How much motion is too much motion? Determining motion thresholds by sample size for reproducibility in developmental resting-state MRI

Abstract

Citation: Leonard J, Flournoy J, Lewis-de los Angeles CP, Whitaker K (2017) How much motion is too much motion? Determining motion thresholds by sample size for reproducibility in developmental resting-state MRI.Research Ideas and Outcomes 3: e12569. https://doi.org/10.3897/rio.3.e12569 AbstractA constant problem developmental neuroimagers face is in-scanner head motion. Children move more than adults and this has led to concerns that developmental changes in restingstate connectivity measures may be artefactual. Furthermore, children are challenging to recruit into studies and therefore researchers have tended to take a permissive stance when setting exclusion criteria on head motion. The literature is not clear regarding our central question: How much motion is too much? Here, we systematically examine the effects of multiple motion exclusion criteria at different sample sizes and age ranges in a large openly available developmental cohort (ABIDE; http://preprocessed-connectomesproject.org/abide). We checked 1) the reliability of resting-state functional magnetic resonance imaging (rs-fMRI) pairwise connectivity measures across the brain and 2) the accuracy with which we can separate participants with autism spectrum disorder from typically developing controls based on their rs-fMRI scans using machine learning. We find that reliability on average is primarily sensitive to the number of participants considered, but that increasingly permissive motion thresholds lower case-control prediction accuracy for all sample sizes. ‡ § | ¶,#

Improving practices and inferences in developmental cognitive neuroscience

“…For example, researchers might have used a 6 mm smoothing kernel, but could have chosen 4 mm or 8 mm. Or they may have chosen to exclude several participants based on a motion artifact threshold of 10 % of volumes, but it could have been specified at 15 % or 20 % instead (for such an exploration, see Leonard et al, 2017 ). Reasonable model specifications are defined as being: consistent with theory, statistically valid, and non-redundant ( Simonsohn et al, 2015 ).…”

Predictive water virology using regularized regression analyses for projecting virus inactivation efficiency in ozone disinfection

“…The numbers of LRV data point were 68 (norovirus), 65 (rotavirus), 182 (poliovirus), 111 (coxsackievirus), 29 (echovirus) and 29 (adenovirus), respectively ( Lim et al., 2010 ; Brie et al., 2018 ; Hirneisen et al., 2011 ; Tondera et al., 2015 ; Shin and Sobsey, 2003 ; Thurston-Enriquez et al., 2005 ; Harakeh and Bulter, 1984a ; Harakeh and Bulter, 1984b ; Meunier et al., 2006 ; Vaughn et al., 1987 ; Finch and Fairbairin, 1991 ; Farooq and Akhlaque, 1983 ; Katzenelson and Biedermann, 1976 ; Moore and Magolin, 1994 ; Roy et al., 1981 ; Roy et al., 1982 ; Engelbrecht et al., 1980 ; Katzenelson et al., 1979 ; Emerson et al., 1982 ; Wang et al., 2018 ; Wolf et al., 2018 ). The Scikit-learn library ( https://scikit-learn.org/stable/tutorial/machine_learning_map/index.html ) and previous studies suggest that more than 50 datasets should be analyzed ( Cui and Gong, 2018 ; Leonard et al., 2017 ), so we excluded echovirus and adenovirus data from our analyses.…”

Regularized regression analysis for the prediction of virus inactivation efficiency by chloramine disinfection

“…17,[31][32][33][34][35][36][37][38][39][40][41][42] The number of LRV data points was 120 (norovirus), 353 (adenovirus), 82 (poliovirus), 59 (coxsackievirus) and 52 (echovirus), respectively (Table 2), which correspond to the number of datasets recommended by scikit-learn and previous reports. 43,44 All the datasets of LRVs were calculated using the infectious titer. LRVs of coxsackievirus and echovirus strains were examined for two strains, which were expressed using a dummy variable (one: 1, another: 0).…”