Consistency of fNIRS parameters: A parallel forms and test-retest reliability study

• ¹ İzmir University of Economics, Türkiye
• ² Ankara University, Türkiye

To address the need to explore the functional activity in the human brain various data have been collected by using different neuroimaging techniques. No matter what technique is used, without reliable results no study can effectively contribute to the assessment of this functional activity. Thus, to get similar results when the experiment is repeated has become an important issue to be investigated for the researchers who work with these techniques. Thus, recent studies on reliability of brain imaging data mostly focused on test-retest reliability that investigates stability of the measures over time. In present study, along with test-retest reliability alternate or parallel forms reliability which is one of the primary classifications of psychometric reliability will be used to assess the consistency of the results of two tasks/tests constructed in the same way at a single point in time. Hence, the consistency of brain imaging data will be obtained both across different forms and different times. Fourteen healthy non-smoker right-handed participants were recruited for the test-retest reliability of hemodynamic measures. The mean age of the participants was 20.64 (SDage= 1.34) and the age range was 18-24. Participants, who took part in the first session of the study, participated again in the second session after 3 weeks (Mretest interval = 20.36 days, SD = 3.14) as Plichta et al. suggested. For the alternate forms reliability, 13 healthy non-smoker right-handed participantswith a mean age of 20.38 (SDage = 0.51) took part in the study. Verbal version of n-back was used as a working memory task. Letters (6 letters (D, G, K, M, T, and V) for test-retest reliability, different 6 letters (F, H, L, R, Y, and C) for the alternate forms reliability) were used as stimuli that were obtained from a web page which generates random letters.N-back task consisting of these letters was presented to participants in three trials. Each trial consisted of four levels of cognitive work-load which were 0-,1-,2-, and 3-back. Therefore, each participant saw 12 n-back conditions in total. In 0-back condition, participants were required to press left click if the letter that appeared on the screen was the pre-specified target letter and press right click for the other letters, in other words non-target letters. In 1-back condition, the target letter was any letter identical to the one immediately preceding it.In 2- and 3-back conditions, the target letter was any letter that was identical to the one presented two and three letters ago, respectively. Each n-back condition included 30 letters, 10 of them were target letters, and 20 of them were non-target letters. Hence, 120 letters were presented to each participant in total. Prior to each n-back condition, 1500 millisecond (msec) resting period was arranged. During this period, participants were required nothing but resting in order to measure their baseline light intensity via fNIRS. Following each resting period, prior to each n-back condition, 1 second beep sound was presented on a white screen in order to take attention of participants. After the beep sound, a single letter appeared on the center of the screen for 500 msec. Inter-stimulus-interval (ISI) was 2500 msec and inter-trail-interval (ITI) was 4000 msec. The order of the presentation for the n-back conditions were randomized based on the Latin Square block design. Accuracy scores and reaction time, as behavioral, oxy-Hb, deoxy-Hb, total-Hb, and oxy-change measures, as hemodynamic measurements were obtained. The same experimental procedure was conducted for the second session of the test-retest and alternate forms reliability investigations. Hemodynamic activity of the prefrontal cortex of the participants during the experimental task was assessed by a 16-channel continuous wave fNIR200A stand-alone functional brain imaging system. Before the analysis of fNIRS data, each participants’ raw light intensity data was refined to eliminate artifacts by applying a low-pass filter with a finite impulse response (FIR) and a linear phase filter with an order 20 and cut-off frequency of 0.1 Hz. To refine data from motion artifacts that cause abrupt spikes in fNIR signals, Sliding-window Motion Artifact Rejection algorithm (SMAR) was applied to previously FIR filtered data.After time synchronization markers were determined, refined fNIR light data were converted to hemodynamic responses by using modified Beer-Lambert Law (mBLL) with respect to 15000 msec resting periods at the beginning of each n-back condition to obtain oxy-hemoglobin, deoxy-hemoglobin, total hemoglobin, and oxy-change parameters. Hemodynamic activity at each channel was averaged for each n-back conditions across three trials to provide an average hemodynamic activity.For each channel,reliability estimates of hemodynamic measures obtained via fNIRS (oxy-Hb, deoxy-Hb, total-Hb, oxygenation change) were obtained by using the parameter of intraclass correlation coefficient (ICC).ICC is a commonly used reliability parameter in the field of behavioral science and neuroimaging studies that ranges from 0 (unreliable) to 1 (excellent reliability).In this study, single measure ICCs were calculated for estimating reliability in terms of absolute agreement, ICC(2,1), and consistency, ICC(3,1), of fNIR parameters in test-retest and alternate forms types of reliability. For the oxy-Hb measures, between two testing sessions; fair (0.40-0.59) to good (0.60-¬0.74) agreement and consistency were found for Channels 2, 3, 4, 6, 7, 8, 9, 10, 12, 14 and 16; between two forms, fair to good agreement and fair to excellent (0.75-1.00) consistency were found for Channels 3, 4, 5, 6, 7, 9, 10, 11, 12, 13, and 14.For the oxy-change measures, between two testing session; fair to excellent agreement and consistency were found for Channels 3, 4, 5,6, 7, 8, 9, 10, 11, 12, 13, 14, and 16; between two forms fair to excellent agreement and consistency were obtained for Channels 3, 4, 5, 6, 7, 9, 10, 11, 12, 13, and 14. Present study is the first study to utilize alternative forms reliability in search of consistency of the hemodynamic measures by using parallel forms of a task. In addition to this, test-retest reliability estimates were obtained for all of the fNIR parameters. Based on overall findings, it can be concluded that oxy-Hb and oxygenation measures were stable; but deoxy-Hb and total-Hb measures were not so stable across two sessions and versions.