PMCC PMCC

Search tips
Search criteria

Advanced
Results 1-5 (5)
 

Clipboard (0)
None
Journals
Authors
more »
Year of Publication
Document Types
1.  Machine learning to predict extubation outcome in premature infants 
Though treatment of the ventilated premature infant has experienced many advances over the past decades, determining the best time point for extubation of these infants remains challenging and the incidence of extubation failures largely unchanged. The objective was to provide clinicians with a decision-support tool to determine whether to extubate a mechanically ventilated premature infant by using a set of machine learning algorithms on a dataset assembled from 486 premature infants receiving mechanical ventilation.
Algorithms included artificial neural networks (ANN), support vector machine (SVM), naïve Bayesian classifier (NBC), boosted decision trees (BDT), and multivariable logistic regression (MLR). Results for ANN, MLR, and NBC were satisfactory (area under the curve [AUC]: 0.63–0.76); however, SVM and BDT consistently showed poor performance (AUC ~0.5).
Complex medical data such as the data set used for this study require further preprocessing steps before prediction models can be developed that achieve similar or better performance than clinicians.
doi:10.1109/IJCNN.2013.6707058
PMCID: PMC4255563  PMID: 25485175
2.  Can Machine Learning Methods Predict Extubation Outcome in Premature Infants as well as Clinicians? 
Journal of neonatal biology  2013;2:1000118.
Rationale
Though treatment of the prematurely born infant breathing with assistance of a mechanical ventilator has much advanced in the past decades, predicting extubation outcome at a given point in time remains challenging. Numerous studies have been conducted to identify predictors for extubation outcome; however, the rate of infants failing extubation attempts has not declined.
Objective
To develop a decision-support tool for the prediction of extubation outcome in premature infants using a set of machine learning algorithms
Methods
A dataset assembled from 486 premature infants on mechanical ventilation was used to develop predictive models using machine learning algorithms such as artificial neural networks (ANN), support vector machine (SVM), naïve Bayesian classifier (NBC), boosted decision trees (BDT), and multivariable logistic regression (MLR). Performance of all models was evaluated using area under the curve (AUC).
Results
For some of the models (ANN, MLR and NBC) results were satisfactory (AUC: 0.63–0.76); however, two algorithms (SVM and BDT) showed poor performance with AUCs of ~0.5.
Conclusion
Clinician's predictions still outperform machine learning due to the complexity of the data and contextual information that may not be captured in clinical data used as input for the development of the machine learning algorithms. Inclusion of preprocessing steps in future studies may improve the performance of prediction models.
doi:10.4172/2167-0897.1000118
PMCID: PMC4238927  PMID: 25419493
Premature infant; mechanical ventilation; extubation; prediction; machine learning
3.  Identification of Diagnostic Urinary Biomarkers for Acute Kidney Injury 
Acute kidney injury (AKI) is an important cause of death among hospitalized patients. The two most common causes of AKI are acute tubular necrosis (ATN) and prerenal azotemia (PRA). Appropriate diagnosis of the disease is important but often difficult. We analyzed urine proteins by 2-DE from 38 patients with AKI. Patients were randomly assigned to a training set, an internal test set or an external validation set. Spot abundances were analyzed by artificial neural networks (ANN) to identify biomarkers which differentiate between ATN and PRA. When the trained neural network algorithm was tested against the training data it identified the diagnosis for 16/18 patients in the training set and all 10 patients in the internal test set. The accuracy was validated in the novel external set of patients where 9/10 subjects were correctly diagnosed including 5/5 with ATN and 4/5 with PRA. Plasma retinol binding protein (PRBP) was identified in one spot and a fragment of albumin and PRBP in the other. These proteins are candidate markers for diagnostic assays of AKI.
doi:10.231/JIM.0b013e3181d473e7
PMCID: PMC2864920  PMID: 20224435
Acute kidney injury; Biomarkers; Diagnosis; Kidney; Urine
4.  Prediction of urinary protein markers in lupus nephritis 
Kidney international  2005;68(6):2588-2592.
Background
Lupus nephritis is divided into six classes and scored according to activity and chronicity indices based on histologic findings. Treatment differs based on the pathologic findings. Renal biopsy is currently the only way to accurately predict class and activity and chronicity indices. We propose to use patterns of abundance of urine proteins to identify class and disease indices.
Methods
Urine was collected from 20 consecutive patients immediately prior to biopsy for evaluation of lupus nephritis. The International Society of Nephrology/Renal Pathology Society (ISN/RPS) class of lupus nephritis, activity, and chronicity indices were determined by a renal pathologist. Proteins were separated by two-dimensional gel electrophoresis. Artificial neural networks were trained on normalized spot abundance values.
Results
Biopsy specimens were classified in the database according to ISN/RPS class, activity, and chronicity. Nine samples had characteristics of more than one class present. Receiver operating characteristic (ROC) curves of the trained networks demonstrated areas under the curve ranging from 0.85 to 0.95. The sensitivity and specificity for the ISN/RPS classes were class II 100%, 100%; III 86%, 100%; IV 100%, 92%; and V 92%, 50%. Activity and chronicity indices had r values of 0.77 and 0.87, respectively. A list of spots was obtained that provided diagnostic sensitivity to the analysis.
Conclusion
We have identified a list of protein spots that can be used to develop a clinical assay to predict ISN/RPS class and chronicity for patients with lupus nephritis. An assay based on antibodies against these spots could eliminate the need for renal biopsy, allow frequent evaluation of disease status, and begin specific therapy for patients with lupus nephritis.
doi:10.1111/j.1523-1755.2005.00730.x
PMCID: PMC2667626  PMID: 16316334
lupus nephritis; biomarkers; urine; electrophoresis; two-dimensional gel
5.  A Semantic Web Management Model for Integrative Biomedical Informatics 
PLoS ONE  2008;3(8):e2946.
Background
Data, data everywhere. The diversity and magnitude of the data generated in the Life Sciences defies automated articulation among complementary efforts. The additional need in this field for managing property and access permissions compounds the difficulty very significantly. This is particularly the case when the integration involves multiple domains and disciplines, even more so when it includes clinical and high throughput molecular data.
Methodology/Principal Findings
The emergence of Semantic Web technologies brings the promise of meaningful interoperation between data and analysis resources. In this report we identify a core model for biomedical Knowledge Engineering applications and demonstrate how this new technology can be used to weave a management model where multiple intertwined data structures can be hosted and managed by multiple authorities in a distributed management infrastructure. Specifically, the demonstration is performed by linking data sources associated with the Lung Cancer SPORE awarded to The University of Texas MDAnderson Cancer Center at Houston and the Southwestern Medical Center at Dallas. A software prototype, available with open source at www.s3db.org, was developed and its proposed design has been made publicly available as an open source instrument for shared, distributed data management.
Conclusions/Significance
The Semantic Web technologies have the potential to addresses the need for distributed and evolvable representations that are critical for systems Biology and translational biomedical research. As this technology is incorporated into application development we can expect that both general purpose productivity software and domain specific software installed on our personal computers will become increasingly integrated with the relevant remote resources. In this scenario, the acquisition of a new dataset should automatically trigger the delegation of its analysis.
doi:10.1371/journal.pone.0002946
PMCID: PMC2491554  PMID: 18698353

Results 1-5 (5)