Artificial Intelligence in Medicine - Articles in Press

Memetic algorithms for de novo motif-finding in biomedical sequences - Corrected Proof

Chengpeng Bi — 2012-05-22

Abstract: Objectives: The objectives of this study are to design and implement a new memetic algorithm for de novo motif discovery, which is then applied to detect important signals hidden in various biomedical molecular sequences.Methods and materials: In this paper, memetic algorithms are developed and tested in de novo motif-finding problems. Several strategies in the algorithm design are employed that are to not only efficiently explore the multiple sequence local alignment space, but also effectively uncover the molecular signals. As a result, there are a number of key features in the implementation of the memetic motif-finding algorithm (MaMotif), including a chromosome replacement operator, a chromosome alteration-aware local search operator, a truncated local search strategy, and a stochastic operation of local search imposed on individual learning. To test the new algorithm, we compare MaMotif with a few of other similar algorithms using simulated and experimental data including genomic DNA, primary microRNA sequences (let-7 family), and transmembrane protein sequences.Results: The new memetic motif-finding algorithm is successfully implemented in C++, and exhaustively tested with various simulated and real biological sequences. In the simulation, it shows that MaMotif is the most time-efficient algorithm compared with others, that is, it runs 2 times faster than the expectation maximization (EM) method and 16 times faster than the genetic algorithm-based EM hybrid. In both simulated and experimental testing, results show that the new algorithm is compared favorably or superior to other algorithms. Notably, MaMotif is able to successfully discover the transcription factors’ binding sites in the chromatin immunoprecipitation followed by massively parallel sequencing (ChIP-Seq) data, correctly uncover the RNA splicing signals in gene expression, and precisely find the highly conserved helix motif in the transmembrane protein sequences, as well as rightly detect the palindromic segments in the primary microRNA sequences.Conclusions: The memetic motif-finding algorithm is effectively designed and implemented, and its applications demonstrate it is not only time-efficient, but also exhibits excellent performance while compared with other popular algorithms.

Feasibility of an objective electrophysiological loudness scaling: A kernel-based novelty detection approach - Corrected Proof

Mai Mariam, Wolfgang Delb, Bernhard Schick, Daniel J. Strauss — 2012-05-17

Abstract: Objective: The objective of our research is to structure a foundation for an electrophysiological loudness scaling measurement, in particular to estimate an uncomfortable loudness (UCL) level by using the hybrid wavelet-kernel novelty detection (HWND).Methods and materials: Late auditory evoked potentials (LAEPs) were obtained from 20 normal hearing adults. These LAEPs were stimulated by 4 intensity levels (60 decibel (dB) sound pressure level (SPL), 70dB SPL, 80dB SPL, and 90dB SPL). We have extracted the habituation correlates in LAEPs by using HWND. For this, we employed a lattice structure-based wavelet frame decompositions for feature extraction combined with a kernel-based novelty detector.Results: The group results showed that the habituation correlates degrees, i.e., relative changes within the sweep sequences, were significantly different among 60dB SPL, 70dB SPL, 80dB SPL, and 90dB SPL stimulation level, independently from the intensity related amplitude information in the averaged LAEPs. At these particular intensities, 60% of the subjects show the correlation between the novelty measures and the stimulation levels resembles a loudness scaling function, in reverse. In this paper, we have found a correlation in between the novelty measures and loudness perception as well. We have found that high ranges of loudness levels such as loud, upper level and too loud show generally 4.88% of novelty measures and comfortable ranges of loudness levels, i.e., soft, comfortable but soft, comfortable loud and comfortable but loud are generally have 12.29% of novelty measures.Additionally, we demonstrated that our sweep-to-sweep basis of post processing scheme is reliable for habituation extraction and offers an advantage of reducing experimental time as the proposed scheme need less than 20% of single sweeps in comparison to the amount that are commonly used in arithmetical average for a meaningful result.Conclusions: We assessed the feasibility of habituation correlates for an objective loudness scaling. With respect to this first feasibility study, the presented results are promising when using the described signal processing and machine learning methodology. For the group results, the novelty measures approach is able to discriminate 60dB, 70dB, 80dB and 90dB stimulated sweeps. In addition, a correlation between the novelty measures and the subjective loudness scaling is observed. However, more loudness perception and frequency specific experiments need to be conducted to determine the UCL novelty measures threshold as well as clinically oriented studies are necessary to evaluate whether this approach might be used in the objective hearing instrument fitting procedures.

Predicting warfarin dosage from clinical data: A supervised learning approach - Corrected Proof

Ya-Han Hu, Fan Wu, Chia-Lun Lo, Chun-Tien Tai — 2012-04-26

Abstract: Objective: Safety of anticoagulant administration has been a primary concern of the Joint Commission on Accreditation of Healthcare Organizations. Among all anticoagulants, warfarin has long been listed among the top ten drugs causing adverse drug events. Due to narrow therapeutic range and significant side effects, warfarin dosage determination becomes a challenging task in clinical practice. For superior clinical decision making, this study attempts to build a warfarin dosage prediction model utilizing a number of supervised learning techniques.Methods and materials: The data consists of complete historical records of 587 Taiwan clinical cases who received warfarin treatment as well as warfarin dose adjustment. A number of supervised learning techniques were investigated, including multilayer perceptron, model tree, k nearest neighbors, and support vector regression (SVR). To achieve higher prediction accuracy, we further consider both homogeneous and heterogeneous ensembles (i.e., bagging and voting). For performance evaluation, the initial dose of warfarin prescribed by clinicians is established as the baseline. The mean absolute error (MAE) and standard deviation of errors (σ(E)) are considered as evaluation indicators.Results: The overall evaluation results show that all of the learning based systems are significantly more accurate than the baseline (MAE=0.394, σ(E)=0.558). Among all prediction models, both Bagged Voting (MAE=0.210, σ(E)=0.357) with four classifiers and Bagged SVR (MAE=0.210, σ(E)=0.366) are suggested as the two most effective prediction models due to their lower MAE and σ(E).Conclusion: The investigated models can not only facilitate clinicians in dosage decision-making, but also help reduce patient risk from adverse drug events.

A knowledge-based clinical toxicology consultant for diagnosing single exposures - Corrected Proof

Joel D. Schipper, Douglas D. Dankel, A. Antonio Arroyo, Jay L. Schauben — 2012-04-23

Abstract: Objective: Every year, toxic exposures kill 1200 Americans. To aid in the timely diagnosis and treatment of such exposures, this research investigates the feasibility of a knowledge-based system capable of generating differential diagnoses for human exposures involving unknown toxins.Methods: Data mining techniques automatically extract prior probabilities and likelihood ratios from a database managed by the Florida Poison Information Center. Using observed clinical effects, the trained system produces a ranked list of plausible toxic exposures. The resulting system was evaluated using 30,152 single exposure cases. In addition, the effects of two filters for refining diagnosis based on a minimum number of exposure cases and a minimum number of clinical effects were also explored.Results: The system achieved accuracies (calculated as the percentage of exposures correctly identified in top 10% of trained diagnoses) as high as 79.8% when diagnosing by substance and 78.9% when diagnosing by the major and minor categories of toxins.Conclusions: The results of this research are modest, yet promising. At this time, no similar systems are currently in use in the United States and it is hoped that these studies will yield an effective medical decision support system for clinical toxicology.

Acute leukemia classification by ensemble particle swarm model selection - Corrected Proof

2012-04-17

Abstract: Objective: Acute leukemia is a malignant disease that affects a large proportion of the world population. Different types and subtypes of acute leukemia require different treatments. In order to assign the correct treatment, a physician must identify the leukemia type or subtype. Advanced and precise methods are available for identifying leukemia types, but they are very expensive and not available in most hospitals in developing countries. Thus, alternative methods have been proposed. An option explored in this paper is based on the morphological properties of bone marrow images, where features are extracted from medical images and standard machine learning techniques are used to build leukemia type classifiers.Methods and materials: This paper studies the use of ensemble particle swarm model selection (EPSMS), which is an automated tool for the selection of classification models, in the context of acute leukemia classification. EPSMS is the application of particle swarm optimization to the exploration of the search space of ensembles that can be formed by heterogeneous classification models in a machine learning toolbox. EPSMS does not require prior domain knowledge and it is able to select highly accurate classification models without user intervention. Furthermore, specific models can be used for different classification tasks.Results: We report experimental results for acute leukemia classification with real data and show that EPSMS outperformed the best results obtained using manually designed classifiers with the same data. The highest performance using EPSMS was of 97.68% for two-type classification problems and of 94.21% for more than two types problems. To the best of our knowledge, these are the best results reported for this data set. Compared with previous studies, these improvements were consistent among different type/subtype classification tasks, different features extracted from images, and different feature extraction regions. The performance improvements were statistically significant. We improved previous results by an average of 6% and there are improvements of more than 20% with some settings. In addition to the performance improvements, we demonstrated that no manual effort was required during acute leukemia type/subtype classification.Conclusions: Morphological classification of acute leukemia using EPSMS provides an alternative to expensive diagnostic methods in developing countries. EPSMS is a highly effective method for the automated construction of ensemble classifiers for acute leukemia classification, which requires no significant user intervention. EPSMS could also be used to address other medical classification tasks.

An implicit approach to deal with periodically repeated medical data - Corrected Proof

2012-04-16

Abstract: Context: Temporal information plays a crucial role in medicine, so that in medical informatics there is an increasing awareness that suitable database approaches are needed to store and support it. Specifically, a great amount of clinical data (e.g., therapeutic data) are periodically repeated. Although an explicit treatment is possible in most cases, it causes severe storage and disk I/O problems.Objective: In this paper, we propose an innovative approach to cope with periodic relational medical data in an implicit way.Methods: We propose a new data model, representing periodic data in a compact (implicit) way, which is a consistent extension of TSQL2 consensus approach. Then, we identify some important types of temporal queries, and present query answering algorithms to answer them. Finally, we also run experiments to evaluate our approach.Results: The experiments show that our approach outperforms current explicit approaches, especially as regard disk I/O.Conclusion: We have provided an implicit approach to periodic data with is a consistent extension of TSQL2 (and which is thus grant interoperable with it), and we have experimentally proven that it outperforms current explicit approaches.

Channel selection and classification of electroencephalogram signals: An artificial neural network and genetic algorithm-based approach - Corrected Proof

2012-04-13

Abstract: Objective: An electroencephalogram-based (EEG-based) brain–computer-interface (BCI) provides a new communication channel between the human brain and a computer. Amongst the various available techniques, artificial neural networks (ANNs) are well established in BCI research and have numerous successful applications. However, one of the drawbacks of conventional ANNs is the lack of an explicit input optimization mechanism. In addition, results of ANN learning are usually not easily interpretable. In this paper, we have applied an ANN-based method, the genetic neural mathematic method (GNMM), to two EEG channel selection and classification problems, aiming to address the issues above.Methods and materials: Pre-processing steps include: least-square (LS) approximation to determine the overall signal increase/decrease rate; locally weighted polynomial regression (Loess) and fast Fourier transform (FFT) to smooth the signals to determine the signal strength and variations. The GNMM method consists of three successive steps: (1) a genetic algorithm-based (GA-based) input selection process; (2) multi-layer perceptron-based (MLP-based) modelling; and (3) rule extraction based upon successful training. The fitness function used in the GA is the training error when an MLP is trained for a limited number of epochs. By averaging the appearance of a particular channel in the winning chromosome over several runs, we were able to minimize the error due to randomness and to obtain an energy distribution around the scalp. In the second step, a threshold was used to select a subset of channels to be fed into an MLP, which performed modelling with a large number of iterations, thus fine-tuning the input/output relationship. Upon successful training, neurons in the input layer are divided into four sub-spaces to produce if-then rules (step 3).Two datasets were used as case studies to perform three classifications. The first data were electrocorticography (ECoG) recordings that have been used in the BCI competition III. The data belonged to two categories, imagined movements of either a finger or the tongue. The data were recorded using an 8×8 ECoG platinum electrode grid at a sampling rate of 1000Hz for a total of 378 trials. The second dataset consisted of a 32-channel, 256Hz EEG recording of 960 trials where participants had to execute a left- or right-hand button-press in response to left- or right-pointing arrow stimuli. The data were used to classify correct/incorrect responses and left/right hand movements.Results: For the first dataset, 100 samples were reserved for testing, and those remaining were for training and validation with a ratio of 90%:10% using K-fold cross-validation. Using the top 10 channels selected by GNMM, we achieved a classification accuracy of 0.80±0.04 for the testing dataset, which compares favourably with results reported in the literature. For the second case, we performed multi-time-windows pre-processing over a single trial. By selecting 6 channels out of 32, we were able to achieve a classification accuracy of about 0.86 for the response correctness classification and 0.82 for the actual responding hand classification, respectively. Furthermore, 139 regression rules were identified after training was completed.Conclusions: We demonstrate that GNMM is able to perform effective channel selections/reductions, which not only reduces the difficulty of data collection, but also greatly improves the generalization of the classifier. An important step that affects the effectiveness of GNMM is the pre-processing method. In this paper, we also highlight the importance of choosing an appropriate time window position.

An automated methodology for levodopa-induced dyskinesia: Assessment based on gyroscope and accelerometer signals - Corrected Proof

2012-04-09

Abstract: Objective: In this study, a methodology is presented for an automated levodopa-induced dyskinesia (LID) assessment in patients suffering from Parkinson's disease (PD) under real-life conditions.Methods and Material: The methodology is based on the analysis of signals recorded from several accelerometers and gyroscopes, which are placed on the subjects’ body while they were performing a series of standardised motor tasks as well as voluntary movements. Sixteen subjects were enrolled in the study. The recordings were analysed in order to extract several features and, based on these features, a classification technique was used for LID assessment, i.e. detection of LID symptoms and classification of their severity.Results: The results were compared with the clinical annotation of the signals, provided by two expert neurologists. The analysis was performed related to the number and topology of sensors used; several different experimental settings were evaluated while a 10-fold stratified cross validation technique was employed in all cases. Moreover, several different classification techniques were examined. The ability of the methodology to be generalised was also evaluated using leave-one-patient-out cross validation. The sensitivity and positive predictive values (average for all LID severities) were 80.35% and 76.84%, respectively.Conclusions: The proposed methodology can be applied in real-life conditions since it can perform LID assessment in recordings which include various PD symptoms (such as tremor, dyskinesia and freezing of gait) of several motor tasks and random voluntary movements.

Detecting and resolving inconsistencies between domain experts’ different perspectives on (classification) tasks - Corrected Proof

Derek Sleeman, Laura Moss, Andy Aiken, Martin Hughes, John Kinsella, Malcolm Sim — 2012-04-06

Abstract: Objectives: The work reported here focuses on developing novel techniques which enable an expert to detect inconsistencies in 2 (or more) perspectives that the expert might have on the same (classification) task. The high level task which the experts (physicians) had set themselves was to classify, on a 5-point severity scale (A–E), the hourly reports produced by an intensive care unit's patient management system.Method: The INSIGHT system has been developed to support domain experts exploring, and removing inconsistencies in their conceptualization of a task. We report here a study of intensive care physicians reconciling 2 perspectives on their patients. The 2 perspectives provided to INSIGHT were an annotated set of patient records where the expert had selected the appropriate category to describe that snapshot of the patient, and a set of rules which are able to classify the various time points on the same 5-point scale. Inconsistencies between these 2 perspectives are displayed as a confusion matrix; moreover INSIGHT then allows the expert to revise both the annotated datasets (correcting data errors, or changing the assigned categories) and the actual rule-set.Results: Each of the 3 experts achieved a very high degree of consensus (∼97%) between his refined knowledge sources (i.e., annotated hourly patient records and the rule-set). We then had the experts produce a common rule-set and then refine their several sets of annotations against it; this again resulted in inter-expert agreements of ∼97%. The resulting rule-set can then be used in applications with considerable confidence.Conclusion: This study has shown that under some circumstances, it is possible for domain experts to achieve a high degree of correlation between 2 perspectives of the same task. The experts agreed that the immediate feedback provided by INSIGHT was a significant contribution to this successful outcome.

Pathway-based identification of a smoking associated 6-gene signature predictive of lung cancer risk and survival - Corrected Proof

Nancy Lan Guo, Ying-Wooi Wan — 2012-02-13

Abstract: Objective: Smoking is a prominent risk factor for lung cancer. However, it is not an established prognostic factor for lung cancer in clinics. To date, no gene test is available for diagnostic screening of lung cancer risk or prognostication of clinical outcome in smokers. This study sought to identify a smoking associated gene signature in order to provide a more precise diagnosis and prognosis of lung cancer in smokers.Methods and materials: An implication network based methodology was used to identify biomarkers by modeling crosstalk with major lung cancer signaling pathways. Specifically, the methodology contains the following steps: (1) identifying genes significantly associated with lung cancer survival; (2) selecting candidate genes which are differentially expressed in smokers versus non-smokers from the survival genes identified in Step 1; (3) from these candidate genes, constructing gene coexpression networks based on prediction logic for the smoker group and the non-smoker group, respectively; (4) identifying smoking-mediated differential components, i.e., the unique gene coexpression patterns specific to each group; and (5) from the differential components, identifying genes directly co-expressed with major lung cancer signaling hallmarks.Results: A smoking-associated 6-gene signature was identified for prognosis of lung cancer from a training cohort (n=256). The 6-gene signature could separate lung cancer patients into two risk groups with distinct post-operative survival (log-rank P<0.04, Kaplan–Meier analyses) in three independent cohorts (n=427). The expression-defined prognostic prediction is strongly related to smoking association and smoking cessation (P<0.02; Pearson's Chi-squared tests). The 6-gene signature is an accurate prognostic factor (hazard ratio=1.89, 95% CI: [1.04, 3.43]) compared to common clinical covariates in multivariate Cox analysis. The 6-gene signature also provides an accurate diagnosis of lung cancer with an overall accuracy of 73% in a cohort of smokers (n=164). The coexpression patterns derived from the implication networks were validated with interactions reported in the literature retrieved with STRING8, Ingenuity Pathway Analysis, and Pathway Studio.Conclusions: The pathway-based approach identified a smoking-associated 6-gene signature that predicts lung cancer risk and survival. This gene signature has potential clinical implications in the diagnosis and prognosis of lung cancer in smokers.

Prediction of human major histocompatibility complex class II binding peptides by continuous kernel discrimination method - Corrected Proof

Ju He, Guobing Yang, Hanbing Rao, Zerong Li, Xianping Ding, Yuzong Chen — 2012-02-10

Abstract: Objective: Accurate prediction of major histocompatibility complex (MHC) class II binding peptides helps reducing the experimental cost for identifying helper T cell epitopes, which has been a challenging problem partly because of the variable length of the binding peptides. This work is to develop an accurate model for predicting MHC-binding peptides using machine learning methods.Methods: In this work, a machine learning method, continuous kernel discrimination (CKD), was used for predicting MHC class II binders of variable lengths. The composition transition and distribution features were used for encoding peptide sequence and the Metropolis Monte Carlo simulated annealing approach was used for feature selection.Results: Feature selection was found to significantly improve the performance of the model. For benchmark dataset Dataset-1, the number of features is reduced from 147 to 24 and the area under the receiver operating characteristic curve (AUC) is improved from 0.8088 to 0.9034, while for benchmark dataset Dataset-2, the number of features is reduced from 147 to 44 and the AUC is improved from 0.7349 to 0.8499. An optimal CKD model was derived from the feature selection and bandwidth optimization using 10-fold cross-validation. Its AUC values are between 0.831 and 0.980 evaluated on benchmark datasets BM-Set1 and are between 0.806 and 0.949 on benchmark datasets BM-Set2 for MHC class II alleles. These results indicate a significantly better performance for our CKD model over other earlier models based on the training and testing of the same datasets.Conclusions: Our study suggested that the CKD method outperforms other machine learning methods proposed earlier in the prediction of MHC class II biding peptides. Moreover, the choice of the cut-off for CKD classifier is crucial for its performance.