Large Medical Data Research Articles

Uniformization and bounded Taylor series in Newton-Raphsonmethod improves computational performance for a multistate transition model estimation and inference.

Multistate transition models (MSTMs) are valuable tools depicting disease progression. However, due to the complexity of MSTMs, larger sample size and longer follow-up time in real-world data, the computation of statistical estimation and inference for MSTMs becomes challenging. A bounded Taylor series in Newton-Raphson procedure is proposed which leverages the uniformization technique to derive maximum likelihood estimates and corresponding covariance matrix. The proposed method, namely uniformization Taylor-bounded Newton-Raphson, is validated in three simulation studies, which demonstrate the accuracy in parameter estimation, the efficiency in computation time and robustness in terms of different situations. This method is also illustrated using a large electronic medical record data related to statin-induced side effects and discontinuation.

Co-occurring autism, ADHD, and gender dysphoria in children, adolescents, and young adults with eating disorders: an examination of pre- vs. post-COVID pandemic outbreak trends with real-time electronic health record data.

Incidence rates of autism, attention-deficit/hyperactivity disorder (ADHD), and gender dysphoria (GD) are rising not only in the general population, but particularly among children, adolescents, and young adults with eating disorders (EDs). While ED rates have risen during the COVID pandemic, trends in co-occurring autism, ADHD, and GD have yet to be investigated in detail or at scale by way of large electronic medical record data. To investigate trends in rates of co-occurring autism, ADHD, and GD among children, adolescents, and young adults with EDs in years prior to and during the COVID-19 pandemic. We utilized a de-identified multinational electronic health records database (TriNetX) with 48,558 individuals aged 5-26 diagnosed with eating disorders (EDs) at least twice between 2017 and 2022. The primary predictor variable differentiated between the years of each person's index (first) ED diagnosis (2017-2019 vs. 2020-2022). The primary outcome variable was the rate of new co-occurring psychiatric diagnoses of autism, ADHD, and GD in the year following each patient's first ED diagnosis. We applied propensity score-matched multivariable logistic regressions to compare primary outcomes between 2017-2019 and 2020-2022. Our analysis included 17,445 individuals diagnosed with EDs in 2017-2019 (8% autism, 13.5% ADHD, 1.9% GD) and 31,113 diagnosed with EDs in 2020-2022 (8% autism, 14.6% ADHD, 3.2% GD). After 1:1 propensity score matching, 17,202 individuals from the 2017-2019 cohort were matched to peers mirroring the 2020-2022 cohort. Those diagnosed in 2020-2022 showed a 19% (aOR[95%CI]=1.19[1.07-1.33]), 25% (aOR=1.25[1.04-1.49]), and 36% (aOR=1.36[1.07-1.74]) increase in odds for autism, ADHD, and GD diagnoses, respectively, within the 365 days after the index EDs diagnosis, compared to the 2017-2019 cohort. Rates of autism, ADHD, and GD are significantly higher in individuals with ED in the post-pandemic 2020-2022 cohort in comparison to the pre-pandemic 2017-2019 cohort, even after controlling for baseline levels of co-occurring psychiatric diagnoses. Such findings reveal a critical gap in our current understanding of the totality of ways in which COVID-19 may have impacted the onset and clinical course of EDs, autism, ADHD, and GD among children, adolescents, and young adults.

A Secure Parallel Pattern Mining System for Medical Internet of Things.

In this paper, a new generic parallel pattern mining framework called multi-objective Decomposition for Parallel Pattern-Mining (MD-PPM) is developed to solve challenges in the Internet of Medical Things through big data exploration. MD-PPM discovers important patterns by using decomposition and parallel mining methods to explore connectivity between medical data. First, a new technique, the multi-objective k-means algorithm, is used to aggregate medical data. A parallel pattern mining approach based on GPU and MapReduce architectures is also used to create useful patterns. To ensure complete privacy and security of the medical data, blockchain technology has been integrated throughout the system. Several tests were conducted to demonstrate the high performance of two sequential and graph pattern mining problems on large medical data and to evaluate the developed MD-PPM framework. From our results, our proposed MD-PPM has achieved strong results in terms of memory usage and computation time in terms of efficiency. Moreover, MD-PPM performs well in terms of accuracy and feasibility compared to existing models.

Role of artificial intelligence in medical radiology and imaging

Abstract Artificial intelligence (AI) is revolutionizing radiology, oncology, and other medicine and veterinary care areas. Adopting deep learning algorithms has significantly advanced image analysis and disease detection. This study explores how AI is reshaping the roles of radiologists and radiographers. It highlights its vital function in infection detection and control, as evidenced by its impact during the coronavirus disease 2019 (COVID-19) pandemic. In veterinary radiation oncology, AI supports complex contouring and treatment planning. However, while AI offers numerous advantages, its implementation must be cautiously approached. Radiologists face challenges, particularly the overwhelming volume of imaging data, which AI helps manage through artificial neural networks and machine learning (ML) algorithms—two significant innovations in this field. In veterinary radiation oncology, AI facilitates collaboration, standardization of data, and the creation of standard operating procedures. Early disease detection, enabled by AI, is essential for initiating treatments that can improve patient outcomes and prognosis. AI is crucial in analyzing large medical datasets, including imaging and clinical data, through advanced algorithms and ML techniques. In veterinary medicine, AI is key to addressing complex challenges in host–pathogen interactions, precision medicine, and predictive epidemiology. AI-powered solutions for continuous monitoring ensure that at-risk patients receive ongoing observation, enabling the rapid detection of changes in health markers. This approach is especially advantageous in managing chronic conditions, enabling proactive healthcare, and facilitating early intervention.

Impact of polygenic risk score for triglyceride trajectory and diabetic complications in subjects with type 2 diabetes based on large electronic medical record data from Taiwan: a case control study.

The prevalence of diabetic dyslipidemia has gradually increased worldwide and individuals with hypertriglyceridemia often have a high polygenic burden of triglyceride (TG)-increasing variants. However, the contribution of genetic variants to dyslipidemia in patients with type 2 diabetes (T2D) remains limited. Therefore, in this study, we aimed to investigate the genetic characteristics of longitudinal changes in TG levels among patients with T2D and summarize the genetic effects of polygenic risk score (PRS) on TG trajectory and risk of diabetic complications. We conducted a case-control study. A total of 11,312 patients with T2D with longitudinal TG and genetic data were identified from a large hospital database in Taiwan. We then performed a genome-wide association study and calculated the relative PRS. In total, 21 single-nucleotide polymorphisms (SNPs) related to TG trajectory were identified and yielded an area under the receiver operating characteristic curve (ROC) of 0.712 for high TG trajectory risk among Taiwanese patients with T2D. A cumulative genetic effect was observed for high TG trajectory, even when considering the adherence of a lipid-lowering agent in stratified analysis. An increased PRS increases high TG trajectory risk in a logistic regression model (odds ratio = 1.55; 95% confidence interval [CI] = 1.31-1.83 in the validation cohort). The TG-specific PRS was associated with the risk of diabetic microvascular complications, including diabetic retinopathy and nephropathy (with hazard ratios of 1.11 [95% CI = 1.01-1.21, P = 0.027] and 1.05 [95% CI = 1.01-1.1, P = 0.018], respectively). This study may contribute to the identification of patients with T2D who are at risk of abnormal TG levels and diabetic microvascular complications using polygenic information.

Artificial intelligence and skin cancer.

Artificial intelligence is poised to rapidly reshape many fields, including that of skin cancer screening and diagnosis, both as a disruptive and assistive technology. Together with the collection and availability of large medical data sets, artificial intelligence will become a powerful tool that can be leveraged by physicians in their diagnoses and treatment plans for patients. This comprehensive review focuses on current progress toward AI applications for patients, primary care providers, dermatologists, and dermatopathologists, explores the diverse applications of image and molecular processing for skin cancer, and highlights AI's potential for patient self-screening and improving diagnostic accuracy for non-dermatologists. We additionally delve into the challenges and barriers to clinical implementation, paths forward for implementation and areas of active research.

Personal Data Protection Model in IOMT-Blockchain on Secured Bit-Count Transmutation Data Encryption Approach

The Internet of Medical Things (IoMT) has paved the way for innovative approaches to collecting and managing medical data. With the large and sensitive medical data being processed hence, the need for a strong identity and privacy become necessary. The present paper suggests a comprehensive method of PriMedGuard which aims at protection of the personal medical information. The first stage will be data collection from devices and sensors, then data cleaning to transform the data into the required format. There is also a safety system in the system that registers and authenticates authorized entities as well as ETDO (Enhanced Tasmanian Devil Optimization algorithm) is used for generating asymmetric cryptographic keys. The data is encrypted using the Secure Bit-Count Transmutation (SBCT) Data Encryption Algorithm and then put in the locations provided by the InterPlanetary File System (IPFS), a decentralized and distributed storage system. A safe smart contract on the blockchain is created so that the data retrieval is secure and MedSecEnsemble Detection is proposed as an intrusion detection technique in the IoMT network. By using this method, data will stay available while at the same time integrity, confidentiality and protection against vulnerabilities are ensured. Hence, the Internet of Medical Things ecosystem will be secured from unauthorized access and possible security threats…

Machine learning application in ischemic stroke diagnosis, management, and outcome prediction: a narrative review

Stroke is a leading cause of disability and death. The condition requires prompt diagnosis and treatment. The quality of care provided to patients with stroke can vary depending on the availability of medical resources, which in turn, can affect prognosis. Recently, there has been growing interest in using machine learning (ML) to support stroke diagnosis and treatment decisions based on large medical data sets. Current ML applications in stroke care can be divided into two categories: analysis of neuroimaging data and clinical informationbased predictive models. Using ML to analyze neuroimaging data can increase the efficiency and accuracy of diagnoses. Commercial software that uses ML algorithms is already being used in the medical field. Additionally, the accuracy of predictive ML models is improving with the integration of radiomics and clinical data. is expected to be important for improving the quality of care for patients with stroke.

Diabetes mellitus prediction using machine learning within the scope of a generic framework

<span>Artificial intelligence (AI) based automated disease prediction has recently taken a significant place in the field of health informatics. However, due to unavailability of real time large scale medical data, the dynamic learning of prediction models remains principally subsided. This paper, therefore proposes a dynamic predictive modelling framework for chronic diseases prediction in real-time. The framework premise suggests creation of a centralized patient-indexed medical database to dynamically train machine learning (ML) models and predict risk levels of chronic diseases in real time. In this study, comprehensive empirical evaluations to train seven state-of-the-art ML models for diabetes risk prediction are performed in context of phase 2 of the suggested framework. The selected optimal model can then be dynamically applied to predict diabetes in phase 3 of the framework. Various metrics such as accuracy, precision, Recall, F1-score and receiver operating characteristic (ROC) curve are employed for evaluating performances of the trained models. Parameter tunings using different type of kernels, different number of neighbors and estimators are rigorously performed in order to create a suggestive literature for healthcare prediction ecosystem. Comparative analysis indicates high prediction accuracies on diabetes test data records for neural network and support vector machine (SVM) models as compared to other applied models.</span>

An efficient semi-supervised quality control system trained using physics-based MRI-artefact generators and adversarial training

Large medical imaging data sets are becoming increasingly available. A common challenge in these data sets is to ensure that each sample meets minimum quality requirements devoid of significant artefacts. Despite a wide range of existing automatic methods having been developed to identify imperfections and artefacts in medical imaging, they mostly rely on data-hungry methods. In particular, the scarcity of artefact-containing scans available for training has been a major obstacle in the development and implementation of machine learning in clinical research. To tackle this problem, we propose a novel framework having four main components: (1) a set of artefact generators inspired by magnetic resonance physics to corrupt brain MRI scans and augment a training dataset, (2) a set of abstract and engineered features to represent images compactly, (3) a feature selection process that depends on the class of artefact to improve classification performance, and (4) a set of Support Vector Machine (SVM) classifiers trained to identify artefacts. Our novel contributions are threefold: first, we use the novel physics-based artefact generators to generate synthetic brain MRI scans with controlled artefacts as a data augmentation technique. This will avoid the labour-intensive collection and labelling process of scans with rare artefacts. Second, we propose a large pool of abstract and engineered image features developed to identify 9 different artefacts for structural MRI. Finally, we use an artefact-based feature selection block that, for each class of artefacts, finds the set of features that provide the best classification performance. We performed validation experiments on a large data set of scans with artificially-generated artefacts, and in a multiple sclerosis clinical trial where real artefacts were identified by experts, showing that the proposed pipeline outperforms traditional methods. In particular, our data augmentation increases performance by up to 12.5 percentage points on the accuracy, F1, F2, precision and recall. At the same time, the computation cost of our pipeline remains low –less than a second to process a single scan– with the potential for real-time deployment. Our artefact simulators obtained using adversarial learning enable the training of a quality control system for brain MRI that otherwise would have required a much larger number of scans in both supervised and unsupervised settings. We believe that systems for quality control will enable a wide range of high-throughput clinical applications based on the use of automatic image-processing pipelines.

Artificial intelligence for dementia drug discovery and trials optimization.

Drug discovery and clinical trial design for dementia have historically been challenging. In part these challenges have arisen from patient heterogeneity, length of disease course, and the tractability of a target for the brain. Applying big data analytics and machine learning tools for drug discovery and utilizing them to inform successful clinical trial design has the potential to accelerate progress. Opportunities arise at multiple stages in the therapy pipeline and the growing availability of large medical data sets opens possibilities for big data analyses to answer key questions in clinical and therapeutic challenges. However, before this goal is reached, several challenges need to be overcome and only a multi-disciplinary approach can promote data-driven decision-making to its full potential. Herein we review the current state of machine learning applications to clinical trial design and drug discovery, while presenting opportunities and recommendations that can break down the barriers to implementation.

Cybersecurity of Medical Data Based on Big Data and Privacy Protection Method

Big data brings new opportunities to discover the new value of healthcare industry, since it can help us understand the hidden value of data deeply. This also brings new challenges: how to effectively manage and organize these datasets. Throughout the whole life cycle of publishing, storing, mining, and using big data in health care, different users are involved, so there are corresponding privacy protection methods and technologies for different life cycles. Data usage is the last and most important part of the whole life cycle. Therefore, this article proposes a privacy protection method for large medical data: an access control based on credibility of the requesting user. This model evaluates and quantifies doctors' credibility from the perspective of behavioral trust. Comparative experiments show that under the background of linear, geometric and exponential distribution trends and mixed trends, the regression model in this article is better than the existing methods in predicting trust accuracy and trust trends.

Prescription patterns of analgesics for oral conditions in india - Analysis of large medical audit data of outpatients in India's private healthcare sector.

Analgesic use needs to be regulated due to its adverse effects. This study aimed to analyse the change in prescription rates and patterns of the analgesics prescribed for various oral conditions and to analyse their trends across different age groups and gender to promote rational prescription of drugs and eventually influence regulatory policies. Secondary analysis was conducted on medical audit data collected from the private health sector in India. The prescription rate per 1000 persons per year was calculated from May 2013 to April 2016 using the mean projected population (PP) of India. Cross-tabulations were conducted to analyse the prescription rate and their changes across different age groups, gender and oral conditions. The mean analgesic prescription rate was highest among the 20-40 age group, and the highest increase was noted in 'non-steroidal anti-inflammatory drug (NSAID) combinations' (3.56 per 1000 persons per year) from May 2013 to April 2016. The 'NSAID combinations' group was also the most prescribed medication across all the oral conditions, with 'diseases of hard tissues' having the highest prescription rate (41.4 and 45.6 per 1000 persons per year, respectively, for 2013-14 and 2015-16). The results indicate an overall increase in the analgesic prescription rate, especially 'NSAID combinations' for each dental disease and age group, a finding that is hard to explain. Due to the lack of prescription guidelines in India, it is difficult to assess whether these analgesics were prescribed rationally or not.

The Integration of Hadoop and a Smart Utility Scoring with a Case-Based Reasoning System for Managing Large and Complex Medical Case Base

In case-based reasoning structure, the quality and the complexity of the case data play a significant role in searching, retrieving, updating, and holding distinct case data at or from distinct databases for various purposes. However, managing the very complex and large case-data sets is not easy. Furthermore, the complex algorithm for maintaining case-based reasoning (CBR) structure makes it more critical and increases the time complexity considerably when the dataset is large and complex. Consequently, distinct available techniques related to the maintenance of the CBR system concentrate on deleting the less important and very complex case data to reduce the number of cases. At the same time, it reduces the efficiency and effectiveness of handling the CBR system. Hence, this research employs an integrated platform with the combination of the Hadoop parallel platform and an intelligent utility scoring mechanism. The smart indexing system and the parallel processing capacity of Hadoop reduce the time complexity for processing large and complex datasets. The distributed storage capability effectively manages the data repository and retrieval system for enormous case datasets without compromising case data. The intelligent utility scoring system employs supervised learning and a unique CBR system to quickly and efficiently retrieve the effective solution for each particular case. Furthermore, Hadoop offers a distributed structure that helps the user to access the case data and update the database through the network. This research uses distinct healthcare text-based datasets for testing the performances of the proposed integrated technique in different aspects. The experimental results show the superiorities in producing a higher percentage of accuracy with consuming less Retrieval time over other related techniques. It further shows its time efficiency by offering higher throughput and lower Cyclomatic Complexity with lower read and write Latency Times. Finally, the performances of the proposed technique in these various aspects have been compared with distinct existing techniques and the performances of various databases in managing large and complex medical data to establish its superiority over them.

DAAC: Digital Asset Access Control in a Unified Blockchain Based E-Health System

The use of the Internet of Things and modern technologies has boosted the expansion of e-health solutions significantly and allowed access to better health services and remote monitoring of patients. Every service provider usually implements its information system to manage and access patient data for its unique purpose. Hence, the interoperability among independent e-health service providers is still a major challenge. From the structure of stored data to its large volume, the design of each such big data system varies, hence the cooperation among different e-health systems is almost impossible. In addition to this, the security and privacy of patient information is a challenging task. Building a unified solution for all creates significant business and economic issues. In this article, we present a solution to migrate existing e-health systems to a unified Blockchain-based model, where access to large scale medical data of patients can be achieved seamlessly by any service provider. A core blockchain network connects individual & independent e-health systems without requiring them to modify their internal processes. Access to patient data in the form of digital assets stored in off-chain storage is controlled through patient-centric channels and policy transactions. Through emulation, we show that the proposed solution can interconnect different e-health systems efficiently.

Application of Adaptive Synthetic Nominal and Extreme Gradient Boosting Methods in Determining Factors Affecting Obesity: A Case Study of Indonesian Basic Health Research Survey 2013

Obesity is the accumulation of excessive body fat and can be harmful to health. According to recent studies, several factors that contribute to the increasing prevalence of obesity in Indonesia include poor diet, lack of consumption of vegetables and fruits, high consumption of fast food, area of residence, and lack of physical activity. In addition, psychological factors, high consumption of alcohol and cigarettes, cultural differences, and stress factors also trigger obesity. The rapid development of the medical field cannot be separated from the availability of data that is increasingly easy to access and increasing knowledge in the medical field. This makes machine learning increasingly needed for pattern recognition from very large medical data, including obesity data. In this study, the factors that influence obesity status in Indonesia will be determined. In order to achieve this, Extreme Gradient Boosting (XGBoost) was used. This method is one of the classification methods that has better scalability and more efficient over its previous methods. Besides that, to overcome the imbalanced data, Adaptive Synthetic Nominal Algorithm (ADASYN-N) is used in order to balance the data and improve its prediction accuracy. Both the ADASYN-N and XGBoost methods will be applied to obesity data from the Indonesian Basic Health Research Survey in 2013. This study shows that female is more at risk in determining obesity status in Indonesia based on the highest gain value (37%). In addition, age 35-54 years, strenuous activity, and eating vegetables for 6 days are also risk factors of obesity.

Trends in strong opioid prescription for cancer patients in Japan from 2010 to 2019: an analysis with large medical claims data

Consumption of opioids, essential drugs for pain relief, has seen rapid growth worldwide. In Japan, where total opioid consumption still remains low among developed countries, little is known about trends in the clinical patterns of opioids in terminally ill cancer patients. Patients who died of cancer from 2010 to 2019 were included in this study. Morphine, oxycodone, fentanyl, tapentadol, methadone and hydromorphone were examined as opioids for cancer pain. We calculated the prevalence of prescribed opioids prior to death by year and age group and the average opioid dose 30days before death. The total number of patients was 221 598. We found that the prescription prevalence of opioids increased from 60.8 to 65.9% (5.1%). Morphine was most prescribed in 2010 but had decreased prevalence (-9.0%) during the 10-year period. Oxycodone had the highest increase in prescription prevalence (13.7%), and fentanyl prevalence decreased (-4.9%). In the subgroup comparison, the prescription prevalence of opioids in the elderly was lower than that in the younger group; however, the increasing trend in the elderly was greater than that in the younger group. The percentage of patients prescribed low-dose opioids (<60mg/day) during the 30days before death increased by 4.9% and was the highest throughout the study period. The prevalence of opioid prescriptions for terminally ill cancer patients has increased from 2010 to 2019 in Japan. The opioid-specific trends were similar to the global trend but differed by palliative care specialty.

Major areas of interest of artificial intelligence research applied to health care administrative data: a scoping review.

Introduction: The ongoing collection of large medical data has created conditions for application of artificial intelligence (AI) in research. This scoping review aimed to identify major areas of interest of AI applied to health care administrative data. Methods: The search was performed in seven databases: Medline, Embase, CINAHL, Web of science, IEEE, ICM digital library, and Compendex. We included articles published between January 2001 and March 2021, that described research with AI applied to medical diagnostics, pharmacotherapy, and health outcomes data. We screened the full text content and used natural language processing to automatically extract health areas of interest, principal AI methods, and names of medications. Results: Out of 14,864 articles, 343 were included. We determined ten areas of interest, the most common being health diagnostic or treatment outcome prediction (32%); representation of medical data, clinical pathways, and data temporality (i.e., transformation of raw medical data into compact and analysis-friendly format) (22%); and adverse drug effects, drug-drug interactions, and medication cascades (15%). Less attention has been devoted to areas such as health effects of polypharmacy (1%); and reinforcement learning (1%). The most common AI methods were decision trees, cluster analysis, random forests, and support vector machines. Most frequently mentioned medications included insulin, metformin, vitamins, acetaminophen, and heparin. Conclusions: The scoping review revealed the potential of AI application to health-related studies. However, several areas of interest in pharmacoepidemiology are sparsely reported, and the lack of details in studies related to pharmacotherapy suggests that AI could be used more optimally in pharmacoepidemiologic research.

Data-Efficient Vision Transformers for Multi-Label Disease Classification on Chest Radiographs

Abstract Radiographs are a versatile diagnostic tool for the detection and assessment of pathologies, for treatment planning or for navigation and localization purposes in clinical interventions. However, their interpretation and assessment by radiologists can be tedious and error-prone. Thus, a wide variety of deep learning methods have been proposed to support radiologists interpreting radiographs. Mostly, these approaches rely on convolutional neural networks (CNN) to extract features from images. Especially for the multi-label classification of pathologies on chest radiographs (Chest X-Rays, CXR), CNNs have proven to be well suited. On the Contrary, Vision Transformers (ViTs) have not been applied to this task despite their high classification performance on generic images and interpretable local saliency maps which could add value to clinical interventions. ViTs do not rely on convolutions but on patch-based self-attention and in contrast to CNNs, no prior knowledge of local connectivity is present. While this leads to increased capacity, ViTs typically require an excessive amount of training data which represents a hurdle in the medical domain as high costs are associated with collecting large medical data sets. In this work, we systematically compare the classification performance of ViTs and CNNs for different data set sizes and evaluate more data-efficient ViT variants (DeiT). Our results show that while the performance between ViTs and CNNs is on par with a small benefit for ViTs, DeiTs outperform the former if a reasonably large data set is available for training.

Attentive Hierarchical ANFIS with interpretability for cancer diagnostic

Prediction of clinical outcomes using the patient’s medical data enhances clinical decision making and improves prognostic accuracy. Deep learning (DL) for medical decision support systems has particularly shown expert-level accuracy in predicting clinical outcomes. However, most of these machine learning and Artificial Intelligent (AI) models lack interpretability which causes major trust-related problems in healthcare. This necessitates the need for interpretable AI systems that can explain their decisions. This paper is designed to address the problem of how clinical decision support systems can be designed to be transparent, interpretable and therefore comprehensible for humans. In this view, this paper proposes an attentive hierarchical adaptive neuro-fuzzy inference system (AH-ANFIS), that combines fuzzy inference in a hierarchical architecture with attention for selecting the important rules. The proposed model benefits from the rule-based structure of ANFIS that enables the user to interpret the abstractions of hidden layers. The hierarchical structure in fuzzy modeling helps to overcome the rule explosion problem that arises with large medical data and improve interpretability. To ensure the improvement of interpretability with hierarchical architecture, an evolutionary algorithm (EA) is used to decompose the input space into an optimal permutation of input subsets which results in subsystems with independent meaning. The attention-based rule selector identifies the most activated rule to select important patient-specific features for predicting the clinical outcome. To verify the performance of the proposed AH-ANFIS and analyze the important features for classification, we perform experiments with two cancer diagnostic datasets. By pruning fuzzy if-then rules using recursive rule elimination (RRE), the complexity of the model is largely reduced while maintaining the performance of the system making it more interpretable.