Journal Logo

Integrative Systems

Cigarette craving modulation is more feasible than resistance modulation for heavy cigarette smokers: empirical evidence from functional MRI data

Kim, Dong-Youla; Tegethoff, Marionb,,c; Meinlschmidt, Guntherd,,e,,f; Yoo, Seung-Schikg; Lee, Jong-Hwana

Author Information
doi: 10.1097/WNR.0000000000001653
  • Open

Abstract

Introduction

Reducing the rates of tobacco smoking is one of the most urgent health priorities because smoking is a major cause of preventable deaths and diseases [1]. Despite the wide range of strategies that have been employed to promote tobacco/cigarette smoking cessation, including tobacco control policies (e.g. tax increases, reimbursement and telephone-based quitlines) [2], the introduction of noncombustible tobacco products (e.g. smokeless tobacco products, e-cigarettes and dissolvable tobacco products) [3–5] and pharmacotherapeutic options (e.g. nicotine acetylcholine receptors) [6], about one in five US adults still smoke [3].

Recent neuroimaging-based intervention efforts have raised the possibility of modulating the strength of cigarette cravings elicited by visual cues via noninvasive real-time functional MRI (fMRI)-based neurofeedback (fMRI-NF) [7–10]. In these studies, the anterior cingulate cortex (ACC) and medial prefrontal cortex (MPFC) areas are defined as regions-of-interest (ROIs) implicated in cigarette craving and resistance, respectively [8]. This is because previous fMRI studies have identified distinct neuronal underpinnings for cigarette smoking and craving, [11–17] such as in the ACC [13–16], posterior cingulate cortex (PCC) [11–14] and precuneus [11–13]. For example, hyper-activations in the dorsal part of the medial prefrontal area for smoking resistance and in the ventral part of the medial prefrontal area, including the ACC, for smoking craving have been reported [8]. Hypo-activations have also been reported in part of the ACC and dorsolateral prefrontal cortex for smoking craving [9,18]. Consequently, smokers have received real-time fMRI-NF (rtfMRI-NF) training to gain volitional control over neuronal activity in these ROIs and have learned to regulate their neuronal activity levels to reduce their cigarette cravings [7,8].

Previous studies have also reported clear differences in functional connectivity, which refers to co-activation between brain regions, of smokers and nonsmokers [19–21]. For example, Fedota et al. [20] investigated changes in the functional connectivity of smokers and nonsmokers in response to a cognitive control task and reported lower functional connectivity between the right anterior insula and the right superior frontal gyrus in smokers. Similarly, Bi et al. [19] reported that the resting-state functional connectivity of the bilateral insula differed between smokers and nonsmokers. However, no research has explicitly compared the efficacy of using functional connectivity patterns as a neurofeedback signal with that of conventional neuronal activations.

Based on these studies, we conducted a novel method of rtfMRI-NF training that utilized information on the functional connectivity between ROIs as the basis for the neurofeedback signal (i.e. functional connectivity-added rtfMRI-NF) to reduce cigarette craving and compared this with traditional activity-based rtfMRI-NF [22]. The aim of the present study was to quantitatively evaluate the efficacy of functional connectivity-added rtfMRI-NF training using non-neurofeedback fMRI data in terms of the change in subjective craving scores (CRSs) obtained while participants craved or resisted cigarette smoking without neurofeedback information.

Materials and methods

Participants

The Institutional Review Board of Korea University approved the entire study protocol. The participants included male right-handed treatment-seeking heavy cigarette smokers (N = 14; see [22] for their sociodemographic and cognitive information) who were not claustrophobic and who provided informed written consent. The self-reported inclusion criteria were (a) an Fagerström Test for Nicotine Dependence (FTND) score >4 [23], (b) a smoking history >5 years, (c) >10 cigarettes per day, (d) an exhaled carbon monoxide (CO) level on the interview day of >15 (piCO Smokerlyzer; Bedfont Scientific, Ltd., Rochester, UK), (e) the absence of any neurological or mental disorders, (f) no current use of a nicotine-replacement substance other than cigarettes and (g) the absence of any current nicotine addiction therapy, including bupropion, varenicline and nortriptyline [7,8,13,24]. During the fMRI runs, the participants were asked to report any issues, including discomfort, when lying down in a supine position by squeezing a ball with their hand to alert the MRI technician and staff. We also monitored the participants’ status after each fMRI run and decided whether to continue with the subsequent fMRI run or not because the inside of an MRI bore can be uncomfortable and discomfort may disrupt the results.

rtfMRI runs for neurofeedback training

An updated version of our in-house rtfMRI-NF software toolbox implemented in a MATLAB environment was used [22,25–27]. Brain ROIs related to cigarette craving and cigarette resistance were defined in the anterior (i.e. ACC and MPFC) and posterior regions (i.e. PCC and precuneus), respectively [7–9,13]. These anterior and posterior ROIs were defined using standard anatomical templates [i.e. automated anatomical labeling (AAL) and Brodmann’s area] from the registration of a subject’s first echo-planar imaging (EPI) volume to Montreal Neurological Institute (MNI) space. The templates and spatial normalization algorithms available in SPM8 were used. Table 1 summarizes the anatomical information for the ROIs. The corresponding regions defined by the AAL and Brodmann’s area atlases were intersected to obtain fine-grained ROIs for each target brain region [25]. Registration and the definition of the ROIs were conducted at the beginning of the rtfMRI-NF session during the calibration and fixation periods in each rtfMRI-NF run (45 s) before the neurofeedback signal was provided to the participants. The size of the anterior and posterior ROIs (mean ± SD of the number of voxels within an ROI) was 193.2 ± 12.1 and 73.8 ± 8.9, respectively.

Table 1 - Definition of the regions-of-interest (ROIs) using automated anatomical labeling (AAL) and Brodmann’s area templates. The regions defined using the AAL and Brodmann’s area templates were intersected and defined as the anterior and posterior ROIs employed in real-time functional-neurofeedback (rtfMRI-NF) training
Anterior ROI Posterior ROI
AAL L/R mOFC (25, 26), L/R ACC (31, 32) L/R PCC (35, 36), L/R PrCN (67, 68)
BA BA 10, 11, 12, 24, 32, 33 BA 23, 26, 29, 30, 31
The ROI number for the AAL template was obtained from MRIcron (https://www.nitrc.org/projects/mricron).
ACC, anterior cingulate cortex; L, left; mOFC, medial orbitofrontal cortex; PCC, posterior cingulate cortex; PrCN, precuneus; R, right.

Activity-based rtfMRI-NF was randomly assigned to half of the participants, with the neurofeedback signal defined using the median percentage of the blood-oxygenation-level-dependent (BOLD) intensity within the anterior ROIs [22]. Our proposed rtfMRI-NF with functional connectivity capability was assigned to the remaining subjects, with the neurofeedback signal defined based on the average of (a) the median percentage of the BOLD intensity within the anterior ROIs and (b) the functional connectivity measured using Pearson’s correlation coefficient between the BOLD signals from the anterior and posterior ROIs (i.e. functional connectivity-added neurofeedback) (Fig. 1). The resulting percentage of the BOLD intensity for the anterior ROIs was restricted between −1 and 1% to match the functional connectivity for the anterior and posterior ROIs. The neurofeedback signals were mapped to the opacity gradient for smoking scenes in a video clip, in which a maximum value of +1 made the video opaque, whereas zero or negative values did not alter the opacity of the video. The AlphaData attribute in the image object of the video scene was updated to change the opacity in a MATLAB environment. The computational time required to extract the neurofeedback signal for each EPI volume was less than a second in our computing environment (Intel Core i5 2.4 GHz, 8 GB RAM, 256 GB SSD and Windows 7). Thus, there was no backlog due to unprocessed EPI volumes.

Fig. 1
Fig. 1:
Overall flow diagram of the adopted experimental paradigm, which consists of nonneurofeedback functional MRI (fMRI) and real-time functional MRI-neurofeedback (rtfMRI-NF) runs over two visits. A pair of nonneurofeedback fMRI runs of cigarette craving or resistance were acquired before and after the rtfMRI-NF runs. In each rtfMRI-NF run, once the first EPI volume had been registered to the automated anatomical labeling and Brodmann’s area atlases, the anterior/posterior regions-of-interest (ROIs) implicated in cigarette craving and resistance were defined. The neuronal activation level of each echo-planar imaging volume was defined using the median blood-oxygenation-level-dependent (BOLD) intensity across the voxels in each ROI. Functional connectivity was defined using the Pearson’s correlation coefficient between the BOLD signals of the anterior and posterior ROIs. ACC, anterior cingulate cortex; mOFC, medial orbitofrontal cortex; mPFC, medial prefrontal cortex; PCC, posterior cingulate cortex.

Participants were asked to make the scenes opaque when increasing their resistance against the urge to smoke. They practiced controlling the neurofeedback interface by changing the opacity of a video of someone smoking a cigarette. More specifically, the participants were informed that the video stimuli would darken as their smoking resistance increased or their smoking craving reduced. They were given time to learn how to increase their resistance to smoking cravings using the video by freely up- or downregulating their resistance level. Information regarding the mapping between the opacity of the video and their brain signals was unknown to the participants. Due to the time delay in the neurophysiological hemodynamic coupling between the neuronal activations and the BOLD signal, participants were informed that their brain signals would be reflected in the video opacity about 5–6 s later.

Participants engaged in two independent rtfMRI-NF sessions 1 week apart. There were six rtfMRI-NF training runs in each of the two sessions (a total of 12 rtfMRI-NF runs) and each rtfMRI-NF run was fixed at the same duration (258 s). All 14 subjects finished all 12 rtfMRI-NF runs. More detailed information regarding the rtfMRI-NF training runs can be found in our earlier report [22]. The average period of abstinence, FTND scores, number of years smoking, number of cigarettes per day and CO levels on the interview day did not differ between the two neurofeedback groups [22].

Non-neurofeedback fMRI runs to evaluate cigarette craving modulation

Participants performed a pair of two non-neurofeedback fMRI runs (i.e. a craving and a resistance run) before and after the rtfMRI-NF training, with a total of four pairs of craving and resistance runs across the two rtfMRI-NF visits (Fig. 1). In each non-neurofeedback fMRI run, pseudo-random alternating and counterbalanced blocks (15 s per block) of (a) cigarette smoking images and (b) nonsmoking images interleaved by fixation images as a baseline block were presented to the participants via visual goggles [12]. The participants were asked to crave a cigarette when presented with the smoking images during the craving run and to resist the urge to smoke during the resistance run. The order of the craving and resistance runs was pseudo-randomized. After each non-neurofeedback fMRI run, participants reported their cigarette CRS(1 being the minimum, 10 being the maximum) using a magnetic resonance-compatible button response (Current Design, Pennsylvania, USA; www.curdes.com).

Imaging parameters

A 3-Tesla MRI scanner (Tim Trio, Siemens, Erlangen, Germany) and a 12-channel head coil were used to measure neuronal activation levels on the basis of the BOLD intensity acquired from a gradient-echo EPI pulse sequence (repetition time = 2000 ms; echo time = 30 ms; field of view = 24 × 24 cm2; in-plane voxel = 64 × 64; voxel size = 3.75 × 3.75 × 4 mm3; flip-angle = 90°; 36 interleaved slices without a gap) [22].

Data analysis

The CRSs recorded for the two visits were analyzed using SPSS as the dependent variable using a mixed analysis of variance (ANOVA) test with the visit as the within-subject independent variable and the neurofeedback training condition/group (i.e. activity-based neurofeedback vs. functional connectivity-added neurofeedback) as the between-subject independent variable (IBM SPSS Statistics 21, New York, USA). In detail, a two-way mixed ANOVA test was conducted on the CRS of subject i for the jth visit, with Yij = μ + ri + fj + eij, where ri is the random effect of group factor r at level i (i.e. between-subject), fj is the fixed effect of visit factor f at level j (i.e. within-subject), μ is the grand mean and eij is the residual error. For the post hoc multiple comparison corrections, the observed means were tested using Tukey’s test for the homogeneity of within-group variance and a studentized residual for outlier detection among the residuals. The main effects for the group and visit factors and their interaction were also reported.

Non-neurofeedback fMRI runs were preprocessed using SPM8 software (www.fil.ion.ucl.ac.uk/spm), followed by a general linear model (GLM) [12,28]. In detail, hemodynamic response function models for the smoking image and nonsmoking image blocks were used as regressors of the GLM. Thus, BOLD responses for the smoking image blocks and nonsmoking image blocks that were considerably higher compared to the baseline fixation block were evaluated using the beta value estimates of the regressor for each of the corresponding blocks. The neuronal activations (i.e. the beta values) from the two regressors of the smoking and nonsmoking images were then subtracted from each of the voxels and considered to be the neuronal activation levels associated with the desire to smoke.

During the first visit, 14 non-neurofeedback fMRI runs (across the 14 participants) for the craving and resistance conditions were conducted before the rtfMRI-NF training and were thus not influenced by this training. Thus, neuronal activity levels that were significantly different (P < 0.05, with a minimum of 10 voxels defined as a voxel cluster) between the craving and resistance conditions across the participants were calculated voxel-wise using a random-effects model via paired t-tests. Brain areas that exhibited significantly increased neuronal activations (P < 0.05) under the craving or resistance conditions were used as an explicit mask in the paired t-tests. The average beta values for each of the voxel clusters identified from the paired t-tests were calculated across four non-neurofeedback fMRI runs for each of the craving and resistance conditions, and across seven participants for each rtfMRI-NF training group. These beta values and the participants’ CRSs were further analyzed using linear regression to identify voxel clusters where the corresponding neuronal activation levels were modulated by rtfMRI-NF training.

Results

Brain regions implicated in either craving or resistance

Figure 2 shows that 10 voxel clusters exhibited greater neuronal activity levels for the craving condition when compared to the resistance condition, and these were found mostly in the visual areas, such as the middle/inferior occipital gyri, calcarine and cuneus. In contrast, four voxel clusters in the angular gyrus and (medial) superior frontal gyrus showed significantly greater neuronal activation levels for the resistance condition compared to the craving condition (Table 2).

Table 2 - Brain regions (i.e. voxel clusters) that represent statistically different neuronal activation levels (i.e. beta values) for each of the cigarette craving or cigarette resistance conditions compared to the other conditions, identified using the first nonneurofeedback functional MRI (fMRI) run from the 14 subjects during the first visit
Cluster index Brain region (AAL/BA) Size x y z Peak t-score
Craving > Resistance
1 FG, Cb 18 −18 −43 −14 3.20
2 MOG, IOG, MTG/BA37, 19 13 −54 −73 −2 3.18
3 Olfactory, ACC 12 12 32 −2 4.02
4 Insula/BA13 23 −30 26 −2 3.83
5 Calcarine 17 27 −76 4 4.32
6 Calcarine, cuneus/BA 31 17 −6 −70 19 2.67
7 Cuneus, SOG/BA 19 15 −9 −85 28 4.01
8 SOG, cunues/BA 19, 7 19 −15 −85 40 5.18
9 SPL, PoCG/BA 7, 5 93 33 −49 64 4.19
10 SPL/BA 5 18 −30 −46 64 3.14
Craving < Resistance
1 Cb 72 −30 −79 −32 4.36
2 MTG 14 54 −34 −2 3.80
3 AG, SMG 21 −51 −61 31 3.16
4 mSFG, SFG/BA 8, 6 34 12 29 64 4.18
AAL, automated anatomical labeling; ACC, anterior cingulate cortex; AG, angular gyrus; BA, Brodmann’s area(Size is denoted as the number of voxels); Cb, cerebellum; FG, fusiform gyrus; IOG, inferior occipital gyrus; MOG, middle occipital gyrus; mSFG, medial superior frontal gyrus; MTG, middle temporal gyrus; PoCG, postcentral gyrus; SFG, superior frontal gyrus; SMG, supramarginal gyrus; SOG, superior occipital gyrus; SPL, superior parietal lobule.

Fig. 2
Fig. 2:
Brain regions (i.e. voxel clusters) that represent statistically different (P < 0.05, with a minimum of 10 voxels) neuronal activation levels (i.e. beta values) between the cigarette craving (i.e. CRV) and resistance (i.e. RST) conditions during nonneurofeedback functional MRI runs, acquired before real-time functional MRI-neurofeedback training during the first visit, were defined as regions-of-interest. Bar graphs of the beta values from each of the two conditions across all 14 subjects are shown for each cluster.

Craving score modulation

Figure 3 shows the CRSs across the four non-neurofeedback fMRI runs in the craving or resistance conditions from the two visits. In the craving condition, although there was no significant main effect for the neurofeedback training and non-RT fMRI runs, there was a significant interaction [F (3,36) = 5.506; P = 0.003; effect size η2 = 0.164]. Further analysis of this interaction indicated that there was a significant decrease across the four-time points for the functional connectivity-added neurofeedback training only [F (3,10) = 5.534; P = 0.017; effect size η2 = 0.138]. In addition, there was a significant difference across the two types of neurofeedback training in the fourth run [F (1,12) = 5.297; P = 0.040; effect size d = 1.05]. On the other hand, using the CRSs from the resistance runs, there was no significant main effect and no significant interaction.

Fig. 3
Fig. 3:
Craving scores across the four nonneurofeedback functional MRI runs obtained from each of the two types of real-time functional-neurofeedback training (i.e. activity-based and functional connectivity-added neurofeedback training) and from each of the craving or resistance condition. FC, functional connectivity; NF, neurofeedback.

Voxel clusters whose neuronal activation levels are associated with craving scores

Figure 4 shows that the voxel clusters identified in the craving condition (in the fusiform gyrus, insula, calcarine, cuneus, superior occipital gyrus and superior parietal lobule) demonstrated a statistically significant correlation (P < 0.05) between the beta values (i.e. the neuronal activation levels) and the CRSs in the functional connectivity-added, but not in the activity-based, neurofeedback training. However, there was no statistically significant correlation between the beta values of the voxel clusters identified in the resistance condition and the CRSs.

Fig. 4
Fig. 4:
Statistically significant correlation between the craving scores and neuronal activation levels (i.e. the beta values) was obtained only from the craving-related voxel clusters in the functional connectivity-added neurofeedback training. FC, functional connectivity; NF, neurofeedback; PoCG, postcentral gyrus; SOG, superior occipital gyrus; SPL, superior parietal lobule.

Discussion

In this study, we reported that functional connectivity-added rtfMRI-NF facilitated the reduction of CRSs among nicotine-deprived heavy cigarette smokers. Furthermore, the CRSs were positively correlated with the neuronal activation levels of the brain regions implicated in cigarette craving but not in those implicated in cigarette resistance. To our knowledge, there has been no systematic analysis to date of the use of fMRI data to evaluate the efficacy of rtfMRI-NF training in the modulation of cigarette craving or cigarette resistance. Our findings that neuronal activation levels in cigarette craving regions, but not those in resistance regions, were correlated with CRSs is in line with a previously reported rtfMRI-NF study [8]. This indicates that craving modulation is more feasible than resistance modulation for heavy cigarette smokers when employing the rtfMRI-NF modality.

The fact that the majority of voxel clusters (i.e. six out of ten) associated with craving conditions were found in the visual areas is in agreement with a recent review that asserts that the visual areas are crucial in addiction research using visual cues/stimuli [29]. The insula may be implicated in cigarette craving due to interoceptive awareness of cigarettes during the craving process [13]. The strong increase in neuronal activation in parts of the frontoparietal network or the cognitive control network, such as in the (medial) superior frontal areas, angular gyrus and supramarginal gyrus, in the resistance rather than the craving condition, is also in line with previous studies [11,13]. During rtfMRI-NF training, multiple brain regions and functional networks may have been involved as ROIs because of the presence of functional processes such as external cognitive control and interoceptive responses in addition to the ROIs directly related to the target conditions, such as smoking addiction, in our study [30]. Thus, multivariate models that accommodate the information from multiple ROIs or functional networks, such as mediation analysis [31] or machine-learning deep-learning models [32–35] would be beneficial to further enhance the efficacy of rtfMRI-NF training.

In our study, the ROIs used in the neurofeedback runs were defined using anatomical atlases and the warping process from the EPI volume to the standard MNI template, which may have resulted in misalignment due to an imperfect registration algorithm [36]. In addition, the ROIs could be further divided into subregions on the basis of distinct functional information. For example, the ACC can be subdivided into subgenual, rostral and dorsal regions [37,38]. Therefore, fine-grained ROIs on the basis of functional information and anatomical boundaries can further enhance the reliability of the neurofeedback information provided to the participants [39–41]. It is also important to note that hyper- and hypo-activation have been found in the proximity of brain regions that are associated with smoking craving or resistance, as described in a previous review article [42]. Thus, the delineation of individual-specific ROIs is warranted to accommodate the potential variability in neuronal associations with smoking craving or resistance.

It should be noted that the present study has a number of limitations, including that nonsmokers were not recruited as a control group in our study. This was because our main goal was to explore a potential alternative/supplementary noninvasive treatment option for treatment-seeking heavy smokers. Nonetheless, the inclusion of a nonsmoker group by stratifying the rtfMRI-NF training conditions with activity-based neurofeedback and functional connectivity-added neurofeedback would provide supporting information on the neurophysiological substrates of smoking addiction for heavy cigarette smokers in comparison to nonsmokers in the context of rtfMRI-NF training. In addition, alternative smoking craving assessment measures to the CRS, such as the Questionnaire on Smoking Urges, the Minnesota Withdrawal Scale and the Shiffman–Jarvik Withdrawal Scale – Short Version, may provide richer information on the association between neuronal activation and subjective craving levels. Finally, only male smokers were recruited in our study so future research that includes female heavy cigarette smokers can extend the findings to the general population.

In general, our empirical findings should also be interpreted carefully because of (a) the relatively small number of participants (although our sample size is comparable to those of previous rtfMRI-NF clinical studies [7]) and (b) the use of subjective CRSs as the sole indicator of cigarette craving levels, rather than incorporating objective indicators, such as the CO levels in exhaled breath or the nicotine levels in the blood [7,14]. Future research could investigate the possibility that the independent regulation of functional connectivity levels only could lead to the clear modulation of neuronal activation or functional connectivity levels associated with cigarette craving or resistance. Further research is also warranted to systematically evaluate our empirical findings using a large sample and more frequent rtfMRI-NF visits [8], while also taking into account the severity of cigarette/nicotine addiction [7,15] and assessing the duration of cigarette smoking cessation to identify the potential for a ceiling effect [11]. Only 1 of the 14 participants reported discomfort during the fMRI acquisition process, partly because of our efforts to maximize their comfort during each fMRI run and because each MRI session was less than an hour. Additional cushions for head and body support, such as memory foam, may further enhance the comfort of the participants.

Our proposed rtfMRI-NF method, which includes functional connectivity information in addition to neuronal activity, may be potentially useful for the analysis or treatment of neuropsychiatric/mental disorders other than smoking addiction. This is because neuropsychiatric/mental disorders such as schizophrenia and attention deficit hyperactivity disorder have also been reported to involve aberrant neuronal connectivity patterns across brain regions [40,43,44]. For example, Hoffman et al.[43] reported reduced cortical connectivity in schizophrenic patients due to a suppressed mesocortico-limbic dopaminergic system. In another study, Suskauer et al. [44] reported aberrant functional connectivity patterns between the rostral supplementary motor area and the anterior prefrontal areas in children with attention deficit hyperactivity disorder [44]. Applications related to sensorimotor and cognitive dysfunctions due to neuropsychiatric disorders could also potentially benefit from employing rtfMRI-NF informed by functional network-based or seed region-based connectivity, as opposed to regional activity-based neurofeedback [40,45].

Conclusions

In this study, we provide evidence for the efficacy of rtfMRI-NF with functional connectivity capability in reducing cigarette craving, particularly based on data from craving modulation runs (as opposed to craving resistance runs) using non-neurofeedback fMRI data acquired before and after rtfMRI-NF training. Based on our empirical findings, our proposed functional connectivity-added rtfMRI-NF method has the potential for use in rtfMRI-NF training for cigarette smoking cessation, thus complementing tobacco control policies, noncombustible tobacco products and pharmacotherapeutic and behavioral options.

Acknowledgements

The authors would like to thank Mr. Han-Gil Lee, Hojung Kang, Ms. Minkyung Oh, Eunkyung Jung, and Drs. Yong-Hwan Kim, Junghoe Kim and Hyun-Chul Kim for their logistic support, helpful discussion and comments.

This work was supported by the National Research Foundation (NRF) grant, MSIP of Korea (NRF-2015R1A2A2A03004462, NRF-2016M3C7A1914450 and NRF-2017R1E1A1A01077288), and in part by the National Research Council of Science & Technology (NST) grant by the Korea government (MSIT) [No. CAP-18-01-KIST]. These sponsors were not involved in the study design, data collection, analysis or interpretation of data, manuscript preparation, or the decision to submit for publication.

Conflicts of interest

There are no conflicts of interest.

References

1. Kulik MC, Menvielle G, Eikemo TA, Bopp M, Jasilionis D, Kulhánová I, et al.; EURO-GBD-SE Consortium. Educational inequalities in three smoking-related causes of death in 18 European populations. Nicotine Tob Res. 2014; 16:507–518.
2. Brown T, Platt S, Amos A. Equity impact of population-level interventions and policies to reduce smoking in adults: a systematic review. Drug Alcohol Depend. 2014; 138:7–16.
3. Fagerstrom K, Rutqvist LE, Hughes JR. Snus as a smoking cessation aid: a randomized placebo-controlled trial. Nicotine Tob Res. 2012; 14:306–312.
4. Kasza KA, Bansal-Travers M, O’Connor RJ, Compton WM, Kettermann A, Borek N, et al. Cigarette smokers’ use of unconventional tobacco products and associations with quitting activity: findings from the ITC-4 U.S. cohort. Nicotine Tob Res. 2014; 16:672–681.
5. Mishina EV, Hoffman AC. Clinical pharmacology research strategy for dissolvable tobacco products. Nicotine Tob Res. 2014; 16:253–262.
6. Bierut LJ. Nicotine dependence and genetic variation in the nicotinic receptors. Drug Alcohol Depend. 2009; 104 (Suppl 1):S64–S69.
7. Canterberry M, Hanlon CA, Hartwell KJ, Li X, Owens M, LeMatty T, et al. Sustained reduction of nicotine craving with real-time neurofeedback: exploring the role of severity of dependence. Nicotine Tob Res. 2013; 15:2120–2124.
8. Hanlon CA, Hartwell KJ, Canterberry M, Li X, Owens M, Lematty T, et al. Reduction of cue-induced craving through realtime neurofeedback in nicotine users: the role of region of interest selection and multiple visits. Psychiatry Res. 2013; 213:79–81.
9. Li X, Hartwell KJ, Borckardt J, Prisciandaro JJ, Saladin ME, Morgan PS, et al. Volitional reduction of anterior cingulate cortex activity produces decreased cue craving in smoking cessation: a preliminary real-time fMRI study. Addict Biol. 2013; 18:739–748.
10. Qi Y. Random forest for bioinformatics. Zhang C, Ma Y, editors. In: Ensemble Machine Learning. Springer; 2012. pp. 307–323.
11. Brody AL, Mandelkern MA, Olmstead RE, Jou J, Tiongson E, Allen V, et al. Neural substrates of resisting craving during cigarette cue exposure. Biol Psychiatry. 2007; 62:642–651.
12. Lee JH, Kim DY, Kim J. Mesocorticolimbic hyperactivity of deprived smokers and brain imaging. Neuroreport. 2012; 23:1039–1043.
13. Hartwell KJ, Johnson KA, Li X, Myrick H, LeMatty T, George MS, Brady KT. Neural correlates of craving and resisting craving for tobacco in nicotine dependent smokers. Addict Biol. 2011; 16:654–666.
14. Azizian A, Nestor LJ, Payer D, Monterosso JR, Brody AL, London ED. Smoking reduces conflict-related anterior cingulate activity in abstinent cigarette smokers performing a Stroop task. Neuropsychopharmacology. 2010; 35:775–782.
15. Smolka MN, Bühler M, Klein S, Zimmermann U, Mann K, Heinz A, Braus DF. Severity of nicotine dependence modulates cue-induced brain activity in regions involved in motor preparation and imagery. Psychopharmacology (Berl). 2006; 184:577–588.
16. Due DL, Huettel SA, Hall WG, Rubin DC. Activation in mesolimbic and visuospatial neural circuits elicited by smoking cues: evidence from functional magnetic resonance imaging. Am J Psychiatry. 2002; 159:954–960.
17. Ernst M, Matochik JA, Heishman SJ, Van Horn JD, Jons PH, Henningfield JE, London ED. Effect of nicotine on brain activation during performance of a working memory task. Proc Natl Acad Sci USA. 2001; 98:4728–4733.
18. Karch S, Paolini M, Gschwendtner S, Jeanty H, Reckenfelderbäumer A, Yaseen O, et al. Real-time fMRI neurofeedback in patients with tobacco use disorder during smoking cessation: functional differences and implications of the first training session in regard to future abstinence or relapse. Front Hum Neurosci. 2019; 13:65.
19. Bi Y, Yuan K, Guan Y, Cheng J, Zhang Y, Li Y, et al. Altered resting state functional connectivity of anterior insula in young smokers. Brain Imaging Behav. 2017; 11:155–165.
20. Fedota JR, Matous AL, Salmeron BJ, Gu H, Ross TJ, Stein EA. Insula demonstrates a non-linear response to varying demand for cognitive control and weaker resting connectivity with the executive control network in smokers. Neuropsychopharmacology. 2016; 41:2557–2565.
21. Fedota JR, Stein EA. Resting-state functional connectivity and nicotine addiction: prospects for biomarker development. Ann N Y Acad Sci. 2015; 1349:64–82.
22. Kim DY, Yoo SS, Tegethoff M, Meinlschmidt G, Lee JH. The inclusion of functional connectivity information into fMRI-based neurofeedback improves its efficacy in the reduction of cigarette cravings. J Cogn Neurosci. 2015; 27:1552–1572.
23. Heatherton TF, Kozlowski LT, Frecker RC, Fagerström KO. The fagerström test for nicotine dependence: a revision of the fagerström tolerance questionnaire. Br J Addict. 1991; 86:1119–1127.
24. Brody AL, Mandelkern MA, Lee G, Smith E, Sadeghi M, Saxena S, et al. Attenuation of cue-induced cigarette craving and anterior cingulate cortex activation in bupropion-treated smokers: a preliminary study. Psychiatry Res. 2004; 130:269–281.
25. Lee JH, O’Leary HM, Park H, Jolesz FA, Yoo SS. Atlas-based multichannel monitoring of functional MRI signals in real-time: automated approach. Hum Brain Mapp. 2008; 29:157–166.
26. Lee JH, Ryu J, Jolesz FA, Cho ZH, Yoo SS. Brain-machine interface via real-time fMRI: preliminary study on thought-controlled robotic arm. Neurosci Lett. 2009; 450:1–6.
27. Lee JH, Kim J, Yoo SS. Real-time fMRI-based neurofeedback reinforces causality of attention networks. Neurosci Res. 2012; 72:347–354.
28. Worsley KJ, Friston KJ. Analysis of fMRI time-series revisited–again. Neuroimage. 1995; 2:173–181.
29. Hanlon CA, Dowdle LT, Naselaris T, Canterberry M, Cortese BM. Visual cortex activation to drug cues: a meta-analysis of functional neuroimaging papers in addiction and substance abuse literature. Drug Alcohol Depend. 2014; 143:206–212.
30. Sitaram R, Ros T, Stoeckel L, Haller S, Scharnowski F, Lewis-Peacock J, et al. Closed-loop brain training: the science of neurofeedback. Nat Rev Neurosci. 2017; 18:86–100.
31. Kim HC, Tegethoff M, Meinlschmidt G, Stalujanis E, Belardi A, Jo S, et al. Mediation analysis of triple networks revealed functional feature of mindfulness from real-time fMRI neurofeedback. Neuroimage. 2019; 195:409–432.
32. Vu H, Kim HC, Jung M, Lee JH. fMRI volume classification using a 3D convolutional neural network robust to shifted and scaled neuronal activations. Neuroimage. 2020; 223:117328.
33. Kim HC, Bandettini PA, Lee JH. Deep neural network predicts emotional responses of the human brain from functional magnetic resonance imaging. Neuroimage. 2019; 186:607–627.
34. Kim J, Calhoun VD, Shim E, Lee JH. Deep neural network with weight sparsity control and pre-training extracts hierarchical features and enhances classification performance: evidence from whole-brain resting-state functional connectivity patterns of schizophrenia. Neuroimage. 2016; 124:127–146.
35. Jang H, Plis SM, Calhoun VD, Lee JH. Task-specific feature extraction and classification of fMRI volumes using a deep neural network initialized with a deep belief network: evaluation using sensorimotor tasks. Neuroimage. 2017; 145:314–328.
36. Klein A, Andersson J, Ardekani BA, Ashburner J, Avants B, Chiang MC, et al. Evaluation of 14 nonlinear deformation algorithms applied to human brain MRI registration. Neuroimage. 2009; 46:786–802.
37. Margulies DS, Kelly AM, Uddin LQ, Biswal BB, Castellanos FX, Milham MP. Mapping the functional connectivity of anterior cingulate cortex. Neuroimage. 2007; 37:579–588.
38. Torta DM, Cauda F. Different functions in the cingulate cortex, a meta-analytic connectivity modeling study. Neuroimage. 2011; 56:2157–2172.
39. Hamilton JP, Glover GH, Hsu JJ, Johnson RF, Gotlib IH. Modulation of subgenual anterior cingulate cortex activity with real-time neurofeedback. Hum Brain Mapp. 2011; 32:22–31.
40. Sulzer J, Haller S, Scharnowski F, Weiskopf N, Birbaumer N, Blefari ML, et al. Real-time fMRI neurofeedback: progress and challenges. Neuroimage. 2013; 76:386–399.
41. Veit R, Singh V, Sitaram R, Caria A, Rauss K, Birbaumer N. Using real-time fMRI to learn voluntary regulation of the anterior insula in the presence of threat-related stimuli. Soc Cogn Affect Neurosci. 2012; 7:623–634.
42. Tursic A, Eck J, Lührs M, Linden DEJ, Goebel R. A systematic review of fMRI neurofeedback reporting and effects in clinical populations. Neuroimage Clin. 2020; 28:102496.
43. Hoffman RE, Rapaport J, Ameli R, McGlashan TH, Harcherik D, Servan-Schreiber D. A neural network simulation of hallucinated “voices” and associated speech perception impairments in schizophrenic patients. J Cogn Neurosci. 1995; 7:479–496.
44. Suskauer SJ, Simmonds DJ, Fotedar S, Blankner JG, Pekar JJ, Denckla MB, Mostofsky SH. Functional magnetic resonance imaging evidence for abnormalities in response selection in attention deficit hyperactivity disorder: differences in activation associated with response inhibition but not habitual motor response. J Cogn Neurosci. 2008; 20:478–493.
45. Ruiz S, Buyukturkoglu K, Rana M, Birbaumer N, Sitaram R. Real-time fMRI brain computer interfaces: self-regulation of single brain regions to networks. Biol Psychol. 2014; 95:4–20.
Keywords:

cigarette craving; cigarette resistance; functional MRI; neurofeedback; real-time fMRI neurofeedback

Copyright © 2021 The Author(s). Published by Wolters Kluwer Health, Inc.