A non-parametric approach to automatic change detection in MRI images of the brain

Abstract

We present a novel approach to change detection between two brain MRI scans (reference and target.) The proposed method uses a single modality to find subtle changes; and does not require prior knowledge (learning) of the type of changes to be sought. The method is based on the computation of a local kernel from the reference image, which measures the likeness of a pixel to its surroundings. This kernel is then used as a feature and compared against analogous features from the target image. This comparison is made using cosine similarity. The overall algorithm yields a scalar dissimilarity map (DM), indicating the local statistical likelihood of dissimilarity between the reference and target images. DM values exceeding a threshold then identify meaningful and relevant changes. The proposed method is robust to various challenging conditions including unequal signal strength.

Change detection of medical images using dictionary learning techniques and principal component analysis

“…The Seo and Milanfar method 4 obtained an average value of 0.879 for SI. As an overall measure of performance, they were able to achieve sensitivity ¼ 0.877, specificity ¼ 0.998, and similarity index ðSIÞ ¼ 0.879.…”

“…4 Rousseau et al 13 evaluated the performance of their methods on simulated lesion images and reported their SI (kapa index) value of around 0.75. Shen et al…”

“…However, the tissue classification task [1][2][3] still remains very difficult and the whole process is inherently time consuming. Seo and Milanfar 4 introduced a general nonparametric statistical framework and a cosine similarity measure to detect changes in two MR images from a single MRI modality. However, their work does not address the effect of misalignment.…”

Multiscale NeighborhoodWise Decision Fusion for Redundancy Detection in Image Pairs

“…For a recent survey, see [83]. Actually, change detection is of significant interest in an increasing number of applications, such as video-surveillance (e.g., in airports, museums, shops, etc), medical diagnosis [18,81,45,88,90], cell biology imaging [78,19] and remote sensing [23,54]. The challenge lies in distinguishing between meaningful changes related to unusual scene events and changes corresponding to camera motion, camera noise or atmospheric/lighting conditions.…”

“…7.29, our approach produces similar detection masks if the search windows is higher than 5 × 5 pixels. In future work, we plan to compare these results to those obtained by more dedicated methods [88,90] in medical imaging. 8.…”