Texture analysis for muscular dystrophy classification in MRI with improved class activation mapping

Abstract

The muscular dystrophies are made up of a diverse group of rare genetic diseases characterized by progressive loss of muscle strength and muscle damage. Since there is no cure for muscular dystrophy and clinical outcome measures are limited, it is critical to assess the progression of MD objectively. Imaging muscle replacement by fibrofatty tissue has been shown to be a robust biomarker to monitor disease progression in DMD. In magnetic resonance imaging (MRI) data, specific texture patterns are found to correlate to certain MD subtypes and thus present a potential way for automatic assessment. In this paper, we first apply state-of-the-art convolutional neural networks (CNNs) to perform accurate MD image classification and then propose an effective visualization method to highlight the important image textures. With a dystrophic MRI dataset, we found that the best CNN model delivers an 91.7% classification accuracy, which significantly outperforms non-deep learning methods, e.g., >40% improvement has been found over the traditional mean fat fraction (MFF) criterion for DMD and CMD classification. After investigating every single neuron at the top layer of CNN model, we found the superior classification ability of CNN can be explained by its 91 and 118 neurons were performing better than the MFF criterion under the measurements of Euclidean and Chi-square distance, respectively. In order to further interpret CNNs predictions, we tested an improved class activation mapping (ICAM) method to visualize the important regions in the MRI images. With this ICAM, CNNs are able to locate the most discriminative texture patterns of DMD in soleus, lateral gastrocnemius, and medial gastrocnemius; for CMD, the critical texture patterns are highlighted in soleus, tibialis posterior, and peroneus.

Advancements in magnetic resonance imaging‐based biomarkers for muscular dystrophy

“…The use of radiomics offers a novel way of overcoming the inherent subjectivity of traditional scoring systems by employing fully automated and systematic methods of quantitatively analyzing imaging data across multiple sequences. These methods may also improve tissue characterization by detecting muscle features that cannot be perceived by visual inspection …”

“…These methods may also improve tissue characterization by detecting muscle features that cannot be perceived by visual inspection. 119,120 Another potential challenge in developing quantitative analysis techniques is the difficulty of establishing reproducibility. Manual muscle segmentation may be subject to variability in how muscles are selected and defined.…”

Multiexponential Analysis of the Water T2‐Relaxation in the Skeletal Muscle Provides Distinct Markers of Disease Activity Between Inflammatory and Dystrophic Myopathies

“…31,32 Moreover, MRI offers the possibility for texture analysis, which has already shown promising results regarding the classification of muscular dystrophy subtypes. 33 However, to our knowledge, most available water T 2 mapping methods assume a monoexponential model for the water T 2 -relaxation in skeletal muscle. Furthermore, in most quantitative MRI studies to date, the exploitation of water T 2 maps remain restricted to the analysis by regions-of-interest, where each muscle or muscle group is characterized by an average water T 2 value.…”

Super resolution microscopy and deep learning identify Zika virus reorganization of the endoplasmic reticulum

“…This demonstrates that the neural network is detecting the ZIKV-induced reorganization of the CER that leads to the formation of the tubular matrix associated with ZIKV replication. Accuracy obtained for VGG16 classification of ZIKV-infected cells from 3D STED image stacks is comparable to prior classification using VGG16 37,38,[50][51][52] . CNN deep learning analysis was better able to distinguish ZIKVinfected from mock-infected cells based on ERmoxGFP compared to Sec61β-GFP labeling.…”