Harnessing spatial MRI normalization: patch individual filter layers for CNNs
This work addresses a domain-specific problem for neuroimaging researchers by improving CNN performance on MRI data, though it is incremental as it builds on existing preprocessing methods.
The paper tackles the problem of applying CNNs to spatially normalized MRI data by proposing a new patch individual filter (PIF) layer that trains filters locally, assuming abstract features are region-specific. It shows that CNNs with PIF layers outperform standard CNNs in tasks like sex classification and disease detection, particularly in low sample size settings.
Neuroimaging studies based on magnetic resonance imaging (MRI) typically employ rigorous forms of preprocessing. Images are spatially normalized to a standard template using linear and non-linear transformations. Thus, one can assume that a patch at location (x, y, height, width) contains the same brain region across the entire data set. Most analyses applied on brain MRI using convolutional neural networks (CNNs) ignore this distinction from natural images. Here, we suggest a new layer type called patch individual filter (PIF) layer, which trains higher-level filters locally as we assume that more abstract features are locally specific after spatial normalization. We evaluate PIF layers on three different tasks, namely sex classification as well as either Alzheimer's disease (AD) or multiple sclerosis (MS) detection. We demonstrate that CNNs using PIF layers outperform their counterparts in several, especially low sample size settings.