Biology subjects

Airola, A.

Publications and source records attributed to Airola, A..

3 recordsLinked to original sources

bioRxiv2025

skiftiTools: An R package for reading, writing, analysing, and visualising, tract-based spatial statistics (TBSS) derived diffusion MR images

skiftiTools processes three- and four-dimensional neuroimaging data, facilitating advanced statistical modelling with voxelwise data in any software of choice. Tract-Based Spatial Statistics (TBSS) is a conventionally used tool to make statistical calculations in voxel space for brain imaging data. While pre-existing software packages provide support for general linear model based statistics, there is a clear need for more sophisticated modeling. skiftiTools writes subject-per-volume NIfTI files as tab-separated value ASCII files, which are easily readable by most commonly used statistical tools such as R language (RStudio), SPSS, SAS, and GraphPad Prism. This facilitates a wide range of voxel-level statistical analyses from TBSS data, including estimation of standardised effect sizes, clustering, dimensionality reduction, non-linear and machine learning predictive modelling, which we showcase in this article using FinnBrain and developing Human Connectome Project diffusion MRI data. After statistical processing, the resulting ASCII data can then be read again for visualization. The package supports NIfTI image format, tab-separated ASCII format, and its own stand-alone format for efficient disk usage. It is open source (https://github.com/haanme/skiftiTools), built on R-language and has easy installation from Rs CRAN package repository. In addition, we provide basic functions available in Docker containers for further platform independence. HighlightsO_LIThe skiftiTools R package is an open-source, user-friendly interface for analysing voxelwise diffusion tensor imaging (DTI) data following tract-based spatial statistics (TBSS) processing C_LIO_LIIt supports reading, writing, visualization, mathematical operations, and data manipulation and thus allows comprehensive conventional and advanced statistics, including machine learning C_LIO_LIskiftiTools bridges a critical gap between statistical tools in R and voxelwise neuroimaging data - including comparable means to perform multiple comparison corrections and much needed possibility to use non-linear statistics C_LI

neuroscience↗

bioRxiv2020

Resting state EEG as a biomarker of Parkinson's disease: Influence of measurement conditions

Resting state electroencephalographic (EEG) recording could provide cost-effective means to aid in the detection of neurological disorders such as Parkinsons disease (PD). We examined how many electrodes are needed for classification of PD based on EEG, which electrode locations provide most value for classification, and whether data recorded eyes open or closed yield comparable results. We used a nested cross-validated classifier which included a budget-based search algorithm for selecting the optimal electrodes for classification. By iterating over variable budgets, we show that with eyes open recording, only 10 electrodes, localized over motor and occipital areas enable relatively accurate classification (AUC = .82) between PD patients (N=20) and age-matched healthy control participants (N=20). Classification accuracy only slightly increased when all 64 electrodes were included (AUC = .85). With the data recorded eyes closed, classification was not statistically significantly above chance even with full set of 64 electrodes (AUC = .55). These results show that classification based on small number of EEG electrodes is a promising tool for classifying PD, but measurement conditions and electrode locations can have a significant effect on classifier performance.

neuroscience↗

bioRxiv2018

Algebraic Shortcuts for Leave-One-Out Cross-Validation in Supervised Network Inference

MotivationSupervised machine learning techniques have traditionally been very successful at reconstructing biological networks, such as protein-ligand interaction, protein-protein interaction and gene regulatory networks. Recently, much emphasis has been placed on the correct evaluation of such supervised models. It is vital to distinguish between using the model to either predict new interactions in a given network or to predict interactions for a new vertex not present in the original network. Specific cross-validation schemes need to be used to assess the performance in such different prediction settings.\n\nResultsWe present a series of leave-one-out cross-validation shortcuts to rapidly estimate the performance of state-of-the-art kernel-based network inference techniques.\n\nAvailabilityThe machine learning techniques with the algebraic shortcuts are implemented in the RLScore software package.

systems biology↗