RESUMEN

Alzheimer's disease (AD) is affecting a growing number of individuals. As a result, there is a pressing need for accurate and early diagnosis methods. This study aims to achieve this goal by developing an optimal data analysis strategy to enhance computational diagnosis. Although various modalities of AD diagnostic data are collected, past research on computational methods of AD diagnosis has mainly focused on using single-modal inputs. We hypothesize that integrating, or "fusing," various data modalities as inputs to prediction models could enhance diagnostic accuracy by offering a more comprehensive view of an individual's health profile. However, a potential challenge arises as this fusion of multiple modalities may result in significantly higher dimensional data. We hypothesize that employing suitable dimensionality reduction methods across heterogeneous modalities would not only help diagnosis models extract latent information but also enhance accuracy. Therefore, it is imperative to identify optimal strategies for both data fusion and dimensionality reduction. In this paper, we have conducted a comprehensive comparison of over 80 statistical machine learning methods, considering various classifiers, dimensionality reduction techniques, and data fusion strategies to assess our hypotheses. Specifically, we have explored three primary strategies: (1) Simple data fusion, which involves straightforward concatenation (fusion) of datasets before inputting them into a classifier; (2) Early data fusion, in which datasets are concatenated first, and then a dimensionality reduction technique is applied before feeding the resulting data into a classifier; and (3) Intermediate data fusion, in which dimensionality reduction methods are applied individually to each dataset before concatenating them to construct a classifier. For dimensionality reduction, we have explored several commonly-used techniques such as principal component analysis (PCA), autoencoder (AE), and LASSO. Additionally, we have implemented a new dimensionality-reduction method called the supervised encoder (SE), which involves slight modifications to standard deep neural networks. Our results show that SE substantially improves prediction accuracy compared to PCA, AE, and LASSO, especially in combination with intermediate fusion for multiclass diagnosis prediction.

RESUMEN

Diagnosis prediction, a key factor in enhancing healthcare efficiency, remains a focal point in clinical decision support research. However, the time-series, sparse and multi-noise characteristics of electronic health record (EHR) data make it a great challenge. Existing methods commonly address these issues using RNNs and incorporating medical prior knowledge from medical knowledge bases, but they neglect the local spatial characteristics and spatial-temporal correlation of the data. Consequently, we propose MDPG, a diagnosis prediction model based on patient knowledge graphs. Initially, we represent the electronic visit records of patients as a patient-centered temporal knowledge graph, capturing the local spatial structure and temporal characteristics of the visit information. Subsequently, we design the spatial graph convolution block, temporal self-attention block, and spatial-temporal synchronous graph convolution block to capture the spatial, temporal, and spatial-temporal correlations embedded in them, respectively. Ultimately, we accomplish the prediction of patients' future states through multi-label classification. We conduct comprehensive experiments on two real-world datasets independently and evaluate the results using visit-level precision@k and code-level accuracy@k metrics. The experimental results demonstrate that MDPG outperforms all baseline models, yielding the best performance.

RESUMEN

General practitioners are supposed to be better diagnostics to detect patients with serious diseases earlier, and conduct early interventions and appropriate referrals of patients. However, in the current general practice, primary general practitioners lack sufficient clinical experiences, and the correct rate of general disease diagnosis is low. To assist general practitioners in diagnosis, this paper proposes a multi-label hierarchical classification method based on graph neural network, which integrates medical knowledge and electronic health record (EHR) data to build a disease prediction model. The experimental results based on data consist of 231,783 visits from EHR show that the proposed model outperforms all baseline models in the general disease prediction task with a top-3 recall of 0.865. The interpretable results of the model can effectively help clinicians understand the basis of the model's decision-making.

Asunto(s)

RESUMEN

PURPOSE: There are currently more than 480 primary immune deficiency (PID) diseases and about 7000 rare diseases that together afflict around 1 in every 17 humans. Computational aids based on data mining and machine learning might facilitate the diagnostic task by extracting rules from large datasets and making predictions when faced with new problem cases. In a proof-of-concept data mining study, we aimed to predict PID diagnoses using a supervised machine learning algorithm based on classification tree boosting. METHODS: Through a data query at the USIDNET registry we obtained a database of 2396 patients with common diagnoses of PID, including their clinical and laboratory features. We kept 286 features and all 12 diagnoses to include in the model. We used the XGBoost package with parallel tree boosting for the supervised classification model, and SHAP for variable importance interpretation, on Python v3.7. The patient database was split into training and testing subsets, and after boosting through gradient descent, the predictive model provides measures of diagnostic prediction accuracy and individual feature importance. After a baseline performance test, we used the Class Weighting Hyperparameter, or scale_pos_weight to correct for imbalanced classification. RESULTS: The twelve PID diagnoses were CVID (1098 patients), DiGeorge syndrome, Chronic granulomatous disease, Congenital agammaglobulinemia, PID not otherwise classified, Specific antibody deficiency, Complement deficiency, Hyper-IgM, Leukocyte adhesion deficiency, ectodermal dysplasia with immune deficiency, Severe combined immune deficiency, and Wiskott-Aldrich syndrome. For CVID, the model found an accuracy on the train sample of 0.80, with an area under the ROC curve (AUC) of 0.80, and a Gini coefficient of 0.60. In the test subset, accuracy was 0.76, AUC 0.75, and Gini 0.51. The positive feature value to predict CVID was highest for upper respiratory infections, asthma, autoimmunity and hypogammaglobulinemia. Features with the highest negative predictive value were high IgE, growth delay, abscess, lymphopenia, and congenital heart disease. For the rest of the diagnoses, accuracy stayed between 0.75 and 0.99, AUC 0.46-0.87, Gini 0.07-0.75, and LogLoss 0.09-8.55. DISCUSSION: Clinicians should remember to consider the negative predictive features together with the positives. We are calling this a proof-of-concept study to continue with our explorations. A good performance is encouraging, and feature importance might aid feature selection for future endeavors. In the meantime, we can learn from the rules derived by the model and build a user-friendly decision tree to generate differential diagnoses.

Asunto(s)

RESUMEN

The ability to detect cancer at an early stage in patients who would benefit from effective therapy is a key factor in increasing survivability. This work proposes an evolutionary supervised learning method called CancerSig to identify cancer stage-specific microRNA (miRNA) signatures for early cancer predictions. CancerSig established a compact panel of miRNA signatures as potential markers from 4,667 patients with 15 different types of cancers for the cancer stage prediction, and achieved a mean performance: 10-fold cross-validation accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve of 84.27% ± 6.31%, 0.81 ± 0.12, 0.80 ± 0.10, and 0.80 ± 0.06, respectively. The pan-cancer analysis of miRNA signatures suggested that three miRNAs, hsa-let-7i-3p, hsa-miR-362-3p, and hsa-miR-3651, contributed significantly toward stage prediction across 8 cancers, and each of the 67 miRNAs of the panel was a biomarker of stage prediction in more than one cancer. CancerSig may serve as the basis for cancer screening and therapeutic selection..

Asunto(s)

RESUMEN

The Internet of Things (IoT) has been influential in predicting major diseases in current practice. The deep learning (DL) technique is vital in monitoring and controlling the functioning of the healthcare system and ensuring an effective decision-making process. In this study, we aimed to develop a framework implementing the IoT and DL to identify lung cancer. The accurate and efficient prediction of disease is a challenging task. The proposed model deploys a DL process with a multi-layered non-local Bayes (NL Bayes) model to manage the process of early diagnosis. The Internet of Medical Things (IoMT) could be useful in determining factors that could enable the effective sorting of quality values through the use of sensors and image processing techniques. We studied the proposed model by analyzing its results with regard to specific attributes such as accuracy, quality, and system process efficiency. In this study, we aimed to overcome problems in the existing process through the practical results of a computational comparison process. The proposed model provided a low error rate (2%, 5%) and an increase in the number of instance values. The experimental results led us to conclude that the proposed model can make predictions based on images with high sensitivity and better precision values compared to other specific results. The proposed model achieved the expected accuracy (81%, 95%), the expected specificity (80%, 98%), and the expected sensitivity (80%, 99%). This model is adequate for real-time health monitoring systems in the prediction of lung cancer and can enable effective decision-making with the use of DL techniques.

RESUMEN

Purpose/Significance The recent applications of machine learning in epilepsy seizure prediction,diagnosis prediction,seizure detection,efficacy prediction of antiepileptic drugs,and epilepsy surgery prediction are summarized and analyzed.Method/Processs Literatures are searched through PubMed to summarize the performance of each machine learning model and the challenges exist-ing in machine learning technology.Result/Conclusion Machine learning plays an important role in the diagnosis and treatment of epi-lepsy,and can provide reference for clinical doctors'diagnosis and treatment work.

RESUMEN

Chronic diseases have become an important public health problem for people under 70 years of age worldwide， while also causing a great economic burden. The establishment of clinical prediction models can help to predict the risk of a disease or the prognostic effect of a study subject in advance by means of index testing at the early stage of chronic diseases， and plays an increasingly important role in clinical practice. This study introduces clinical diagnostic prediction models and clinical prognostic prediction models， and reviews clinical data processing， clinical prediction model building， visualization methods and model evaluation from the perspective of the application of clinical prediction models， which contribute to the correct and reasonable use of prediction models in clinical research.

RESUMEN

Electronic health records naturally contain most of the medical information in the form of doctor's notes as unstructured or semi-structured texts. Current deep learning text analysis approaches allow researchers to reveal the inner semantics of text information and even identify hidden consequences that can offer extra decision support to doctors. In the presented article, we offer a new automated analysis of Polish summary texts of patient hospitalizations. The presented models were found to be able to predict the final diagnosis with almost 70% accuracy based just on the patient's medical history (only 132 words on average), with possible accuracy increases when adding further sentences from hospitalization results; even one sentence was found to improve the results by 4%, and the best accuracy of 78% was achieved with five extra sentences. In addition to detailed descriptions of the data and methodology, we present an evaluation of the analysis using more than 50,000 Polish cardiology patient texts and dive into a detailed error analysis of the approach. The results indicate that the deep analysis of just the medical history summary can suggest the direction of diagnosis with a high probability that can be further increased just by supplementing the records with further examination results.

RESUMEN

WD is caused by ATP7B variants disrupting copper efflux resulting in excessive copper accumulation mainly in liver and brain. The diagnosis of WD is challenged by its variable clinical course, onset, morbidity, and ATP7B variant type. Currently it is diagnosed by a combination of clinical symptoms/signs, aberrant copper metabolism parameters (e.g., low ceruloplasmin serum levels and high urinary and hepatic copper concentrations), and genetic evidence of ATP7B mutations when available. As early diagnosis and treatment are key to favorable outcomes, it is critical to identify subjects before the onset of overtly detrimental clinical manifestations. To this end, we sought to improve WD diagnosis using artificial neural network algorithms (part of artificial intelligence) by integrating available clinical and molecular parameters. Surprisingly, WD diagnosis was based on plasma levels of glutamate, asparagine, taurine, and Fischer's ratio. As these amino acids are linked to the urea-Krebs' cycles, our study not only underscores the central role of hepatic mitochondria in WD pathology but also that most WD patients have underlying hepatic dysfunction. Our study provides novel evidence that artificial intelligence utilized for integrated analysis for WD may result in earlier diagnosis and mechanistically relevant treatments for patients with WD.

Asunto(s)

RESUMEN

BACKGROUND: Early detection of developmental disabilities in children is essential because early intervention can improve the prognosis of children. Meanwhile, a growing body of evidence has indicated a relationship between developmental disability and motor skill, and thus, motor skill is considered in the early diagnosis of developmental disability. However, there are challenges to assessing motor skill in the diagnosis of developmental disorder, such as a lack of specialists and time constraints, and thus it is commonly conducted through informal questions or surveys to parents. OBJECTIVE: This study sought to evaluate the possibility of using drag-and-drop data as a digital biomarker and to develop a classification model based on drag-and-drop data with which to classify children with developmental disabilities. METHODS: We collected drag-and-drop data from children with typical development and developmental disabilities from May 1, 2018, to May 1, 2020, via a mobile application (DoBrain). We used touch coordinates and extracted kinetic variables from these coordinates. A deep learning algorithm was developed to predict potential development disabilities in children. For interpretability of the model results, we identified which coordinates contributed to the classification results by applying gradient-weighted class activation mapping. RESULTS: Of the 370 children in the study, 223 had typical development, and 147 had developmental disabilities. In all games, the number of changes in the acceleration sign based on the direction of progress both in the x- and y-axes showed significant differences between the 2 groups (P<.001; effect size >0.5). The deep learning convolutional neural network model showed that drag-and-drop data can help diagnose developmental disabilities, with an area under the receiving operating characteristics curve of 0.817. A gradient class activation map, which can interpret the results of a deep learning model, was visualized with the game results for specific children. CONCLUSIONS: Through the results of the deep learning model, we confirmed that drag-and-drop data can be a new digital biomarker for the diagnosis of developmental disabilities.

RESUMEN

Background and Aims-Transforming growth factor-beta (TGF-ß) signaling orchestrates tumorigenesis and one of the family members, TGF-ß receptor type III (TGFßR3), are distinctively under-expressed in numerous malignancies. Currently, the clinical impact of TGFßR3 down-regulation and the underlying mechanism remains unclear in hepatocellular carcinoma (HCC). Here, we aimed to identify the tumor-promoting roles of decreased TGFßR3 expression in HCC progression. Materials and Methods-For clinical analysis, plasma and liver specimens were collected from 100 HCC patients who underwent curative resection for the quantification of TGFßR3 by q-PCR and ELISA. To study the tumor-promoting mechanism of TGFßR3 downregulation, HCC mouse models and TGFßR3 knockout cell lines were applied. Results-Significant downregulation of TGFßR3 and its soluble form (sTGFßR3) were found in HCC tissues and plasma compared to healthy individuals (p < 0.01). Patients with <9.4 ng/mL sTGFßR3 exhibited advanced tumor stage, higher recurrence rate and shorter disease-free survival (p < 0.05). The tumor-suppressive function of sTGFßR3 was further revealed in an orthotopic mouse HCC model, resulting in 2-fold tumor volume reduction. In TGFßR3 knockout hepatocyte and HCC cells, increased complement component C5a was observed and strongly correlated with shorter survival and advanced tumor stage (p < 0.01). Interestingly, C5a activated the tumor-promoting Th-17 response in tumor associated macrophages. Conclusion-TGFßR3 suppressed tumor progression, and decreased expression resulted in poor prognosis in HCC patients through upregulation of tumor-promoting complement C5a.

RESUMEN

OBJECTIVES: We aimed to develop machine learning models and evaluate their performance in predicting HIV and sexually transmitted infections (STIs) diagnosis based on a cohort of Australian men who have sex with men (MSM). METHODS: We collected clinical records of 21,273 Australian MSM during 2011-2017. We compared accuracies for predicting HIV and STIs (syphilis, gonorrhoea, chlamydia) diagnosis using four machine learning approaches against a multivariable logistic regression (MLR) model. RESULTS: Machine learning approaches consistently outperformed MLR. Gradient boosting machine (GBM) achieved the highest area under the receiver operator characteristic curve for HIV (76.3%) and STIs (syphilis, 85.8%; gonorrhoea, 75.5%; chlamydia, 68.0%), followed by extreme gradient boosting (71.1%, 82.2%, 70.3%, 66.4%), random forest (72.0%, 81.9%, 67.2%, 64.3%), deep learning (75.8%, 81.0%, 67.5%, 65.4%) and MLR (69.8%, 80.1%, 67.2%, 63.2%). GBM models demonstrated the ten greatest predictors collectively explained 62.7-73.6% of variations in predicting HIV/STIs. STIs symptoms, past syphilis infection, age, time living in Australia, frequency of condom use with casual male sexual partners during receptive anal sex and the number of casual male sexual partners in the past 12 months were most commonly identified predictors. CONCLUSIONS: Machine learning approaches are advantageous over multivariable logistic regression models in predicting HIV/STIs diagnosis.

Asunto(s)

RESUMEN

The onset of liver cancer is insidious. Currently, there is no effective method for the early detection of hepatocellular carcinoma (HCC). Transcriptomic profiles of 826 tissue samples from the Gene Expression Omnibus (GEO), The Cancer Genome Atlas (TCGA), Genotype tissue expression (GTEx), and International Cancer Genome Consortium (ICGC) databases were utilized to establish models for early detection and surveillance of HCC. The overlapping differentially expressed genes (DEGs) were screened by elastic net and robust rank aggregation (RRA) analyses to construct the diagnostic prediction model for early HCC (DP.eHCC). Prognostic prediction genes were screened by univariate cox regression and lasso cox regression analyses to construct the survival risk prediction model for early HCC (SP.eHCC). The relationship between the variation of transcriptome profile and the oncogenic risk-score of early HCC was analyzed by combining Weighted Correlation Network Analysis (WGCNA), Gene Set Enrichment Analysis (GSEA), and genome networks (GeNets). The results showed that the AUC of DP.eHCC model for the diagnosis of early HCC was 0.956 (95% CI: 0.941-0.972; p < 0.001) with a sensitivity of 90.91%, a specificity of 92.97%. The SP.eHCC model performed well for predicting the overall survival risk of HCC patients (HR = 10.79; 95% CI: 6.16-18.89; p < 0.001). The oncogenesis of early HCC was revealed mainly involving in pathways associated with cell proliferation and tumor microenvironment. And the transcription factors including EZH2, EGR1, and SOX17 were screened in the genome networks as the promising targets used for precise treatment in patients with HCC. Our findings provide robust models for the early diagnosis and prognosis of HCC, and are crucial for the development of novel targets applied in the precision therapy of HCC.

RESUMEN

Electronic health records (EHRs) contain important temporal information about the progression of disease and treatment outcomes. This paper proposes a transitive sequencing approach for constructing temporal representations from EHR observations for downstream machine learning. Using clinical data from a cohort of patients with congestive heart failure, we mined temporal representations by transitive sequencing of EHR medication and diagnosis records for classification and prediction tasks. We compared the classification and prediction performances of the transitive sequential representations (bag-of-sequences approach) with the conventional approach of using aggregated vectors of EHR data (aggregated vector representation) across different classifiers. We found that the transitive sequential representations are better phenotype "differentiators" and predictors than the "atemporal" EHR records. Our results also demonstrated that data representations obtained from transitive sequencing of EHR observations can present novel insights about the progression of the disease that are difficult to discern when clinical data are treated independently of the patient's history.

RESUMEN

BACKGROUND: Diagnosis aims to predict the future health status of patients according to their historical electronic health records (EHR), which is an important yet challenging task in healthcare informatics. Existing diagnosis prediction approaches mainly employ recurrent neural networks (RNN) with attention mechanisms to make predictions. However, these approaches ignore the importance of code descriptions, i.e., the medical definitions of diagnosis codes. We believe that taking diagnosis code descriptions into account can help the state-of-the-art models not only to learn meaning code representations, but also to improve the predictive performance, especially when the EHR data are insufficient. METHODS: We propose a simple, but general diagnosis prediction framework, which includes two basic components: diagnosis code embedding and predictive model. To learn the interpretable code embeddings, we apply convolutional neural networks (CNN) to model medical descriptions of diagnosis codes extracted from online medical websites. The learned medical embedding matrix is used to embed the input visits into vector representations, which are fed into the predictive models. Any existing diagnosis prediction approach (referred to as the base model) can be cast into the proposed framework as the predictive model (called the enhanced model). RESULTS: We conduct experiments on two real medical datasets: the MIMIC-III dataset and the Heart Failure claim dataset. Experimental results show that the enhanced diagnosis prediction approaches significantly improve the prediction performance. Moreover, we validate the effectiveness of the proposed framework with insufficient EHR data. Finally, we visualize the learned medical code embeddings to show the interpretability of the proposed framework. CONCLUSIONS: Given the historical visit records of a patient, the proposed framework is able to predict the next visit information by incorporating medical code descriptions.

Asunto(s)

RESUMEN

Alien limb phenomenon is a rare syndrome associated with a feeling of non-belonging and disowning toward one's limb. In contrast, anarchic limb phenomenon leads to involuntary but goal-directed movements. Alien/anarchic limb phenomena are frequent in corticobasal syndrome (CBS), an atypical parkinsonian syndrome characterized by rigidity, akinesia, dystonia, cortical sensory deficit, and apraxia. The structure-function relationship of alien/anarchic limb was investigated in multi-centric structural magnetic resonance imaging (MRI) data. Whole-group and single-subject comparisons were made in 25 CBS and eight CBS-alien/anarchic limb patients versus controls. Support vector machine was used to see if CBS with and without alien/anarchic limb could be distinguished by structural MRI patterns. Whole-group comparison of CBS versus controls revealed asymmetric frontotemporal atrophy. CBS with alien/anarchic limb syndrome versus controls showed frontoparietal atrophy including the supplementary motor area contralateral to the side of the affected limb. Exploratory analysis identified frontotemporal regions encompassing the pre-/and postcentral gyrus as compromised in CBS with alien limb syndrome. Classification of CBS patients yielded accuracies of 79%. CBS-alien/anarchic limb syndrome was differentiated from CBS patients with an accuracy of 81%. Predictive differences were found in the cingulate gyrus spreading to frontomedian cortex, postcentral gyrus, and temporoparietoocipital regions. We present the first MRI-based group analysis on CBS-alien/anarchic limb. Results pave the way for individual clinical syndrome prediction and allow understanding the underlying neurocognitive architecture.

Asunto(s)

RESUMEN

The problems a patient suffers from can be summarized in terms of a list of patient diagnoses. The diagnoses are typically organized in a hierarchy (or a lattice structure) in which many different low-level diagnoses are covered by one or more diagnostic categories. An interesting machine learning problem is related to learning of a wide range of diagnostic models (at different levels of abstraction) that can automatically assign a diagnosis or a diagnostic category to a specific patient. While one can always approach this problem by learning models for each diagnostic task independently, an interesting open question is how one can leverage the knowledge of a diagnostic hierarchy to improve the classification and outperform independent diagnostic models. In this work, we study this problem by designing a new hierarchical classification learning framework in which multiple diagnostic classification targets are explicitly related via diagnostic hierarchy relations. By conducting experiments on MIMIC-III data and ICD-9 diagnosis hierarchy, we demonstrate that our framework leads to improved classification performance on individual diagnostic tasks when compared to independently learned diagnostic models. This improvement is stronger for diagnoses with a low prior and smaller number of positive training examples.

RESUMEN

Information and communications technologies have enabled healthcare institutions to accumulate large amounts of healthcare data that include diagnoses, medications, and additional contextual information such as patient demographics. To gain a better understanding of big healthcare data and to develop better data-driven clinical decision support systems, we propose a novel multiple-channel latent Dirichlet allocation (MCLDA) approach for modeling diagnoses, medications, and contextual information in healthcare data. The proposed MCLDA model assumes that a latent health status group structure is responsible for the observed co-occurrences among diagnoses, medications, and contextual information. Using a real-world research testbed that includes one million healthcare insurance claim records, we investigate the utility of MCLDA. Our empirical evaluation results suggest that MCLDA is capable of capturing the comorbidity structures and linking them with the distribution of medications. Moreover, MCLDA is able to identify the pairing between diagnoses and medications in a record based on the assigned latent groups. MCLDA can also be employed to predict missing medications or diagnoses given partial records. Our evaluation results also show that, in most cases, MCLDA outperforms alternative methods such as logistic regressions and the k-nearest-neighbor (KNN) model for two prediction tasks, i.e., medication and diagnosis prediction. Thus, MCLDA represents a promising approach to modeling healthcare data for clinical decision support.

Asunto(s)

RESUMEN

Obstructive sleep apnea hypopnea syndrome (OSAHS) is a chronic disorder characterized by repetitive apneas, oxygen desaturation and disruption during sleep. The main clinical manifestations of OSAHS are snores, intermittent hypoxia, morning headache, excessive daytime sleepiness, tired and memory descent. OSAHS affects 3%-20% of the general population. Although commonly accepted as the gold standard for diagnosis of OSAHS, PSG is thought to be time-consuming, labor-intensive, costly and uncomfortable. Evidence from epidemiology indicated that 93% of women and 82% of men with moderate to severe OSAHS have not been clinically diagnosed. An epidemiological survey in Shanghai, China, showed that more than 85% of OSAHS was undiagnosed. Studies showed that undiagnosed OSAHS is independently associated with the increased likelihood of hypertension, cardiovascular diseases, stroke, daytime sleepiness, motor vehicle accidents, and diminished quality of life. Thus, researchers have attempted to develop a simple and effective tool to screen patients with OSAHS, which has been reviewed and summarized in our article.

Asunto(s)