The primary targets of interest were study populations consisting of term born infants and young children with focus on examining brain growth and development using MRI. A literature search using PubMed database was originally conducted on the 30th of June in 2016. The search comprised of the following keywords (‘Magnetic Resonance Imaging’[Mesh] OR ‘MR imaging^∗’ OR ‘MRI’ OR ‘NMRI’ OR ‘fMRI’ OR ‘DTI’ OR ‘diffusion tensor imaging’) AND (‘Brain/growth and development’[Mesh] OR ‘brain growth^∗’ OR ‘brain developm^∗’) AND (‘Infant’[Mesh] OR ‘infant^∗’ OR ‘toddler^∗’). No languages were excluded at this point. To capture the most recent and relevant work in the field, a starting date limit was enforced to include only papers published after the 1st of January in 2012. The search was updated on the 9th of March in 2021 and the final search included literature published between 1st of January in 2012 to 1st of January in 2021. After duplicates were removed, the search resulted in a total of 1098 publications.

Titles and abstracts were used to screen articles in the first phase. Exclusion criteria were the following, in a descending order of priority:

If a publication met more than one criterion, only the highest criterion was marked as a reason for exclusion. If the exclusion criterion was found from the title, the abstract was not used to find a higher priority criterion for exclusion. 721 articles were screened out. 377 publications were identified as potentially relevant (Figure 1).

These 377 publications were reviewed based on the abstracts and full texts. At this second phase, we first excluded all review articles and after that, the exclusion criteria (1–5) were applied as in the first phase. Given that we were particularly interested in studies using healthy infants and young children who underwent MRI without sedation (noting that preterm-born children are frequently scanned under anesthesia), only articles that met the following criteria were included:

All information was obtained from the article full-texts and their supplementary data when applicable. While we were interested in the methods to calm subjects in the scanner, we also investigated how studies have reported the used procedures before and at the scanner. Thus, reviewed studies may contain overlapping participant populations. 291 publications did not meet the criteria and were excluded.

Finally, a total of 86 original articles published in English were identified and included in this review. The included studies are listed in Supplementary Table 1. While we acknowledge that the inclusion of a longer time frame and studies with prematurely born participants would provide somewhat more information on the topic, we opted to include the most recent studies performed without clinical grounds.

For all the included study populations (n = 86) MRI scans were performed on the subjects between 0 and 6 years of age. The number of participants per one study ranged from 9 to 288. Majority of the subjects were term born, but in some cases gestational age at birth was not provided. Sample sizes and ages at scan are shown in Supplementary Table 1. Fifty-two studies were completely cross-sectional and performed only one scan per each subject (Hazlett et al., 2012; Deniz Can et al., 2013; Deoni et al., 2013, 2015; Gao et al., 2013; O’Muircheartaigh et al., 2013, 2014; Qiu et al., 2013a, 2015b; Broekman et al., 2014; Dean et al., 2014b, c, 2017, 2018a,b; Grewen et al., 2014; Travis et al., 2014; Zhang et al., 2014; Donald et al., 2015; Langer et al., 2015; Ou et al., 2015; Poh et al., 2015; Spann et al., 2015a, b, 2020a,b; Ferradal et al., 2016, 2019; Li et al., 2016; Sethna et al., 2016; Adibpour et al., 2018, 2020; Lugo-Candelas et al., 2018; Monnelly et al., 2018; Chen et al., 2019; Hernandez-Castillo et al., 2019; Lebenberg et al., 2019; Lehtola et al., 2019; Ong et al., 2019; Tuulari et al., 2019; Acosta et al., 2020a, b, 2021; Alexander et al., 2020; Bruchhage et al., 2020; Camacho et al., 2020; Dowe et al., 2020; Fenchel et al., 2020; Gale-Grant et al., 2020; Graham et al., 2020; Merhar et al., 2020; Merz et al., 2020). The remaining studies (n = 34) were longitudinal and conducted serial scans on the same individuals (Geng et al., 2012, 2016; Choe et al., 2013; Li et al., 2013, 2014a,b,c, 2015a,b; Qiu et al., 2013b; Sadeghi et al., 2013; Alcauter et al., 2014, 2015; Bompard et al., 2014; Chen et al., 2014; Dean et al., 2014a, 2015a,b; Gao et al., 2014a, b; Swanson et al., 2015; Chang et al., 2016; Croteau-Chonka et al., 2016; Deoni et al., 2016; Kim et al., 2016; Meng et al., 2017; Ahn et al., 2019; Dai et al., 2019a, b; Hu et al., 2019; Wang et al., 2019, 2012; Remer et al., 2020; Schmied et al., 2020). Serial scans included two to seven scans per subject, most typically three scans per subject. The majority of the studies (n = 70) used 3 Tesla MRI scanners (Geng et al., 2012, 2016; Hazlett et al., 2012; Deoni et al., 2013, 2015; Gao et al., 2013, 2014a,b; Li et al., 2013, 2014a,b,c, 2015a,b; Sadeghi et al., 2013; Alcauter et al., 2014, 2015; Bompard et al., 2014; Chen et al., 2014, 2019; Dean et al., 2014a,b,c, 2015a,b, 2017, 2018a,b; Grewen et al., 2014; Zhang et al., 2014; Donald et al., 2015; Langer et al., 2015; Spann et al., 2015a,b, 2020a,b; Swanson et al., 2015; Chang et al., 2016; Croteau-Chonka et al., 2016; Ferradal et al., 2016, 2019; Kim et al., 2016; Meng et al., 2017; Adibpour et al., 2018, 2020; Lugo-Candelas et al., 2018; Monnelly et al., 2018; Ahn et al., 2019; Dai et al., 2019a,b; Hernandez-Castillo et al., 2019; Hu et al., 2019; Lebenberg et al., 2019; Lehtola et al., 2019; Tuulari et al., 2019; Wang et al., 2019, 2012; Acosta et al., 2020a, b, 2021; Alexander et al., 2020; Bruchhage et al., 2020; Camacho et al., 2020; Dowe et al., 2020; Fenchel et al., 2020; Gale-Grant et al., 2020; Merhar et al., 2020; Merz et al., 2020; Remer et al., 2020; Schmied et al., 2020), while 1.5 Tesla scanners were less commonly used (n = 12) (Choe et al., 2013; Deniz Can et al., 2013; Qiu et al., 2013a, b, 2015b; Broekman et al., 2014; Travis et al., 2014; Ou et al., 2015; Poh et al., 2015; Li et al., 2016; Sethna et al., 2016; Ong et al., 2019). Four studies did not report the field strength of the used MR scanner (O’Muircheartaigh et al., 2013, 2014; Deoni et al., 2016; Graham et al., 2020).

All subjects underwent the MRI scanning non-sedated. Infants and children under the age of 4 slept during acquisition in majority of studies (n = 73/86) (Acosta et al., 2020a, b, 2021; Adibpour et al., 2018, 2020; Ahn et al., 2019; Alcauter et al., 2014, 2015; Alexander et al., 2020; Bompard et al., 2014; Broekman et al., 2014; Bruchhage et al., 2020; Camacho et al., 2020; Chang et al., 2016; Chen et al., 2019; Choe et al., 2013; Croteau-Chonka et al., 2016; Dai et al., 2019a, b; Dean et al., 2014a,b,c, 2015a,b, 2017, 2018a,b; Deniz Can et al., 2013; Deoni et al., 2013, 2015, 2016; Donald et al., 2015; Dowe et al., 2020; Ferradal et al., 2016, 2019; Gale-Grant et al., 2020; Gao et al., 2013, 2014a,b; Geng et al., 2016; Graham et al., 2020; Grewen et al., 2014; Hazlett et al., 2012; Hernandez-Castillo et al., 2019; Hu et al., 2019; Kim et al., 2016; Langer et al., 2015; Lebenberg et al., 2019; Lehtola et al., 2019; Li et al., 2014b, 2015a,b, 2016; Lugo-Candelas et al., 2018; Meng et al., 2017; Merhar et al., 2020; Merz et al., 2020; Monnelly et al., 2018; O’Muircheartaigh et al., 2013, 2014; Ong et al., 2019; Poh et al., 2015; Qiu et al., 2013a, b, 2015b; Remer et al., 2020; Sadeghi et al., 2013; Schmied et al., 2020; Sethna et al., 2016; Swanson et al., 2015; Tuulari et al., 2019; Wang et al., 2019; Zhang et al., 2014). Seven studies reported that infants were given time to fall asleep before scanning, but the subject’s state during the scan was not specifically reported (Geng et al., 2012; Chen et al., 2014; Travis et al., 2014; Spann et al., 2015a,b, 2020a,b). One study reported that infants were scanned during natural sleep or while resting quietly (Ferradal et al., 2016). Furthermore, a few studies (n = 6) reported scans without sedation, but subject’s state during the scan was not specifically reported (Wang et al., 2012; Li et al., 2013, 2014a,c; Ou et al., 2015; Fenchel et al., 2020). See Table 1 for a summary.

Fifteen studies of all the total 86 studies in this review scanned also children between the ages 4 to 6 years (O’Muircheartaigh et al., 2013, 2014; Chen et al., 2014, 2019; Dean et al., 2014a,c, 2015a,b; Deoni et al., 2015, 2016, 2013; Croteau-Chonka et al., 2016; Dai et al., 2019a, b; Remer et al., 2020). At this age, the child has typically more ability to cooperate, but in contrast to infants, it might be more difficult to get the child to fall asleep in a strange environment (see Table 2). Eleven studies reported that if tolerated by the child (in most cases ≥ 4 years old), the scan was made while the child was awake, e.g., when watching a movie or a TV show. The remaining four studies scanned during natural sleep. Otherwise preparations, motion prevention, sound attenuation, and monitoring during the scan did not differ from that of the younger children.

Timing of the MRI sessions were frequently scheduled according to participant’s normal sleeping/diurnal rhythm. Scanning schedules were reported in 14 studies (Choe et al., 2013; Deniz Can et al., 2013; Dean et al., 2014a, b, 2017, 2018b; Gao et al., 2014b; Travis et al., 2014; Langer et al., 2015; Lehtola et al., 2019; Tuulari et al., 2019; Bruchhage et al., 2020; Camacho et al., 2020; Dowe et al., 2020), and the remainder 72 studies did not report the timing of the visit. Imaging was frequently performed on naptimes or bedtime. In seven studies, MRI visits were scheduled for the subject’s naptime or bedtime, but the time was not specified (Gao et al., 2014b; Langer et al., 2015; Dean et al., 2017, 2018b; Bruchhage et al., 2020; Camacho et al., 2020; Dowe et al., 2020). One study scheduled visits for naptime in the late morning (Deniz Can et al., 2013) and three others in the late afternoon or early evening (Choe et al., 2013; Lehtola et al., 2019; Tuulari et al., 2019). In a few studies (n = 3), the majority or all the scans were performed in the evening around bedtime or at night (Dean et al., 2014a, b; Travis et al., 2014). Naptime imaging was typically used with the youngest participants, while evening hours and nighttime were typical scan times not only for infants, but also for the older participants.

Getting ready for the upcoming MRI scan often started already at home. To habituate the infant to the scanner noise, families were provided a CD of the scanner sound, and parents were instructed to play the CD, while subjects were sleeping at home (Hazlett et al., 2012; Langer et al., 2015). In one study, mothers were given an MRI prep kit including earplugs and a portable speaker pre-loaded with MRI sounds. In this study, mothers were also encouraged to start swaddling their infant to sleep if they had not already started to do so (Camacho et al., 2020). In two studies, the child was deprived of sleep prior to scans by asking parents to wake the child up earlier in the morning or skip a nap on a research day (Geng et al., 2012; Dean et al., 2014a). To promote sleep at the imaging site, Dean et al. (2014a) reported that parents were asked to keep the child busy throughout the day before the scan. In total, 81 studies did not report preparations at home.

Replicating the typical bed or naptime routines in the MRI facilities was reported in five studies (Choe et al., 2013; Deniz Can et al., 2013; Dean et al., 2014a; Langer et al., 2015; Camacho et al., 2020). Parents were asked to bring along comfort items to create a homely environment at the imaging site (Langer et al., 2015). Private rooms with diaper changing and bathing facilities, rocking chairs, portacribs, blankets, soft lullaby music, and other objects were attempted to make the environment cozier (Choe et al., 2013; Deniz Can et al., 2013; Dean et al., 2014a; Langer et al., 2015). Dimmed lights at the MRI site and also around the facility, when carrying the sleeping child, were also provided (Dean et al., 2014a; Dowe et al., 2020). While creating a comfortable and homely environment, the children got to familiarize with the MRI sound simultaneously before the scanning and when they fell asleep in five studies (Langer et al., 2015; Spann et al., 2015a, b, 2020a,b). Langer et al. (2015) reported the use of interesting tasks prior to the scan to fatigue the child.

Feeding the child before scanning was the most commonly reported preparation (n = 32/86) (Geng et al., 2012; Wang et al., 2012; Gao et al., 2013, 2014a; Li et al., 2013, 2014a,c, 2016; Sadeghi et al., 2013; Alcauter et al., 2014, 2015; Bompard et al., 2014; Chen et al., 2014; Grewen et al., 2014; Zhang et al., 2014; Donald et al., 2015; Ou et al., 2015; Spann et al., 2015a, b, 2020a,b; Kim et al., 2016; Dean et al., 2017, 2018b; Ahn et al., 2019; Hernandez-Castillo et al., 2019; Lehtola et al., 2019; Tuulari et al., 2019; Alexander et al., 2020; Camacho et al., 2020; Dowe et al., 2020; Merhar et al., 2020). Sleep was promoted by adjusting a feeding schedule prior to the scan. After children were fed, they were swaddled or wrapped, or otherwise helped to fall asleep. A few studies explained this so called feed and sleep or feed and wrap method in more detail: children were fed 15–30 min prior to the scan and swaddled in warm sheets (Ou et al., 2015; Li et al., 2016).

Swaddling or wrapping was used not only to make falling asleep easier, but also to reduce potential body movement during the scan. A number of different approaches to wrapping the infants were provided, varying from only wrapping them in sheets to placing them into an immobilizer. Several studies (n = 23) used specific vacuum immobilization mats, bags, or pillows to stabilize the child and reduce natural movement from breathing (Deoni et al., 2013; Zhang et al., 2014; Ou et al., 2015; Croteau-Chonka et al., 2016; Li et al., 2016; Dean et al., 2014a,b,c, 2015b, 2017, 2018a,b; Dai et al., 2019a, b; Hernandez-Castillo et al., 2019; Lehtola et al., 2019; Ong et al., 2019; Tuulari et al., 2019; Camacho et al., 2020; Dowe et al., 2020; Graham et al., 2020; Merhar et al., 2020; Remer et al., 2020). Dean et al. (2014a) reported placing a mat under the child before the child fell asleep and once asleep the immobilizer was wrapped around the child. Subject’s head was separately secured in a vacuum fixation device in eight studies (Geng et al., 2012, 2016; Grewen et al., 2014; Li et al., 2015a, b; Ahn et al., 2019; Wang et al., 2019, 2012), while only half of these also reported swaddling the child. Foam cushions, foam pads, and visco-elastic matters were also commonly used to keep the head in place and occupy the space between the subjects and the head coil. All in all, the majority of studies (n = 51/86) mentioned some method to stabilize the infant prior to the scan.

Acoustic noise levels of the MRI scanner were reduced, and the hearing of the subjects was protected with different methods. The major part of the studies (n = 65/86) reported the use of passive or active measures during acquisition, while 21 studies made no mention of sound attenuation. Six studies out of these 65 reporting noise attenuation mentioned taking precautions to reduce the noise, but did not delineate the methods (Qiu et al., 2013a, b, 2015b; Broekman et al., 2014; Poh et al., 2015; Ong et al., 2019). Most commonly used passive measure was ear protection: 25 studies used double or triple ear protection (Deniz Can et al., 2013; Deoni et al., 2013, 2015, 2016; Dean et al., 2014a,c, 2017, 2018a,b; Donald et al., 2015; Spann et al., 2015a, b, 2020a,b; Li et al., 2016; Monnelly et al., 2018; Dai et al., 2019a,b; Hernandez-Castillo et al., 2019; Lehtola et al., 2019; Tuulari et al., 2019; Alexander et al., 2020; Camacho et al., 2020; Dowe et al., 2020; Graham et al., 2020) and the remainder (n = 33) used single ear protection. For example, MiniMuffs, earplugs, headphones, sound attenuating ear protectors, electrodynamic headphones and a custom-made acoustic hood were used. In four studies, electrodynamic headphones played white noise (Dean et al., 2017, 2018a; Camacho et al., 2020; Dowe et al., 2020) and in one study soothing rain sounds (Bruchhage et al., 2020) during image acquisition. In addition to ear protection, noise levels were lessened by a noise insulating bore liner or foam insert fitted inside of the scanner bore in 17 studies (Deoni et al., 2013, 2015, 2016; Dean et al., 2014a,b,c, 2015b, 2017, 2018a,b; Croteau-Chonka et al., 2016; Dai et al., 2019a,b; Lebenberg et al., 2019; Bruchhage et al., 2020; Dowe et al., 2020; Remer et al., 2020). Furthermore, some studies reported reducing scanner noise actively by selecting specific imaging parameters, slowing the gradient switching rate and reducing the maximum gradient amplitudes (Deoni et al., 2013, 2015, 2016; O’Muircheartaigh et al., 2013, 2014; Dean et al., 2014a,b,c, 2015b, 2017, 2018a,b; Croteau-Chonka et al., 2016; Dai et al., 2019a,b; Lehtola et al., 2019; Tuulari et al., 2019; Adibpour et al., 2020; Dowe et al., 2020; Remer et al., 2020). The changes in the imaging parameters were shown to provide approximately a 35 dB noise reduction during the acquisition (Dean et al., 2014a).

Numerous studies (n = 39/86) mentioned monitoring subjects throughout the scan. To confirm that the child remained asleep, a physician, a nurse, a research assistant or a member of the research team was presented, who visually monitored the subject in 26 studies (Geng et al., 2012; Wang et al., 2012; Qiu et al., 2013a, b, 2015b; Broekman et al., 2014; Dean et al., 2014a, 2017, 2018b; Grewen et al., 2014; Li et al., 2014b, 2015b; Donald et al., 2015; Poh et al., 2015; Chang et al., 2016; Croteau-Chonka et al., 2016; Meng et al., 2017; Monnelly et al., 2018; Hernandez-Castillo et al., 2019; Lehtola et al., 2019; Ong et al., 2019; Tuulari et al., 2019; Bruchhage et al., 2020; Camacho et al., 2020; Dowe et al., 2020; Gale-Grant et al., 2020). Three studies reported using an MRI compatible camera (Ou et al., 2015; Bruchhage et al., 2020; Graham et al., 2020), and in additional five studies an infrared camera to monitor subjects during the scan (Deoni et al., 2013; Dean et al., 2014c; Croteau-Chonka et al., 2016; Dai et al., 2019a, b). Besides visual monitoring, 28 studies used a pulse oximeter to follow the heart rate and oxygen saturation (Geng et al., 2012; Wang et al., 2012; Deoni et al., 2013; Qiu et al., 2013a, b, 2015b; Broekman et al., 2014; Dean et al., 2014a, c; Grewen et al., 2014; Li et al., 2014b, 2015b, 2016; Donald et al., 2015; Ou et al., 2015; Poh et al., 2015; Spann et al., 2015a, b, 2020a,b; Croteau-Chonka et al., 2016; Ferradal et al., 2016; Meng et al., 2017; Monnelly et al., 2018; Dai et al., 2019a, b; Hu et al., 2019; Ong et al., 2019). Bruchhage et al. (2020) reported using pulse socks to monitor pulse and behavior when asleep. Electrocardiography (ECG) was used in four studies (Spann et al., 2015a, b, 2020a,b). One study reported monitoring heart rate, oxygen saturation, and temperature, but did not specify the used equipment (Gale-Grant et al., 2020). In addition to these monitoring methods made by a professional or a member of a research team, only six studies reported parents being invited to remain in the imaging site during the acquisition (Dean et al., 2014a, 2017, 2018b; Lehtola et al., 2019; Tuulari et al., 2019; Dowe et al., 2020).

The majority of studies did not directly report the total duration of the imaging protocol. The acquisition time could often be calculated using given sequences and the number of planes, this however, does not tell the total/maximum time in the scanner. Most of the studies reported using different imaging sequences (e.g., T1, T2, DTI, fMRI), but did not report the individual acquisition times for them. When reported, the acquisition times ranged from 2 min (Chen et al., 2019) up to 2 h (for a typical MRI session) (Ferradal et al., 2016), trying to keep them short to prevent the child from waking up during the acquisition (Dean et al., 2014a). The child’s cooperation and ability to remain asleep enabled additional imaging sequences and longer imaging times (Dean et al., 2015b). In some studies, the imaging times varied depending on the subject’s age due to different protocols used in various age groups (Deoni et al., 2013, 2016; Dean et al., 2014c; O’Muircheartaigh et al., 2014; Dai et al., 2019a, b). Finally, only one study reported the duration of the whole scanning visit (Dean et al., 2014a). In this study, the duration of the visit was highly variable from less than 1 h to more than 5 h.

All 86 studies had successful MRI scans and the number of included scans varied between 9 (Zhang et al., 2014) to 445 (Dai et al., 2019b). In addition to the successful scans, reporting about data losses it varied considerably between studies. 41 studies (n = 41/86) reported the number of excluded scans or alternatively the success rate for included data (Hazlett et al., 2012; Deniz Can et al., 2013; Qiu et al., 2013a, b, 2015b; Sadeghi et al., 2013; Bompard et al., 2014; Broekman et al., 2014; Dean et al., 2014a, b, 2017, 2018a,b; Li et al., 2014c, 2016; Travis et al., 2014; Alcauter et al., 2015; Donald et al., 2015; Langer et al., 2015; Poh et al., 2015; Spann et al., 2015a, 2020b; Chang et al., 2016; Ferradal et al., 2016, 2019; Sethna et al., 2016; Adibpour et al., 2018, 2020; Lugo-Candelas et al., 2018; Dai et al., 2019a; Ong et al., 2019; Tuulari et al., 2019; Acosta et al., 2020a,b, 2021; Bruchhage et al., 2020; Camacho et al., 2020; Dowe et al., 2020; Fenchel et al., 2020; Graham et al., 2020; Merz et al., 2020). Additionally, some studies mentioned reasons for exclusion during the data preprocessing steps but did not report the number of excluded scans. Remainder of the studies used only subjects with successful scans or for some other reason did not report the number of missing scans. The number of included and excluded scans, and if available, reasons for exclusion are shown in Supplementary Table 1. The most common specified reason for losing data was movement during the scan causing motion artifacts to the data. Other mentioned reasons related to the scanning procedure were that the subject did not fall asleep prior to scanning, woke up during transition to the scanning bed or during the MRI acquisition. Exclusions were made also due to demographic reasons, problems with the analysis, age under or over the study-specific range, missed measurements in other parts of the study, and due to a brain anatomical anomaly. In other words, not all exclusion criteria were a consequence of failure at imaging. Due to the different reasons of exclusion, the success rates between studies are not comparable and do not represent exclusively the success of scanning. Instead of the overall success rate, some studies reported separate rates for different parts of the sample, for example, for age groups or for cases and controls separately. For example, Li et al. (2014c) used data acquired at three time points with following success rates: 90% for neonates, 66% for 1-year-olds, and 60% for 2-year-olds. Dean et al. (2014a) observed that scanning during the second or third visit succeeded more often (success rate 100%) than scanning during the first visit (success rate 90%).

We refer to FinnBrain infant MRI measurements in the current article. They were carried out in accordance with the Declaration of Helsinki, and the protocol was approved by the joint Ethics Committee of the University of Turku and the Hospital District of Southwest Finland.

In the FinnBrain Birth Cohort Study, 189 families participated in the scans when their infants were 2–5-weeks old (M = 26.04, SD = 7.6, range = 8–51 days corrected for gestation), and 180 were scanned. The data collection took place between 2012 and 2016. Most scans were conducted from the afternoon to early evening hours (16:30–20:00), but ca. 10 scans took place on Saturday afternoons. The imaging was performed at the Department of Radiology, Turku University Hospital, Finland. All scans were obtained using a Siemens Magnetom Verio 3T scanner (Siemens Medical Solutions, Erlangen, Germany).

At the start of the visit, the families were welcomed by a trained and experienced radiographer and the researchers at the scanning site. The infants were fed with breastmilk or formula and swaddled into a vacuum mattress. No sedatives were used. Deformable wax plugs and custom-sized earmuffs were used for hearing protection. Parents were provided with standard earmuffs, as they stayed in the scanning room during the whole scanning session. The personnel observed the scanning from the control room through a window with a microphone contact to a parent. A loudspeaker sent the sounds from the scanning room to the control room allowing the staff to hear if the infant woke up. The session was ended if the infant did not fall asleep before or did not fall back asleep during the scan.

The sequences comprised of an axial PD-T2-TSE (Dual-Echo Turbo Spin Echo), a sagittal 3D-T1 (T1-weighted MPRAGE) and three diffusion tensor imaging (DTI) sequences respectively. The acquisition times for sequences were 6 min 50 s (axial PD-T2-TSE), 4 min 3 s (3D-T1), 5 min 3 s (DTI 1), 5 min 33 s (DTI 2) and 5 min 42 s (DTI 3). Sequence parameters were optimized so that “whisper” gradient mode could be used in PD-T2 TSE and 3D-T1 sequences to reduce acoustic noise during the scan (Lehtola et al., 2019). Functional MRI sequences were added to the protocol starting in June 2015 and performed until the end of the study (N = 28). We acquired task fMRI measurements investigating touch responses and a resting state fMRI scan. 60 min was the maximum duration of the complete scanning protocol, and total duration of the visit was less than 2 h. The data are still being processed and analyzed, but the current success rates are 125/180 for structural T1 and T2 scans (69%) (Acosta et al., 2020a), 172/180 (95%) for at least 20 good quality diffusion weighted images out of the acquired 96 images (N = 157 for 30 directions, N = 142 for 40 directions and N = 121 for 60 directions) (Merisaari et al., 2019). Success rates for task fMRI are 10/13 (77%) in preliminary findings (Tuulari et al., 2019) and 21/28 (75%) for the resting state data (Rajasilta et al., 2020). Unfortunately, we suffered from some technical difficulties with the T2-weighted images that were identified only after data collection, which significantly impacted the available good data from 92% estimated at the scanner to 69% in the final data. Figure 2 shows representative examples of successful and unsuccessful neonate MRI scans (randomly selected from our data).

To share practical advice, we also report some procedures that we used to calm the infants if they woke up during the scan, and that can be used by gently reaching for the infant even within the scanner bore. These procedures are presented in Figure 3. During the scans, either the investigators or the parents frequently calmed the infants with these relatively simple measures that rely on infant reflexes and/or calming touch. Of important note, the homogeneity of the static magnetic field may be affected when an adult reaches into the scanner bore. Usually, we used the soothing technique for a brief period of time if the baby was sleeping restlessly at the start of the session or started to move during scanning. We let the ongoing sequence to continue during the soothing (as abrupt changes in the surrounding noises risk waking up the infant), but always acquired the sequence again, i.e., the sequences during which soothing was used were considered failed and were never used in analyses. We did not record the total number of infants who woke up during the scan, were successfully calmed and how long it took for babies to be calmed and able to put back into the MRI scanner. This would be valuable data to collect in future studies.

After conducting this review, we suggest that all studies using MRI in infant or child studies should report at least the following points about the procedure: (1) timing of visit, (2) preparations at home and MRI facilities, (3) subject’s state during scanning (asleep/awake), (4) motion prevention techniques, (5) sound attenuation techniques, (6) monitoring during scanning, (7) scan duration, (8) exact number of included and excluded scans, (9) reasons for lost data, and (10) total duration of visit.

This review has its limitations. First, although our literature search was wide, we might have missed articles that were not identified with our keywords (subject to MESH terms). Second, our strict selection criteria set limits especially for scanning children at the older ages. We included only studies, which scanned children at the age of 6 or younger but which also included also at least one subject under the age of 2. Because of this, the review does not give a real view of the methods and challenges in performing MRI with older children and this topic warrants thorough treatment of its own. Reviewed studies did not apply methods like a mock scanner or play therapy probably because the majority of the subjects were still too young to take advantage of them. All in all, our main interest was scanning during the first few years of life, thus the presented methods are best applied in this age group. Third, we did not exclude studies using the same or overlapping sample (the review was performed study-by-study), which can be seen in the frequency of used methods. Fourth, in this review we also highlighted the difficulties of the imaging procedure and listed the reasons for lost data, which critically viewed is reported poorly. Various reasons leading to data exclusion made it difficult to classify reasons for failure and to perform a reliable synthesis about success rates between imaging techniques, e.g., structural MRI vs. diffusion or functional MRI. In addition, some studies identified retrospectively only subjects with good quality data, so the detailed information of success rates was missing. Due to heterogeneity of the reporting, the success rates were not comparable between studies. We were open to conduct meta-analysis over the success rates, reported procedures, and timing of the studies but unfortunately this was not possible due to the high variability in reporting. Finally, we did not cover prospective and retrospective motion correction techniques or best practices of determining the degree of tolerable motion, which deserve a thorough treatment of their own.