The role of noise in denoising models for anomaly detection in medical images

Pathological brain lesions exhibit diverse appearance in brain images, in terms of intensity, texture, shape, size, and location. Comprehensive sets of data and annotations are difficult to acquire. Therefore, unsupervised anomaly detection approaches have been proposed using only normal data for training, with the aim of detecting outlier anomalous voxels at test time. Denoising methods, for instance classical denoising autoencoders (DAEs) and more recently emerging diffusion models, are a promising approach, however naive application of pixelwise noise leads to poor anomaly detection performance. We show that optimization of the spatial resolution and magnitude of the noise improves the performance of different model training regimes, with similar noise parameter adjustments giving good performance for both DAEs and diffusion models. Visual inspection of the reconstructions suggests that the training noise influences the trade-off between the extent of the detail that is reconstructed and the extent of erasure of anomalies, both of which contribute to better anomaly detection performance. We validate our findings on two real-world datasets (tumor detection in brain MRI and hemorrhage/ischemia/tumor detection in brain CT), showing good detection on diverse anomaly appearances. Overall, we find that a DAE trained with coarse noise is a fast and simple method that gives state-of-the-art accuracy. Diffusion models applied to anomaly detection are as yet in their infancy and provide a promising avenue for further research. Code for our DAE model and coarse noise is provided at: https://github.com/AntanasKascenas/DenoisingAE.

A controlled comparison of thickness, volume and surface areas from multiple cortical parcellation packages

BACKGROUND

Cortical parcellation is an essential neuroimaging tool for identifying and characterizing morphometric and connectivity brain changes occurring with age and disease. A variety of software packages have been developed for parcellating the brain's cortical surface into a variable number of regions but interpackage differences can undermine reproducibility. Using a ground truth dataset (Edinburgh_NIH10), we investigated such differences for grey matter thickness (GMth), grey matter volume (GMvol) and white matter surface area (WMsa) for the superior frontal gyrus (SFG), supramarginal gyrus (SMG), and cingulate gyrus (CG) from 4 parcellation protocols as implemented in the FreeSurfer, BrainSuite, and BrainGyrusMapping (BGM) software packages.

RESULTS

Corresponding gyral definitions and morphometry approaches were not identical across the packages. As expected, there were differences in the bordering landmarks of each gyrus as well as in the manner in which variability was addressed. Rostral and caudal SFG and SMG boundaries differed, and in the event of a double CG occurrence, its upper fold was not always addressed. This led to a knock-on effect that was visible at the neighbouring gyri (e.g., knock-on effect at the SFG following CG definition) as well as gyral morphometric measurements of the affected gyri. Statistical analysis showed that the most consistent approaches were FreeSurfer's Desikan-Killiany-Tourville (DKT) protocol for GMth and BrainGyrusMapping for GMvol. Package consistency varied for WMsa, depending on the region of interest.

CONCLUSIONS

Given the significance and implications that a parcellation protocol will have on the classification, and sometimes treatment, of subjects, it is essential to select the protocol which accurately represents their regions of interest and corresponding morphometrics, while embracing cortical variability.