep learning-based ensemble learning (MC-ECNN) to find lung cancer even when the data is not balanced. Firstly, the data imbalance issue is tackled through the synthetic minority oversampling technique (SOMTE); afterwards, a multi-channel analysis of the data is performed to explore a distinct set of abstract features. Lastly, a deep ensemble learning method is used to classify the extracted distinct abstract feature set into the appropriate classes. The proposed method uses the discrete Fast Fourier transform (DFFT) and discrete cosine transform (DCT), along with the actual CT scans, for the multi-channel analysis of the data in different domains. The proposed model yielded 99.60% test accuracy on unseen data, which is at least 3% better than the other state-of-the-art studies considered for the comparison. In addition to the classification accuracy, the efficacy of the proposed model has also been justified through precision, recall, F1-score, support value, and misclassification rate.