Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2016 Dec 1.
Published in final edited form as: Sleep Med. 2015 Apr 24;16(12):1541–1549. doi: 10.1016/j.sleep.2015.03.008

MATPLM1, A MATLAB Script For Scoring Of Periodic Limb Movements: Preliminary Validation With Visual Scoring

Andy S Huang 1, Patrick Skeba 1, Myung S Yang 2, Francis P Sgambati 3, Christopher J Earley 2, Richard P Allen 2,*
PMCID: PMC4619191  NIHMSID: NIHMS689922  PMID: 26210395

Abstract

Background and Purpose

A MATLAB script (MATPLM1) was developed to rigorously apply WASM scoring criteria for PLMS from bilateral EMG leg recordings. This study compares MATPLM1 with both standard technician and expert detailed visual PLMS scoring.

Methods and Subjects

Validation was based on a ‘macro’ level by agreement for PLMS/hr during a night recording and on a ‘micro’ level by agreement for detection of each PLMS from a stratified random sample for each subject. Data available for these analyses were from 15 RLS (age: 61.5 ±8.5, 60% female) and 9 control subjects (age: 61.4 ±7.1, 67% female) participating in another study.

Results

In the ‘micro’ analysis, MATPLM1 and visual detection of PLMS events agreed 87.7% for technician scoring and 94.4 % for expert scoring. The technician and MATPLM1 scoring disagreements were checked for 36 randomly selected events, 97% involved clear technician-scoring error. In the ‘macro’ analysis, MATPLM1 rates of PMLS/hr correlated highly with visual scoring by the technician (r2=0.97) and the expert scorer (r2=0.99) but the technician scoring was consistently less than MATPLM1: Median (quartiles) difference: 10 (5, 23). There was little disagreement with expert scorer [Median (quartile) difference: −0.3(−2.4, 0.3)].

Conclusions

The MATPLM1 produces reliable scoring of PLMS that matches expert scoring. The standard visual scoring without careful measuring of events tends to significantly underscore PLMS. These preliminary results support the use of MATPLM1 as a preferred method of scoring PLMS for EMG recordings that are of a good quality and without significant sleep-disordered breathing events.

Keywords: MATPLM1, Periodic Limb Movements in Sleep (PLMS), Periodic Leg Movements (PLM), Restless Leg Syndrome (RLS)

2. Introduction

Periodic limb movements in sleep (PLMS) occur as a motor sign of the restless leg syndrome (RLS), also known as Willis-Ekbom disease (WED) (1), but also occur in other conditions (2) and tend to become more prominent with age (3). These events have been most precisely defined in the World Associations of Sleep Medicine Criteria (WASM) (4). They are measured from uncalibrated EMG recordings from surface electrodes on bilateral anterior tibialis muscles. The events from each leg are combined following rules specified in the WASM criteria. These events are commonly identified by human visual scoring. The scorer moves through a night's recording of sleep of the patient looking at 30 – 120 second epochs, marking each leg movement event (LM) that meets the criteria for a PLMS. This is often assisted by a scoring program (RemLogic) that is part of the systems used to collect the physiological data from sleep. These programs mark potential PLMS but most are not validated, do not use the WASM/IRLSSG standard scoring criteria and generally need considerable visual correction. The WASM standard requires careful and detailed measurements of the EMG for each potential PLM. It provides an explicit definition of the EMG signal for a PLM rather than relying on judgment of a visual scorer. The tedious nature of measuring PLMS by the WASM criteria, however, produces situations likely to lead to scoring errors. For example, scoring fatigue may occur when care has to be taken to measure hundreds of move, leading to events being missed. Conversely when they are rare, false expectations can lead to failure to measure possible events. Moreover the current visual scoring process focuses on the absolute number of events and does not closely assess start and stop times of events and thus usually does not provide a good measure of the durations and inter-movement intervals between the events. Recent studies have noted that measurement of PLMS should include consideration of these other features that are not reliably produced by the usual visual scoring, particularly for inter-movement interval (IMI) and the periodicity index (5).

In this paper, we demonstrate the ability to score PLMS from the EMG component of patients’ polysomnogram using a program (MATPLM1) on Matrix Laboratory (MATLAB) and validate its accuracy for detection of events in comparison with the traditional visual scoring. The validation focuses on detection of PLMS events independent of related factors such as EEG arousal or respiratory events. Future versions of MATPLM1 will add script to mark the PLMS occurring with significant events. This validation study not only determines at the macro level agreement for total number of PLMS during a night recording, but also evaluates at the micro level the basis for difference from a stratified random selection of specific PLM covering the full night's recording. This MATLAB program provides a general application for use on any EMG with sleep-stage scored data from sleep labs that can be converted to European Data Format (EDF). It also provides the periodicity index and descriptive statistics with arrays for PLM durations, amplitudes, sleep stages, time and IMI for each PLMS. The script allows adjusting significant parameters, e.g. sampling rate, filter densities, minimum IMI and can be used for batch processing of a large set of data.

3. Methods

3.1 MATPLM1 scoring algorithm

The MATPLM1 program reads the sampling rate from the EDF file. It applies the MATLAB implementation of a Butterworth filter with low pass set at 225 and high pass set at 20 Hz to the EMG data and then rectifies the signal. The filter settings are parameters in the program set for the rectified signal and sampling rate for the data used in this study. They can be easily adjusted as appropriate. A separate text array specifies the sleep stage for each 30-second epoch of the night's sleep along with start and stop points for sleep scoring on the EMG. The WASM/IRLSSG criteria (4) are fully implemented in MATPLM1. These criteria suffice for identification of PLMS without any visual review except for determining the resting EMG threshold. MATPLM1also includes an algorithm to identify the resting EMG. MATPLM1provides the results for combining both legs following the WASM criteria, but it can provide results for each leg separately if desired.

3.2 MATPLM1 algorithm for resting EMG level

The MATPLM1algorithm determines the resting EMG level in 2 steps. First it searches from the start of the record for the first consecutive ten-second interval where the patient has a stable EMG. A stable rectified EMG signal is defined by the maximum EMG signal less than 20 μv and also less than 5 standard deviations above the average for a. ten-second period. This process effectively ensures the stable EMG interval does not include a significant EMG movement event, such as a PLM. The resting EMG is defined as the median of the largest EMG reading from the first five ten-second intervals with stable EMG signal. In extreme cases, where five stable ten-second intervals are not found in the entire record, the program arbitrarily sets resting EMG at 18 μV (low threshold for movement end then = 20 μv). Such a case only occurred in this study for one very severe RLS patients who had an exceptionally high number of leg movements and no 10-second interval without some portion of a movement in the entire tracing. The two thresholds for significant leg movement detection according to the WASM criteria are set as 2 and 8 μv above the resting EMG level for movement end and start respectively.

3.3 Visual scoring

The validation was planned for two comparisons of MATLAB1 with visual scoring of PLMS. The first compared MATLAB1 to the usual sleep-laboratory technician scoring of the PLMS on all available records (n=24) at the time of this study. The second compared MATLAB1 to an expert scorer on a subset (n= 9) of all the records that had the larger differences between the technician and MATLAB1 scoring. Experienced certified sleep technicians visually scored the records for the first comparison as they would for any clinical service. The PLM were first identified and marked on the PSG tracing by the recording system (RemLogic-3). The sleep technicians visually inspected the PSG tracing and based on the WASM/IRLSSG criteria corrected the PLMS identifications. Major clinical and research centers normally use this process and this evaluation process is considered standard for sleep medicine practice. This was part of the technicians’ routine service provided for both clinical and research laboratories and the technicians were unaware of the planned comparison with MATPLM1.

There is obvious concern that scorer fatigue for some PSGs, which could contain hundreds of PLMS, and low-rate expectancy for PSGs with few PLMS would lead to errors in carefully applying MATLAB scoring criteria. Therefore, the second comparison provided a check on the technician visual scoring accuracy. The expert scorer had been highly trained and had 5 years of experience scoring PLMS for research studies. He was instructed to take as much time as needed to carefully evaluate and measure each of the PLMS events using expanded displays where appropriate. He rigorously applied the WASM/IRLSSG criteria for PLMS. The subset of 9 records selected for expert analyses included one record of the two records from severe RLS patients with PLMS/hr >200 and all other records with a difference between the technician visual scoring and MATPLM1 that was more than 15 per hour and more than 15% of the technician score. The expert scoring generally required about 2 to 4 hours per record for this detailed visual scoring of only the PLMS. The technician and expert visual scoring were done without knowledge of the PLMS scores from MATPLM1 or from the other visual scorer.

3.4 Subjects

The data from the second of two consecutive nights of polysomnography recordings were available for all analyses from 15 RLS (average age ± SD: 61.5 ± 8.5, 60% female) and 9 control subjects (average age ± SD: 61.4 ± 7.1, 67% female) who were participating in an ongoing RLS study. All subjects gave Johns Hopkins IRB-approved informed consent for the sleep study and any analyses of the de-identified data used in this study. The RLS patients were off all RLS medications for at least 12 days and at that time had an average IRLS of 26.2 ± 5.5 (range 16 – 35). This provides a patient sample covering the full range of moderate (usual low score for clinical trials is 15) to very severe (maximum IRLS score is 40) RLS. Ten of the 15 RLS patients had been on dopamine agonists, 3 on an α2δ and 1 on an opioid. The RLS patients had been screened to include only those with an average PLMS >15/hr over 5 days of recording at home starting when off medication for at least 5 days. This recording used the PAM-RL (Phillips Respironics) activity meter that has been validated for use with RLS patients (6, 7) ; (8). The controls were included only if they had an average of PLMS <10/hr for 5 consecutive days of similar recording at home. Controls and RLS patients were clinically screened to exclude those with significant sleep or mental health disorder aside from RLS. In particular we excluded any control who were abnormal on the Pittsburgh Sleep Quality Index (PSQI score >5). (9) In this study, sleep was recorded for eight hours for all subjects starting with their normal bedtimes. The recording on the first night included respiratory evaluations and any subject with apneahypopnea index ≥ 15 events/hr was excluded from the study. The second night recording used in this study did not include respiratory evaluations.

3.5 Micro-Analysis: agreement for each PLM

The EMG data for each recording was divided into thirds of the night. A 10-minute period of continuous non-rapid eye movement (non-REM) sleep state was randomly selected for detailed analyses from each of these thirds. In addition, 10-minute periods of continuous REM sleep were randomly selected from the first and last REM sleep periods. If the first REM period did not provide a large enough continuous sample, the subsequent REM periods were used.

The records from MATPLM1 and visually scored RemLogic-3 were time synched for each of the 10-minute periods chosen for analyses. The first 15 PLMS, or all PLMS if less than 15, marked in each period by either visual scoring or MATPLM1 were compared for overlap. The number of concurrent events was recorded along with those detected by only one of the scorers noting which identified the PLM. These were summed across all the sections and subjects for REM and NREM separately and combined (See Table 1). Figure 1 gives an example of the matched recordings used to evaluate overlap.

Table 1.

Display the agreement-disagreement matrix for detection of PLMS events by MATPLM1 and Technician (Visual) scoring techniques.

graphic file with name nihms-689922-f0006.jpg
Open in a new tab

The values in each cell represents the number of events (upper value) detected by one or both methods and the percentage (lower value) detected relative to the total number of events for all studies (694 events), only RLS studies (682 events), only control studies (12 events), only during REM sleep (227 events) or only during nonREM sleep (466 events)

MATPLM1+ and Visual+ signify that the method used detected these PLMS events

MATPLM1− and Visual− signify that the method used did not detect these PLMS event that was detected by the other

Figure 1.

Figure 1

Open in a new tab

Technician visual scoring misses significant events. The top tracing is from MATPLM1 and the bottom from REMLOGIC. The PLMS are marked by the top-most dark green bars for MATPLM1 and by purple highlight for REMLOGIC. The arrows show the PLMS that when measured carefully meet WASM criteria but were not identified by the visual scoring. (MATPLM1 marks the leg movements that are candidates for possible PLMS with the lower light green bars in the figure)

Each of the disagreements between MATPLM1 and technician scoring were carefully evaluated with detailed measurement for all periods from 5 randomly selected RLS subjects. These were analyzed to determine if the disagreement occurred because of technician or MATPLM1 detection error based on very strict application of the WASM criteria. Qualitative analyses were done to identify the major sources of disagreement between MATPLM1 and visual scoring.

3.6 Macro analyses; agreement for total PLMS from the recording

The total PLMS/hr for each of the PSGs was analyzed for the visual to the MATPLM1 scoring for all subjects based on the sleep technician scoring. The PLMS/hr scores from the detailed scoring by research expert scorer were similarly evaluated.

3.7 statistics

Agreements between data sets were analyzed using standard linear regressions and correlations, unequal variance t-tests for significant differences, and appropriate Bland-Altman plots to reveal bias (10, 11). The microanalyses permitted calculating the relative disagreement rates as percentages of total samples evaluated.

4. Results

4.1 Threshold comparisons

Independently, two scorers visually measured resting EMGs from 10 patients and had an inter-scorer reliability of r2 =0.90 and 90% differences ≤ ±2 μv. The MATPLM1 auto-threshold had an agreement with the average of the scorers of r2=0.83 and 90% differences ≤ ±1 μv. The auto-thresholds compared to visually-determined thresholds had an average lower resting EMG of 0.5 μV, well within the range of the inter-scorer agreement.

4.2 Micro analyses result

A total of 694 events were detected by either MATPLM1 or technician-dependent, visual scoring. The agreement between MATPLM1 and technician scoring for total 694 events was 87.7% (table 1- Visual+/MATPLM1+) with 91.7% agreement for control studies (12 events) and 87.6% for RLS studies (682 events). Fifty-five of the 664 PLM events (8.3%) identified by MATPLM1 were not identified by technician scoring; conversely, thirty of the 639 PLM events (4%) identified by the technician were not identified by MATPLM. The agreement between these two techniques was 91% for NREM-related events and 82% for REM-related events. In all cases MATPLM1 identified more PLMS events than technician scoring. The detailed analyses of all disagreements between MATPLM1 and technician visual scoring for the 5 randomly selected patients revealed that 35 of the 36 events (97%) occurred because of clear visual scorer errors in apply the WASM criteria. The one error identified with MATPLM1scoring involved a persisting high EMG value between two PLMS. MATPLM1 identified this as an event too long to be a PLMS while the technician correctly identified two separate movements. There were two major types of technician visual scorer errors noted. The most common was simply missing events as shown in figure 1. Technician errors also occurred in estimating duration of movements that led to accepting events as PLM that MATPLM1 rejected (See figure 2). The errors in duration could also be the reverse. Figure 3 shows a PLM detected by both MATPLM1 and visual scoring, but the visual scorer fails to identify the entire movement thus producing error for measurements of duration and IMI.

Figure 2.

Figure 2

Open in a new tab

Technician visual scoring accepts a movement that is too long (>10 seconds). The top tracing is from MATPLM1 and the bottom from REMLOGIC. MATPLM1 marks possible PLM with the lower green bar in this figure, but marks PLMS with a higher green bar seen in figure 1 but not here. This was rejected as a possible PLM by MATPLM1. REMLOGIC marks the PLMS with purple highlight. The total movement lasts about 12 seconds as shown by time marking added to the bottom trace.

Figure 3.

Figure 3

Open in a new tab

Technician visual scoring ends the movement too soon. The top tracing is from MATPLM1 and the bottom from REMLOGIC. The PLMS are marked by the top-most green bars for MATPLM1 and by purple highlight for REMLOGIC. In this case there is no error in number of events but there would be in a measure of duration of the event.

MATPLM1 and Expert-dependent, visual scoring of 414 events had a 94.4% agreement. Ten of the 401 events (2.5%) detected by MATLAB1 were not detected by the expert visual scoring. Conversely, 13 of the 404 events (3.2%) detected by the expert were not detected by MATLAB1. The disagreements between MATPLM1 and Expert scoring, when studies were re-evaluated, revealed in all cases that the expert scorer made errors in determining the length of the event, often by including short events that were very close but not long enough to meet the exact criteria for PLM events.

4.31 Macro-analyses result

As shown in figure 4A, MATPLM1 and technician scoring were highly correlated (r2= 0.97) but the visual scoring was consistently less than MATPLM1. Median (quartiles) of MATPLM1-technician scoring was 10.3 (4.9, 22.6). The linear regression line and Bland-Altman plots in figure 4B indicates this was a consistent difference generally increasing with increasing number of events. The rigorous scoring by the expert scorer largely corrects this bias as shown in figure 5. The linear regression of events detected by the expert versus MATPLM1 had a correlation of r2 = 0.99 with median (quartile) difference of −0.32 (−2.4, 0.28). The rigorous scoring tended to identify slightly more events than MATPLM1 except for the very high rates of PLMS when the expert score identified fewer events (figure 5).

Figure 4.

Figure 4

Open in a new tab

Panel A: Linear regression of MATPLM1 vs. technician visual PLMS/hr. Note the consistent tendency for lower visual PLMS rates. Panel B: Bland-Altman plot shows the visual compared to MATPLM1 scores were consistently less. The regression line and correlation indicates the visual scoring tended to be increasingly less for higher rates of PLMS

Figure 5.

Figure 5

Open in a new tab

Panel A: Linear regression of MATPLM1 vs. Research expert visual PLMS/hr. Note generally close relation r2= 0.99. The research expert compared to MATPLM1 showed a tendency to have slightly more PLMS except for the tracing with a very high rate of PLMS. This is shown in the Bland Altman plots in panel B.

5. Discussion

There are two major findings from this study. First, the MATPLM1 produces reliable and accurate scoring of PLMS that matches well with expert scoring. Second, the standard visual scoring of PLMS by technician tends to significantly under-score PLMS. Visual scoring errors were particularly large for RLS patients who had a large number of PLMS events. This seems hardly surprising. Scoring fatigue has to be a problem when visually checking hundreds of these small events over the entire night's sleep. Thus the more the events, the more the scorer is likely to miss them. This problem of scorer fatigue occurred even for the expert scorer as shown in figure 5. Additionally, there were several records for which the visual scorer reported no PLMS, but MATPLM1 identified a small number of events. A visual check of these records indicated the scorer appeared to create an expectation of no events and failed to detect sections with small events meeting the WASM criteria. Overall MATPLM1 performs better than visual scoring for obtaining an accurate assessment of PLMS from a sleep recording.

A major strength of this paper is the micro (event by event) analysis method used to evaluate accuracy of detection. It is conventional to validate scoring at a more macro level by looking at the summed results from different methods across an entire night. This, however, misses the likely possibility that in any data set there are different factors producing false positives and negatives for PLM events. These might largely cancel out in some data sets to produce a misleading concept that the methods are detecting the same events. Another data set may not have the same balance between errors and thus the methods will not apply. The ideal and preferred evaluation for assessing accuracy of an automatic PLM detector should be based on determining accuracy at the event level. Given the large number of events a reasonable alternative is to use a random sample of the events stratified for NREM and REM sleep and for time of night. This detailed micro analysis, as used in this paper, is equally or more important than using a large number of patients.

There are, however, three significant limitations of MATPLM1. First, it does not have a dynamic threshold adjustment, but uses the same threshold for the entire record. Most EMG recordings from quality sleep centers provide stable resting EMG throughout the night, but if not the case this program should be used with caution.. The current WASM standards allow for visual adjustment of resting EMG over the night but do not provide specific methods defining the process that could be implemented. This problem fortunately appears to be fairly rare.. The second concern is that MATPLM1 does not exclude episodes of fragmentary very brief EMG activity (e.g. fragmentary myoclonus) that can sometimes occur close enough together to meet the WASM criteria for a single leg movement. Some patients may have many of these producing episodes that would count as PLMS. This is a problem for the WASM criteria in general that needs to be addressed in future revisions of the criteria and then incorporated into MATPLM1.. The criteria for marking PLMS in relation to respiratory events is part of the MATPLM1 and it can be adjusted depending on preference regarding alternate criteria relating leg movements to apnea. (12). Marking the PLMS that occur with these events is, however, a separate second step taken after actual detection of the PLMS events themselves. Thus not marking these events does not impact the validation of the PLMS detection. The working concept here is not to reject PLMS occurring with respiratory events, but rather to mark them and allow separate or combined analyses as appropriate. PLMS associated with arousals will be similarly marked for appropriate analyses in a future version.

The results found in this study with visual scoring exemplify the problem of relying upon visual scoring for an accurate account of PLMS events. Agreement between labs and between data sets will always be a problem for visual scoring. Moreover, detailed measurement for strict and accurate application of the WASM criteria to records with lots of PLMS requires an excessive amount of time, more than is usually acceptable for sleep scoring. MATPLM1 in contrast can provide the detailed PLMS scoring from both legs (about 24 million data points at 500hz sampling) with graphic display and full data arrays within about a minute on most standard workstations. Thus the use of a standard program such as MATPLM1 offers improved speed, reliability and accuracy of PLMS measurements while strictly abiding to the WASM criteria. Ultimately, MATPLM1 or similar validated automatic scoring should be considered as a replacement for visual scoring, particularly for research studies.

There are 4 other automatic scoring programs that have been previously reported in the literature. There was no effort made in this study to compare results between these, partly because each program has somewhat different concepts. The two oldest methods by Tauchmann and Pollmächer (13) and Wetter et al (14) were written in JAVA and did not use the current WASM criteria. The Ferri et al (15) was the first program to apply the basic concept of two thresholds (onset and offset of events) used in the WASM criteria, but they used fixed values for the thresholds for all records checking to ensure the resting EMG was at 2 μv for their recordings. The WASM criteria, however, require adjusting the thresholds based on the resting EMG in each record. MATPLM1 provides an automatic determination of EMG resting level for each record as needed to apply the WASM criteria. The most recent automatic scoring program by Moore, et al (16) has been very well developed to match as much as possible results of visual scoring. The visual scoring was based mostly on American Academy of Sleep Medicine (AASM) criteria published in 2007. They evaluated their program against a wide range of patients in several studies. They make several sophisticated modifications of the EMG signal to reduce recording problems with recording artifacts often seen in low quality EMG records. They also use a dynamic threshold based on a moving 20-second interval. Our experience with intervals that long, however, indicates problems for some RLS patients who will have no 20-second period without significant EMG events. Developing a program to match visual scoring from a large clinical data base is also a problem because of the significant disagreement between scorers in different clinical settings, the poor quality EMG recordings sometimes obtained in clinical recordings and the errors visual scorers make, especially for large numbers of PLMS. More fundamentally the concept behind the WASM criteria was to explicitly define the PLMS based upon the EMG signals without requiring visual judgment aside from that of the resting threshold. Thus the approach for developing MATPLM1 was to leave the basic EMG signal intact as much as possible and to rigorously apply the WASM criteria rather than adjusting the scoring to match visual judgments. It is noted that none of the prior automatic PLMS scoring programs have gained acceptance outside of use where they were developed. The major problem has been the prior programs were not made easily accessible. The most accessible of these prior programs may be the MATLAB program developed by Moore et al, but this program is embedded in a larger set of routines. It can, however, be separately identified and downloaded from the internet. .MATPLM1 will be made available free from the Hopkins RLS center. It, however, currently requires a MATLAB license. The MATLAB license is available for free in most academic centers, but is otherwise expensive. Future expansion of MATLAB1 is planned to include adjustments to be complaint with GNU OCTAVE making it available free of charge.

There are some significant advantages to MATPLM1. First it is available as a MATLAB function that provides a full range of evaluations of PLMS characteristics including the periodicity index. This function runs automatically when given the data in a suitable format and produces a MATLAB array for the results that can be copied into an excel sheet. It can also run in one step a group of data sets from several subjects and put out the array with data for each subject in separate rows. It produces MATLAB numeric outputs of the PLMS/hr and periodicity index, a MALAB graphic display that can be visually evaluated at different resolutions and data arrays that permit further analyses of significant PLMS features including inter-movement intervals, time of night, sleep stage, different minimums and durations of events (17). The arrays and the graphs can be saved for further analyses. Thus this program supports research and clinical evaluations of PLMS features in any study that can provide European Data Format EMG tracings with sleep stage scoring.

SUMMARY.

These results validate MATPLM1 as accurate for scoring PLMS following the WASM criteria. The usual sleep technician scoring, in contrast, often fails to correctly identify the PLMS based on the WASM criteria, particularly when there are a large number of events to be measured as possible PLMS. The results thus demonstrate that a validated automatic scoring such as MATPLM1 should be preferred to a technician visual scoring. MATPLM1 is available as a MATLAB script that can be applied to data in European data format (EDF). There is also available an EDF to MATLAB data converter that translates the EDF data into the MATLAB structure used in this program. This opens the process for others to use this program and also enables further evaluation and program adjustments. MATPLM1 as it stands works significantly better than visual scoring for good quality recordings where the resting EMG is stable for most of the night and where sleep disordered breathing is not a significant factor.

Highlights.

  1. MATPLM1 a MATLAB script detects PLMS from both legs based on WASM scoring criteria.

  2. MATPLM1 is accurate compared to detailed expert visual scoring of PLMS.

  3. The usual visual scoring by trained technicians significantly underscores PLMS.

Abbreviation

EDF

European Data Format

EMG

Electromyography

IRB

Institutional Review Board

IMI

Inter-movement interval

IRLSSG scale

International Restless Legs Syndrome Study Group Severity Scale

LM

Leg Movement

PAM-RL

Physical Activity Monitor for Restless Legs

PLM

Periodic Leg movements

PLMS

Periodic limb Movements of sleep

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  • 1.Allen RP, Picchietti DL, Garcia-Borreguero D, Ondo WG, Walters AS, Winkelman JW, et al. Restless legs syndrome/Willis-Ekbom disease diagnostic criteria: updated International Restless Legs Syndrome Study Group (IRLSSG) consensus criteria--history, rationale, description, and significance. Sleep Med. 2014;15(8):860–73. doi: 10.1016/j.sleep.2014.03.025. [DOI] [PubMed] [Google Scholar]
  • 2.Montplaisir J, Michaud M, Denesle R, Gosselin A. Periodic leg movements are not more prevalent in insomnia or hypersomnia but are specifically associated with sleep disorders involving a dopaminergic impairment. Sleep Med. 2000;1(2):163–7. doi: 10.1016/s1389-9457(00)00014-9. [DOI] [PubMed] [Google Scholar]
  • 3.Pennestri MH, Whittom S, Adam B, Petit D, Carrier J, Montplaisir J. PLMS and PLMW in healthy subjects as a function of age: prevalence and interval distribution. Sleep. 2006;29(9):1183–7. doi: 10.1093/sleep/29.9.1183. [DOI] [PubMed] [Google Scholar]
  • 4.Zucconi M, Ferri R, Allen R, Baier PC, Bruni O, Chokroverty S, et al. The official World Association of Sleep Medicine (WASM) standards for recording and scoring periodic leg movements in sleep (PLMS) and wakefulness (PLMW) developed in collaboration with a task force from the International Restless Legs Syndrome Study Group (IRLSSG). Sleep Med. 2006;7(2):175–83. doi: 10.1016/j.sleep.2006.01.001. [DOI] [PubMed] [Google Scholar]
  • 5.Ferri R. The time structure of leg movement activity during sleep: The theory behind the practice. Sleep Medicine. 2012;13(4):433–41. doi: 10.1016/j.sleep.2011.10.027. [DOI] [PubMed] [Google Scholar]
  • 6.Gschliesser V, Frauscher B, Brandauer E, Kohnen R, Ulmer H, Poewe W, et al. PLM detection by actigraphy compared to polysomnography: a validation and comparison of two actigraphs. Sleep Med. 2009;10(3):306–11. doi: 10.1016/j.sleep.2008.03.015. [DOI] [PubMed] [Google Scholar]
  • 7.Kobayashi M, Namba K, Ito E, Nishida S, Nakamura M, Ueki Y, et al. The validity of the PAM-RL device for evaluating periodic limb movements in sleep and an investigation on night-to-night variability of periodic limb movements during sleep in patients with restless legs syndrome or periodic limb movement disorder using this system. Sleep Medicine. (0) doi: 10.1016/j.sleep.2013.08.790. [DOI] [PubMed] [Google Scholar]
  • 8.Sforza E, Johannes M, Claudio B. The PAM-RL ambulatory device for detection of periodic leg movements: a validation study. Sleep Med. 2005;6(5):407–13. doi: 10.1016/j.sleep.2005.01.004. [DOI] [PubMed] [Google Scholar]
  • 9.Buysse DJ, Reynolds CF, 3rd, Monk TH, Berman SR, Kupfer DJ. The Pittsburgh Sleep Quality Index: a new instrument for psychiatric practice and research. Psychiatry Res. 1989;28(2):193–213. doi: 10.1016/0165-1781(89)90047-4. [DOI] [PubMed] [Google Scholar]
  • 10.Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1(8476):307–10. [PubMed] [Google Scholar]
  • 11.Bland JM, Altman DG. Comparing methods of measurement: why plotting difference against standard method is misleading. Lancet. 1995;346(8982):1085–7. doi: 10.1016/s0140-6736(95)91748-9. [DOI] [PubMed] [Google Scholar]
  • 12.Manconi M, Zavalko I, Fanfulla F, Winkelman JW, Fulda S. An Evidence-Based Recommendation for a New Definition of Respiratory-Related Leg Movements. Sleep. 2014 doi: 10.5665/sleep.4418. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Tauchmann N, Pollmacher T. Automatic detection of periodic leg movements (PLM). J Sleep Res. 1996;5(4):273–5. doi: 10.1111/j.1365-2869.1996.00273.x. [DOI] [PubMed] [Google Scholar]
  • 14.Wetter TC, Dirlich G, Streit J, Trenkwalder C, Schuld A, Pollmacher T. An automatic method for scoring leg movements in polygraphic sleep recordings and its validity in comparison to visual scoring. Sleep. 2004;27(2):324–8. doi: 10.1093/sleep/27.2.324. [DOI] [PubMed] [Google Scholar]
  • 15.Ferri R, Zucconi M, Manconi M, Bruni O, Miano S, Plazzi G, et al. Computer-assisted detection of nocturnal leg motor activity in patients with restless legs syndrome and periodic leg movements during sleep. Sleep. 2005;28(8):998–1004. doi: 10.1093/sleep/28.8.998. [DOI] [PubMed] [Google Scholar]
  • 16.Moore H, Leary E, Lee S-Y, Carrillo O, Stubbs R, Peppard P, et al. Design and Validation of a Periodic Leg Movement Detector. PLoS ONE. 2014;9(12):e114565. doi: 10.1371/journal.pone.0114565. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Ferri R, Manconi M, Lanuzza B, Cosentino FI, Bruni O, Ferini-Strambi L, et al. Age-related changes in periodic leg movements during sleep in patients with restless legs syndrome. Sleep Med. 2008;9(7):790–8. doi: 10.1016/j.sleep.2007.08.020. [DOI] [PubMed] [Google Scholar]

RESOURCES