Artif Intell Med.

Ensemble adaptive network-based fuzzy inference system with weighted arithmetical mean and application to diagnosis of optic nerve disease from visual-evoked potential signals.

Publication Date: 2008 May 9 PMID: 18468871
Authors: Akdemir, B. - Kara, S. - Polat, K. - Guven, A. - Gunes, S.
Journal: Artif Intell Med

OBJECTIVE: This paper presents a new method based on combining principal component analysis (PCA) and adaptive network-based fuzzy inference system (ANFIS) to diagnose the optic nerve disease from visual-evoked potential (VEP) signals. The aim of this study is to improve the classification accuracy of ANFIS classifier on diagnosis of optic nerve disease from VEP signals. With this aim, a new classifier ensemble based on ANFIS and PCA is proposed. METHODS AND MATERIAL: The VEP signals dataset include 61 healthy subjects and 68 patients suffered from optic nerve disease. First of all, the dimension of VEP signals dataset with 63 features has been reduced to 4 features using PCA. After applying PCA, ANFIS trained using three different training-testing datasets randomly with 50-50% training-testing partition. RESULTS: The obtained classification results from ANFIS trained separately with three different training-testing datasets are 96.87%, 98.43%, and 98.43%, respectively. And then the results of ANFIS trained with three different training-testing datasets randomly with 50-50% training-testing partition have been combined with three different ways including weighted arithmetical mean that proposed firstly by us, arithmetical mean, and geometrical mean. The classification results of ANFIS combined with three different ways are 98.43%, 100%, and 100%, respectively. Also, ensemble ANFIS has been compared with ANN ensemble. ANN ensemble obtained 98.43%, 100%, and 100% prediction accuracy with three different ways including arithmetical mean, geometrical mean and weighted arithmetical mean. CONCLUSION: These results have shown that the proposed classifier ensemble approach based on ANFIS trained with different train-test datasets and PCA has produced very promising results in the diagnosis of optic nerve disease from VEP signals.

post to: CiteULike

Neuro-genetic non-invasive temperature estimation: Intensity and spatial prediction.

Publication Date: 2008 May 9 PMID: 18468870
Authors: Teixeira, C. A. - Graca Ruano, M. - Ruano, A. E. - Pereira, W. C.
Journal: Artif Intell Med

OBJECTIVES: The existence of proper non-invasive temperature estimators is an essential aspect when thermal therapy applications are envisaged. These estimators must be good predictors to enable temperature estimation at different operational situations, providing better control of the therapeutic instrumentation. In this work, radial basis functions artificial neural networks were constructed to access temperature evolution on an ultrasound insonated medium. METHODS: The employed models were radial basis functions neural networks with external dynamics induced by their inputs. Both the most suited set of model inputs and number of neurons in the network were found using the multi-objective genetic algorithm. The neural models were validated in two situations: the operating ones, as used in the construction of the network; and in 11 unseen situations. The new data addressed two new spatial locations and a new intensity level, assessing the intensity and space prediction capacity of the proposed model. RESULTS: Good performance was obtained during the validation process both in terms of the spatial points considered and whenever the new intensity level was within the range of applied intensities. A maximum absolute error of 0.5 degrees C+/-10% (0.5 degrees C is the gold-standard threshold in hyperthermia/diathermia) was attained with low computationally complex models. CONCLUSION: The results confirm that the proposed neuro-genetic approach enables foreseeing temperature propagation, in connection to intensity and space parameters, thus enabling the assessment of different operating situations with proper temperature resolution.

post to: CiteULike

Complexity and spectral analysis of the heart rate variability dynamics for distant prediction of paroxysmal atrial fibrillation with artificial intelligence methods.

Publication Date: 2008 Apr 30 PMID: 18455375
Authors: Chesnokov, Y. V.
Journal: Artif Intell Med

OBJECTIVE: Paroxysmal atrial fibrillation (PAF) is a serious arrhythmia associated with morbidity and mortality. We explore the possibility of distant prediction of PAF by analyzing changes in heart rate variability (HRV) dynamics of non-PAF rhythms immediately before PAF event. We use that model for distant prognosis of PAF onset with artificial intelligence methods. METHODS AND MATERIALS: We analyzed 30-min non-PAF HRV records from 51 subjects immediately before PAF onset and at least 45min distant from any PAF event. We used spectral and complexity analysis with sample (SmEn) and approximate (ApEn) entropies and their multiscale versions on extracted HRV data. We used that features to train the artificial neural networks (ANNs) and support vector machine (SVM) classifiers to differentiate the subjects. The trained classifiers were further tested for distant PAF event prognosis on 16 subjects from independent database on non-PAF rhythm lasting from 60 to 320min before PAF onset classifying the 30-min segments as distant or leading to PAF. RESULTS: We found statistically significant increase in 30-min non-PAF HRV recordings from 51 subjects in the VLF, LF, HF bands and total power (p<0.0001) before PAF event compared to PAF distant ones. The SmEn and ApEn analysis provided significant decrease in complexity (p<0.0001 and p<0.001) before PAF onset. For training ANN and SVM classifiers the data from 51 subjects were randomly split to training, validation and testing. ANN provided better results in terms of sensitivity (Se), specificity (Sp) and positive predictivity (Pp) compared to SVM which became biased towards positive case. The validation results of the ANN classifier we achieved: Se 76%, Sp 93%, Pp 94%. Testing ANN and SVM classifiers on 16 subjects with non-PAF HRV data preceding PAF events we obtained distant prediction of PAF onset with SVM classifier in 10 subjects (58+/-18min in advance). ANN classifier provided distant prediction of PAF event in 13 subjects (62+/-21min in advance). CONCLUSION: From the results of distant PAF prediction we conclude that ANN and SVM classifiers learned the changes in the HRV dynamics immediately before PAF event and successfully identified them during distant PAF prognosis on independent database. This confirms the reported in the literature results that corresponding changes in the HRV data occur about 60min before PAF onset and proves the possibility of distant PAF prediction with ANN and SVM methods.

post to: CiteULike

Supervised pattern recognition for the prediction of contrast-enhancement appearance in brain tumors from multivariate magnetic resonance imaging and spectroscopy.

Publication Date: 2008 May PMID: 18448318
Authors: Lee, M. C. - Nelson, S. J.
Journal: Artif Intell Med

OBJECTIVE: The purpose of this study was to develop a pattern classification algorithm for use in predicting the location of new contrast-enhancement in brain tumor patients using data obtained via multivariate magnetic resonance (MR) imaging from a prior scan. We also explore the use of feature selection or weighting in improving the accuracy of the pattern classifier. METHODS AND MATERIALS: Contrast-enhanced MR images, perfusion images, diffusion images, and proton spectroscopic imaging data were obtained from 26 patients with glioblastoma multiforme brain tumors, divided into a design set and an unseen test set for verification of results. A k-NN algorithm was implemented to classify unknown data based on a set of training data with ground truth derived from post-treatment contrast-enhanced images; the quality of the k-NN results was evaluated using a leave-one-out cross-validation method. A genetic algorithm was implemented to select optimal features and feature weights for the k-NN algorithm. The binary representation of the weights was varied from 1 to 4bits. Each individual parameter was thresholded as a simple classification technique, and the results compared with the k-NN. RESULTS: The feature selection k-NN was able to achieve a sensitivity of 0.78+/-0.18 and specificity of 0.79+/-0.06 on the holdout test data using only 7 of the 38 original features. Similar results were obtained with non-binary weights, but using a larger number of features. Overfitting was also observed in the higher bit representations. The best single-variable classifier, based on a choline-to-NAA abnormality index computed from spectroscopic data, achieved a sensitivity of 0.79+/-0.20 and specificity of 0.71+/-0.11. The k-NN results had lower variation across patients than the single-variable classifiers. CONCLUSIONS: We have demonstrated that the optimized k-NN rule could be used for quantitative analysis of multivariate images, and be applied to a specific clinical research question. Selecting features was found to be useful in improving the accuracy of feature weighting algorithms and improving the comprehensibility of the results. We believe that in addition to lending insight into parameter relevance, such algorithms may be useful in aiding radiological interpretation of complex multimodality datasets.

post to: CiteULike

Evaluation of an architecture for intelligent query and exploration of time-oriented clinical data.

Publication Date: 2008 May PMID: 18442899
Authors: Martins, S. B. - Shahar, Y. - Goren-Bar, D. - Galperin, M. - Kaizer, H. - Basso, L. V. - McNaughton, D. - Goldstein, M. K.
Journal: Artif Intell Med

OBJECTIVE: Evaluate KNAVE-II, a knowledge-based framework for visualization, interpretation, and exploration of longitudinal clinical data, clinical concepts and patterns. KNAVE-II mediates queries to a distributed temporal-abstraction architecture (IDAN), which uses a knowledge-based problem-solving method specializing in on-the-fly computation of clinical queries. METHODS: A two-phase, balanced cross-over study to compare efficiency and satisfaction of a group of clinicians when answering queries of variable complexity about time-oriented clinical data, typical for oncology protocols, using KNAVE-II, versus standard methods: both paper charts and a popular electronic spreadsheet (ESS) in Phase I; an ESS in Phase II. The measurements included the time required to answer and the correctness of answer for each query and each complexity category, and for all queries, assessed versus a predetermined gold standard set by a domain expert. User satisfaction was assessed by the Standard Usability Score (SUS) tool-specific questionnaire and by a "Usability of Tool Comparison" comparative questionnaire developed for this study. RESULTS: In both evaluations, subjects answered higher-complexity queries significantly faster using KNAVE-II than when using paper charts or an ESS up to a mean of 255s difference per query versus the ESS for hard queries (p=0.0003) in the second evaluation. Average correctness scores when using KNAVE-II versus paper charts, in the first phase, and the ESS, in the second phase, were significantly higher over all queries. In the second evaluation, 91.6% (110/120) of all of the questions asked within queries of all levels produced correct answers using KNAVE-II, opposed to only 57.5% (69/120) using the ESS (p<0.0001). User satisfaction with KNAVE-II was significantly superior compared to using either a paper chart or the ESS (p=0.006). Clinicians ranked KNAVE-II superior to both paper and the ESS. CONCLUSIONS: An evaluation of the functionality and usability of KNAVE-II and its supporting knowledge-based temporal-mediation architecture has produced highly encouraging results regarding saving of physician time, enhancement of accuracy of clinical assessment, and user satisfaction.

post to: CiteULike

A reliable method for cell phenotype image classification.

Publication Date: 2008 Apr 25 PMID: 18440791
Authors: Nanni, L. - Lumini, A.
Journal: Artif Intell Med

OBJECTIVE: Image-based approaches have proven to be of great utility in the automated cell phenotype classification, it is very important to develop a method that efficiently quantifies, distinguishes and classifies sub-cellular images. METHODS AND MATERIALS: In this work, the invariant locally binary patterns (LBP) are applied, for the first time, to the classification of protein sub-cellular localization images. They are tested on three image datasets (available for download), in conjunction with support vector machines (SVMs) and random subspace ensembles of neural networks. Our method based on invariant LBP provides higher accuracy than other well-known methods for feature extraction; moreover, our method does not require to (direct) crop the cells for the classification. RESULTS AND CONCLUSION: The experimental results show that the random subspace ensemble of neural networks outperforms the SVM in this problem. The proposed approach based on the solely LBP features gives accuracies of 85%, 93.9% and 88.4% on the 2D HeLa dataset, LOCATE endogenous and transfected datasets, respectively, and in combination with other state-of-the-art methods for the cell phenotype image classification we obtain a classification accuracy of 94.2%, 98.4% and 96.5%.

post to: CiteULike

Autonomous virtual mobile robot for three-dimensional medical image exploration: Application to micro-CT cochlear images.

Publication Date: 2008 May PMID: 18417327
Authors: Ferrarini, L. - Verbist, B. M. - Olofsen, H. - Vanpoucke, F. - Frijns, J. H. - Reiber, J. H. - Admiraal-Behloul, F.
Journal: Artif Intell Med

OBJECTIVE: In this paper, we present an autonomous virtual mobile robot (AVMR) for three-dimensional (3D) exploration of unknown tubular-like structures in 3D images. METHODS AND MATERIALS: The trajectory planning for 3D central navigation is achieved by combining two neuro-fuzzy controllers, and is based on 3D sensory information; a Hough transform is used to locally fit a cylinder during the exploration, estimating the local radius of the tube. Nonholonomic constraints are applied to assure a smooth, continuous and unique final path. When applied to 3D medical images, the AVMR operates as a virtual endoscope, directly providing anatomical measurements of the organ. After a thorough validation on challenging synthetic environments, we applied our method to eight micro-CT datasets of cochleae. RESULTS: Validation on synthetic environments proved the robustness of our method, and highlighted key parameters for the design of the AVMR. When applied to the micro-CT datasets, the AVMR automatically estimated length and radius of the cochleae: results were compared to manual delineations, proving the accuracy of our approach. CONCLUSIONS: The AVMR presents several advantages when used as a virtual endoscope: the nonholonomic constraint guarantees a unique and smooth central path, which can be reliably used both for qualitative and quantitative investigation of 3D medical datasets. Results on the micro-CT cochleae are a significant step towards the validation of more clinical computed tomography (CT) studies.

post to: CiteULike

Discovery and integration of univariate patterns from daily individual organ-failure scores for intensive care mortality prediction.

Publication Date: 2008 May PMID: 18394871
Authors: Toma, T. - Abu-Hanna, A. - Bosman, R. J.
Journal: Artif Intell Med

OBJECTIVES: The current established mortality predictive models in the intensive care rely only on patient information gathered within the first 24hours of admission. Recent research demonstrated the added prognostic value residing in the sequential organ-failure assessment (SOFA) score which quantifies on each day the cumulative patient organ derangement. The objective of this paper is to develop and study predictive models that also incorporate univariate patterns of the six individual organ systems underlining the SOFA score. A model for a given day d predicts the probability of in-hospital mortality. MATERIALS AND METHODS: We use the logistic framework to combine a summary statistic of the historic SOFA information for a patient together with selected dummy variables indicating the occurrence of univariate frequent temporal patterns of individual organ system functioning. We demonstrate the application of our method to a large real-life data set from an intensive care unit (ICU) in a teaching hospital. Model performance is tested in terms of the AUC and the Brier score. RESULTS: An algorithm for categorization, discovery, and selection of univariate patterns of individual organ scores and the induction of predictive models. The case-study resulted in six daily models corresponding to days 2-7. Their AUC ranged between 0.715 and 0.794 and the Brier scores between 0.161 and 0.216. Models using only admission data but recalibrated for days 2-7 generated AUC ranging between 0.643 and 0.761 and Brier scores ranged between 0.175 and 0.230. CONCLUSIONS: The results show that temporal organ-failure episodes improve predictions' quality in terms of both discrimination and calibration. In addition, they enhance the interpretability of models. Our approach should be applicable to many other medical domains where severity scores and sub-scores are collected.

post to: CiteULike

Pre-operative ordering of minimally invasive surgical tools: A fuzzy inference system approach.

Publication Date: 2008 May PMID: 18378128
Authors: Miller, D. J. - Nelson, C. A. - Oleynikov, D. - Jones, D. D.
Journal: Artif Intell Med

OBJECTIVE: With a limited number of access ports, minimally invasive surgery (MIS) often requires the complete removal of one tool and reinsertion of another in order to provide surgeons with the full functionality necessary to complete a procedure. MATERIALS AND METHODS: Endoscope video from 14 MIS procedures performed at the University of Nebraska Medical Center was used to collect usage statistics for various surgical instruments. This usage data was normalized and input to a fuzzy inference system (FIS) with four membership functions (MFs) to provide a crisp rating value for each instrument. Input membership functions included: number of uses ("Use"), total time used ("Time"), number of changes ("Change") and time per use ("Ave-Time"). Tools were arranged in a simulated cartridge system based on a "Usefulness" output membership function in such a way as to allow easy selection of the next instrument necessary to complete the procedure. Performance was measured by comparing the amount of cartridge indexing needed to complete a procedure using the FIS-generated arrangement against a set of random tool arrangements. RESULTS: The 14 FIS-generated tool arrangements considered in this investigation performed better than 64.11% of randomly generated tool arrangements and as well or better than 80.48% of tool arrangements. Using the FIS in conjunction with a multifunction laparoscopic tool, it is projected that an average cycle savings of 17.75% and 17.39% can be achieved over the mean and median of the random tool arrangements, respectively. CONCLUSIONS: For a given set of tools, the FIS used in this investigation provides an efficient method of arranging tools for MIS that performs at least as well or better than simply placing the tool tips into the chambers in a random configuration. This leads to a decrease in operating room time and corresponding decreases in both patient trauma from insertion and removal of tools and monetary cost, which is directly related to the amount of time spent changing instruments.

post to: CiteULike

Assessment of classification improvement in patients with Alzheimer's disease based on magnetoencephalogram blind source separation.

Publication Date: 2008 May PMID: 18329868
Authors: Escudero, J. - Hornero, R. - Poza, J. - Abasolo, D. - Fernandez, A.
Journal: Artif Intell Med

OBJECTIVES: In this pilot study, we intended to assess whether a procedure based on blind source separation (BSS) and subsequent partial reconstruction of magnetoencephalogram (MEG) recordings might enhance the differences between MEGs from Alzheimer's disease (AD) patients and elderly control subjects. MATERIALS AND METHODS: We analysed MEG background activity recordings acquired with a 148-channel whole-head magnetometer from 21 AD patients and 21 control subjects. Artefact-free epochs of 20s were blindly decomposed using the algorithm for multiple unknown signals extraction (AMUSE), which arranges the extracted components by decreasing linear predictability. Thus, the components of diverse epochs and subjects could be easily compared. Every component was characterised with its median frequency and spectral entropy (denoted by f(median) and SpecEn, respectively). The differences between subject groups in these variables were statistically evaluated to find out which components could improve the subject classification. Then, these significant components were used to partially reconstruct the MEG recordings. RESULTS: The statistical analysis showed that the AMUSE components which provided the largest differences between demented patients and control subjects were ordered together. Considering this analysis, we defined two subsets, denoted by BSS-{15,35} and BSS-{20,30}, which included 21 components (15-35) and 11 components (20-30), respectively. We partially reconstructed the MEGs with these subsets. Then, the classification performance was computed with a leave-one-out cross-validation procedure for the case where no BSS was applied and for the partial reconstructions BSS-{15,35} and BSS-{20,30}. The BSS and component selection procedure improved the classification accuracy from 69.05% to 83.33% using f(median) with BSS-{15,35} and from 61.91% to 73.81% using SpecEn with BSS-{20,30}. CONCLUSION: These preliminary results lead us to think that the proposed procedure based on BSS and selection of significant components may improve the classification of AD patients using straightforward features from MEG recordings.

post to: CiteULike