Biomedical subjects

Martti Juhola

Publications and source records attributed to Martti Juhola.

At least 19 recordsLinked to original sources

PubMed2006

A signal analysis technique of vestibulo-ocular reflex stimulated with impulsive head movements.

Eye movements have been investigated in several areas of medicine and also elsewhere, such as in psychology or even in the development of human-computer interfaces. In the last few years we have designed a technique to stimulate, measure and analyze vestibulo-ocular reflex eye movements. In the otoneurological literature these are seen as a novel and promising means of revealing certain disorders and diseases associated with vertigo. Vestibulo-ocular reflex is stimulated by impulsive head movements. We developed the present pattern recognition technique to detect the stimulus (impulsive head movements) and the vestibulo-ocular reflex (response eye movements) generated from signals and to compute the latency and the gain values between them. Using our technique to calculate these attributes, we obtained clearly different results for a group of 22 dizzy patients than for a group of 30 healthy subjects.

Dizziness↗

PubMed2006

On the neural network classification of medical data and an endeavour to balance non-uniform data sets with artificial data extension.

We studied the efficiency of multilayer perceptron networks to classify eight different medical data sets with typical problems connected to their strongly non-uniform distributions between output classes and relatively small sizes of training sets. We studied especially the possibility mentioned in the literature of balancing a class distribution by artificially extending small classes of a data set. The results obtained supported our hypothesis that principally this does somewhat improve the classification accuracy of small classes, but is also inclined to impair the classification accuracy of majority classes.

Algorithms↗

PubMed2006

Techniques for biased data distributions and variable classification with neural networks applied to otoneurological data.

It is frequently useful and advantageous to investigate not only the classification efficacy of neural networks, but also the reasons for misclassification and relations between input variables and output classes. We have developed novel techniques to disentangle these dilemmas: a network structure and learning strategy for biased output class distributions, a method to measure the classification information incorporated in variables and variable groups, and methods to express properties learned by a network from its structure. We tested these techniques with otoneurological data from the conjunction with vertiginous diseases that we have explored in our previous neural network studies.

Bias↗

PubMed2006

Virtual reality in posturography.

Balance dysfunctions are common, especially among elderly people. Present methods for the diagnosis and evaluation of severity of dysfuntion have limited value. We present a system that makes it easy to implement different visual and mechanical perturbations for clinical investigations of balance and visual-vestibular interaction. The system combines virtual reality visual stimulation with force platform posturography on a moving platform. We evaluate our contruction's utility in a classification task between 33 healthy controls and 77 patients with Ménière's disease, using a series of tests with different visual and mechanical stimuli. Responses of patients and controls differ significantly in parameters computed from stabilograms. We also show that the series of tests achieves a classification accuracy slightly over 80% between controls and patients.

Artificial Intelligence↗

PubMed2006

A neural network model for the simulation of word production errors of Finnish nouns.

We constructed a connectionist model with multilayer perceptron networks for the simulation of word production errors in the Finnish language. Such errors occur as slips of the tongue in healthy subjects and even more often in aphasic patients suffering from brain damage. Word production errors are of theoretical interest as they open an avenue to inner workings of the language production system. Here we present our coding schemes for semantic and phoneme information of words, investigate the general properties of the model, and compare the results of the model against empirical naming error data from ten aphasic patients. The model was reasonable in simulating word production errors in this heterogeneous group of aphasic patients.

Aphasia, Broca↗

PubMed2006

On neural network classification of otoneurological cases on the basis of recognition results of vestibulo-ocular reflex eye movements signal.

We constructed a signal analysis and recognition system to compute revealing results from vestibulo-ocular reflex eye movements and their stimulation head movements of impulsive type that were generated as passive head movements with a special device designed by us for this purpose. Further, we implemented perceptron neural networks to separate healthy subjects from patients suffering from dysfunction of vestibulo-ocular reflex in either ear. We gained high classification accuracies with this method ready for routine use.

Ear, Inner↗

PubMed2005

Genome-wide selection of unique and valid oligonucleotides.

Functional genomics methods are used to investigate the huge amount of information contained in genomes. Numerous experimental methods rely on the use of oligo- or polynucleotides. Nucleotide strand hybridization forms the underlying principle for these methods. For all these techniques, the probes should be unique for analyzed genes. In addition to being unique for the studied genes, the probes should fulfill a large number of criteria to be usable and valid. The criteria include for example, avoidance of self-annealing, suitable melting temperature and nucleotide composition. We developed a method for searching unique and valid oligonucleotides or probes for genes so that there is not even a similar (approximate) occurrence in any other location of the whole genome. By using probe size 25, we analyzed 17 complete genomes representing a wide range of both prokaryotic and eukaryotic organisms. More than 92% of all the genes in the investigated genomes contained valid oligonucleotides. Extensive statistical tests were performed to characterize the properties of unique and valid oligonucleotides. Unique and valid oligonucleotides were relatively evenly distributed in genes except for the beginning and end, which were somewhat overrepresented. The flanking regions in eukaryotes were clearly underrepresented among suitable oligonucleotides. In addition to distributions within genes, the effects on codon and amino acid usage were also studied.

Amino Acids↗

PubMed2005

On exact string matching of unique oligonucleotides.

Unique, gene-specific oligonucleotides are used for many genetic investigations such as polymerase chain reaction, gene cloning, microarray technology and antisense DNA studies. It is a computationally demanding task to extract these oligonucleotides from DNA databases. We studied the problem from the point of view of the string matching problem. We implemented and tested several exact string matching algorithms and modified the implementations to be as effective as possible. Ten different implementations were tested on yeast genomic sequence data. The run times for the best algorithms were significantly improved compared to conventional approaches, while in principle, i.e. in respect of theoretical time complexity, these algorithms do not actually differ essentially from each other.

Algorithms↗

PubMed2005

Neural network screening of electromyographic signals as the first phase to design novel human-computer interaction.

The present aim was to describe the first phase attempts to recognise voluntarily produced changes in electromyographic signals measured from two facial muscles. Thirty subjects voluntarily activated two facial muscles, corrugator supercilii and zygomaticus major. We designed a neural network based recognition system that screened out muscle activations from the electromyographic signals. When several subjects were tested according to the same test protocol, the neural network system was able to correctly recognise more than 95 % of all muscle activations. This is a promising result and we shall next proceed to modify the system for real-time functioning and then design its utilisation for various multimodal human-computer interaction techniques. The subsequent phase in the future will be the interaction backwards: when a computer program first recognised the use of the facial muscles, it will then follow the instructions given by the user. For instance, by using the facial muscles the subject could select or activate objects on the computer screen. This would be one of the opportunities that we develop to help, e.g., disabled persons, who are unable to use their hands.

Electromyography↗

PubMed2004

Influence of lossy compression on eye movement signals.

Eye movements considered in our research are physiological signals that are measured in otoneurological balance tests. They are also investigated in other areas of medicine and in psychology. When great amounts of signals are measured in clinical and research work, signal compression is of great use in storing measurements for later investigations. In this research we assessed the influence of lossy compression on medically interesting parameter values that are computed from eye movement signals. We found that high compression ratios with bit rates lower than 1.5 bits per sample on signals with an original resolution of 13 bits per sample produced results without significant changes to the medical parameters values.

Data Compression↗

PubMed2004

A neural network model of lexicalization for simulating the anomic naming errors of dementia patients.

Word-finding difficulty (anomia) is the most common linguistic deficit in dementia. It is often measured by picture naming tasks as naming a picture taps all the major processes in word production, i.e., activation of a concept, retrieval of lexical-semantic information on that concept, retrieval of the corresponding word form and articulation. Naming and naming errors have extensively been simulated by neural network models of lexicalization (see e.g. [1,2]). A common feature of these models is that they are static, i.e. non-learning. However, naming is a dynamic process that changes as a function of normal learning or re-learning after neural damage. These important patterns cannot be caught by the static models of lexicalization. Therefore we have developed a learning model of lexicalization based on multi-layer-perceptron (MLP) neural networks. We tested the model by fitting it to the naming data of 22 Finnish-speaking dementia patients and 19 neurologically intact control subjects. The tests showed an excellent fit between the model's and the subjects naming response distributions. Thus our model seems be suitable to simulate naming disorders of dementia patients.

Alzheimer Disease↗

PubMed2004

A comparison of different heterogeneous proximity functions and Euclidean distance.

Proximity functions evaluate distances or similarities between objects. Unlike the Euclidean distance, heterogeneous proximity functions process variables differently according to their scale. The correct evaluation of nominal variables, whose values are unordered, is especially important. We compared five heterogeneous functions with the Euclidean distance to study whether functions sensitive to scale are better than a function assuming the same scale. In addition, we were interested of the relative performance of the five heterogeneous functions. The performance of the functions was measured with a nearest neighbor classifier that was applied to 12 medical data sets characterized with different scales. Unexpectedly, the performance of heterogeneous functions did not differ significantly from that of the Euclidean distance. As expected, significant differences between the Heterogeneous Value Difference Metric (HVDM) and the four value-matching-based heterogeneous functions favored HVDM. Additional research is needed to explain why heterogeneous functions did not outperform the Euclidean distance.

Artificial Intelligence↗

PubMed2003

Lossy compression of eye movement and auditory brainstem response signals.

Eye movement and auditory brainstem response signals recorded for balance and hearing investigations were used as a medical test battery for several types of lossy compression techniques. These signals are associated with the function of the ears. The former signals are used to assess the balance problems (especially vertigo) of a subject and the latter his or her hearing problems. New technique is also presented based on successive approximation quantization. The effect of information loss on medical parameters computed from the signals in the course of compression was evaluated for brainstem response signals. It is important to ensure that lossy compression techniques of these biomedical signals do not impair medical parameter values computed from the signals.

Algorithms↗

PubMed2003

Development of virtual reality stimuli for force platform posturography.

People relying much on vision in the control of posture are known to have an elevated risk of falling. Dependence on visual control is an important parameter in the diagnosis of balance disorders. We have previously shown that virtual reality (VR) methods can be used to produce visual stimuli that affect balance, but suitable stimuli need to be found. In this study, the effect of six different VR stimuli on the balance of 22 healthy test subjects was evaluated using force platform posturography. We report in more detail and expand the results published earlier. According to the tests two of the stimuli have a significant destabilizing effect on balance. In addition a significant displacement effect on the subject's center of pressure (COP) was found. Thus it is shown that the design of VR stimuli to cause different effects on the control of balance is possible.

Adult↗

PubMed2002

Generating decision trees from otoneurological data with a variable grouping method.

When medical data sets are modelled by machine learning methods, wealth of variables may be available. This paper deals with variable selection for decision tree induction in the context of two otoneurological data sets: vertigo data, and postoperative nausea and vomiting data. First, a variable grouping method based on measures of association and graph theoretic techniques was used to gain insight into data. Then, representations of learning data were defined using the information from discovered variable groups, and decision trees were generated. The use of variable grouping method was beneficial by revealing interesting associations between variables and enabling generation of accurate and reasonable decision trees that modelled the application areas from different viewpoints.

Data Collection↗

PubMed2002

Lossless compression of otoneurological eye movement signals.

We studied the performance of several lossless compression algorithms on eye movement signals recorded in otoneurological balance and other physiological laboratories. Despite the wide use of these signals their compression has not been studied prior to our research. The compression methods were based on the common model of using a predictor to decorrelate the input and using an entropy coder to encode the residual. We found that these eye movement signals recorded at 400 Hz and with 13 bit amplitude resolution could losslessly be compressed with a compression ratio of about 2.7.

Algorithms↗

PubMed2002

Virtual reality stimuli for force platform posturography.

Adult↗

PubMed2002

Neural networks in recognition of electrical signals recorded from two facial muscle sites.

The present aim was to develop signal analysis methods to recognize muscle action potentials recorded with electromyography (EMG) from two facial muscle sites, corrugator supercilii (the muscle activated in frowning) and zygomaticus major (the muscle activated in smiling). Fourteen subjects produced volitional activations on both muscle sites, first, on a single muscle site activation basis and finally combinatory activations of both muscle sites. Wavelets and neural networks were used to analyse these voluntarily produced bursts of electric signals. Our results showed well over 90% recognition rates for all signal types. It is possible to utilize these communicative signals, for example, for multimodal human-computer interaction.

Action Potentials↗