Antti Saastamoinen

Development and comparison of four sleep spindle detection methods

OBJECTIVE The objective of the present work was to develop and compare methods for automatic detection of bilateral sleep spindles. METHODS AND MATERIALS All-night sleep electroencephalographic (EEG) recordings of 12 healthy subjects with a median age of 40 years were studied. The data contained 6043 visually scored bilateral spindles occurring in frontopolar or central brain location. In the present work a new sigma index for spindle detection was developed, based on the fast Fourier transform (FFT) spectrum, aiming at approximating our previous fuzzy spindle detector. The sigma index was complemented with spindle amplitude analysis, based on finite impulse response (FIR) filtering, to form of a combination detector of bilateral spindles. In this combination detector, the spindle amplitude distribution of each recording was estimated and used to tune two different amplitude thresholds. This combination detector was compared to bilaterally extracted sigma indexes and fuzzy detections, which aim to be independent of absolute spindle amplitudes. As a fourth method a fixed spindle amplitude detector was included. RESULTS The combination detector provided the best overall performance; in S2 sleep a 70% true positive rate was reached with a specificity of 98.6%, and a false-positive rate of 32%. The bilateral sigma indexes provided the second best results, followed by fuzzy detector, while the fixed amplitude detector provided the poorest results so that in S2 sleep a 70% true positive rate was reached with a specificity of 97.7% and false-positive rate of 46%. The spindle amplitude distributions automatically determined for each recording by the combination detector were compared to amplitudes of visually scored spindles and they proved to correspond well. Inter-hemispheric amplitude variation of visually scored bilateral spindles is also presented. CONCLUSION Flexibility is beneficial in the detection of bilateral spindles. The present work advances automated spindle detection and increases the knowledge of bilateral sleep spindle characteristics.

Waveform detection with RBF network – Application to automated EEG analysis

Abstract Automated detection of different waveforms in physiological signals has been one of the most intensively studied applications of signal processing in the clinical medicine. During recent years an increasing amount of neural network based methods have been proposed. In this paper we present a radial basis function (RBF) network based method for automated detection of different interference waveforms in epileptic EEG. This kind of artefact detector is especially useful as a preprocessing system in combination with different kinds of automated EEG analyzers to improve their general applicability. The results show that our neural network based classifier successfully detects artefacts at the rate of over 75% while the correct classification rate for normal segments is as high as about 95%.

Computer program for automated sleep depth estimation

In this article, we present a new implementation of an amplitude-independent method for continuous-scale sleep depth estimation. Having been implemented as an add-on analysis module under commercially available biosignal recording and analysis software, it can be easily applied in clinical routine. The software gives the user full freedom to change all the analysis parameters inside theoretical limits. Computational sleep depth profiles produced by the presented software compare favourably with visual classifications. Future work will concentrate on systematic optimization of analysis parameters, further evaluation of the method with disturbed sleep and application of the method for automated adaptive sleep analysis.

Fuzzy detection of EEG alpha without amplitude thresholding

Intelligent automated systems are needed to assist the tedious visual analysis of polygraphic recordings. Most systems need detection of different electroencephalogram (EEG) waveforms. The problem in automated detection of alpha activity is the large inter-individual variability of its amplitude and duration. In this work, a fuzzy reasoning based method for the detection of alpha activity was designed and tested using a total of 32 recordings from seven different subjects. Intelligence of the method was distributed to features extracted and the way they were combined. The ranges of the fuzzy rules were determined based on feature statistics. The advantage of the detector is that no alpha amplitude threshold needs to be selected. The performance of the alpha detector was assessed with receiver operating characteristic (ROC) curves. When the true positive rate was 94.2%, the false positive rate was 9.2%, which indicates good performance in sleep EEG analysis.

Determination of dominant simulated spindle frequency with different methods

Accurate analysis of EEG sleep spindle frequency is challenging. The frequency content of true sleep spindles is not known. Therefore, simulated spindle activity was studied in the present work. Five types of simulated test signals were designed, all containing a dominant spindle represented by a 13-Hz sine wave as such or with a waxing and waning pattern accompanied by a secondary spindle activity in three test signals. Background EEG was included in four test signals, modeled either as small additional sinusoids across the spindle frequency range or as filtered Gaussian noise segments. The purpose of this study was to investigate how accurately the dominant spindle frequency of 13 Hz could be resolved with different methods in the presence of the interfering waveforms. A matching pursuit (MP) based approach, discrete Fourier transform (DFT) with Hanning windowing with and without zero padding, Hankel total least squares (HTLS) and wavelet methods were compared in the analyses. MP method provided best overall performance, followed closely by DFT with zero padding. Comparative studies like this are important to decide the method of choice in clinical sleep EEG analysis.

Diffuse sleep spindles show similar frequency in central and frontopolar positions.

The objective of the present work was to examine fronto-central spindle frequency. A previously validated spindle detector, providing an electroencephalographic (EEG) amplitude independent spindle detection, was used to detect bilateral sleep spindles from sleep EEG recordings of ten healthy subjects with a time resolution of 0.33-s. A bilateral spindle detected centrally and frontopolarly simultaneously is called here a diffuse spindle. A bilateral spindle detected only frontopolarly or centrally at a given time is called a pure frontopolar and a pure central spindle, respectively. Spindle frequency was obtained with zero-padded discrete Fourier transform (DFT). Waveform phase angle of diffuse spindles was also examined. A total of 1230 diffuse spindles and 5316 pure central and 2595 pure frontopolar spindles were detected. The difference of median spindle frequency between central and frontopolar brain positions was clearly smaller in diffuse spindles than in pure spindles. Moreover, 34% of the diffuse spindles showed a similar frequency in central and frontopolar locations. This figure was up to 50.9% when including the 700 diffuse spindles fulfilling a strict anteroposterior (AP) timing criteria. The timing criteria selection in diffuse spindle analysis is a new functionality, enabled by the present spindle analysis method. Diffuse spindles showed coherent spindle oscillation in a large fronto-central area. Pure frontopolar spindles might be special cases of diffuse spindles, both of them seem to be generated in the nucleus medialis dorsalis (NMD) of the thalamus.

Anteroposterior Difference in EEG Sleep Depth Measure is Reduced in Apnea Patients

In the present work, mean frequencies of FFT amplitude spectra from six EEG derivations were used to provide a frontopolar, a central and an occipital sleep depth measure. Parameters quantifying the anteroposterior differences in these three sleep depth measures during the night were also developed. The method was applied to analysis of 30 all-night recordings from 15 healthy control subjects and 15 apnea patients. Control subjects showed larger differences in sleep depth between frontopolar and central positions than the apnea patients. The relatively reduced frontal sleep depth in apnea patients might reflect the disruption of the dynamic sleep process caused by apneas.

Vigilance stages and performance in OSAS patients in a monotonous reaction time task

OBJECTIVES To develop improved methods for objective assessment of daytime vigilance. This is important in the diagnosis and therapy control of sleep disorders associated with excessive daytime sleepiness (EDS). METHODS Twenty-one patients with EDS due to obstructive sleep apnea were recorded during a daytime 90 min reaction time (RT) test in a monotonous situation. Two EEG, two EOG and a submental EMG channel were recorded simultaneously. The recordings were divided into short, stationary segments of variable length (0.5-2 s) and classified into 7 stages using our previously described system, which includes additional stages for drowsiness. RESULTS The duration of RT was linearly correlated to the vigilance state (P<0.05). The appearance of slow eye movements (SEM) was more consistently related to performance impairment than EEG changes. CONCLUSIONS Our system can provide a better tracking of vigilance changes than the standardized sleep stage scoring. A 1-2 h test is useful in the assessment of the performance of a subject suffering from EDS. We found that SEMs indicate more sensitive and consistent EDS-related performance impairment than changes in EEG activity.

Compressed tracheal sound analysis in screening of sleep-disordered breathing

OBJECTIVE To evaluate the suitability of compressed tracheal sound signal for screening sleep-disordered breathing. METHODS Thirty-three consecutive patients underwent a polysomnography with a tracheal sound analysis. Nineteen patients were healthy except for the sleep complaint, 9 were hypertonic and 3 were hypertonic and had elevated cholesterol. Minimum and maximum values of each consecutive, non-overlapping segment of 15s of original sound data were extracted. All these compressed tracheal sound traces were divided into plain, thin and thick signal periods. Also pure, 10-min episodes of plain, thin and thick tracheal sound periods were selected and the nasal pressure flow shapes during these pure sound periods were examined. RESULTS There was a significant positive correlation between the total nocturnal amount of thick periods and AHI. Apneas and hypopneas were most common during the 10-min episodes of thick sound periods. The proportion of round (normal, non-flattened) inspiratory flow shape was highest during the pure plain periods. CONCLUSIONS Breathing consisting of apneas and hypopneas can quite reliably be visualised with compressed tracheal sound analysis. The other interesting outcome of the study is that even prolonged flow limitation might be revealed with the method. SIGNIFICANCE Compressed tracheal sound analysis might provide a promising screening method for obstructive apneas and hypopneas.

Intelligent methods for identifying respiratory cycle phases from tracheal sound signal during sleep

We present two methods for identifying respiratory cycle phases from tracheal sound signal during sleep. The methods utilize the Hilbert transform in envelope extraction. They determine automatically a patient-specific amplitude threshold to be used in the detection. The core of one method is designed to be amplitude-independent whereas the other method uses solely the amplitude information. The methods provided average sensitivities of 98% and 99%, respectively, and positive prediction values of 100% on the total of 1434 respiratory cycles analysed from six different patients. The developed methods seem promising as such or as tools for analysing sleep disordered breathing.