-
DoubleU-NetPlus: A Novel Attention and Context Guided Dual U-Net with Multi-Scale Residual Feature Fusion Network for Semantic Segmentation of Medical Images
Authors:
Md. Rayhan Ahmed,
Adnan Ferdous Ashrafi,
Raihan Uddin Ahmed,
Swakkhar Shatabda,
A. K. M. Muzahidul Islam,
Salekul Islam
Abstract:
Accurate segmentation of the region of interest in medical images can provide an essential pathway for devising effective treatment plans for life-threatening diseases. It is still challenging for U-Net, and its state-of-the-art variants, such as CE-Net and DoubleU-Net, to effectively model the higher-level output feature maps of the convolutional units of the network mostly due to the presence of…
▽ More
Accurate segmentation of the region of interest in medical images can provide an essential pathway for devising effective treatment plans for life-threatening diseases. It is still challenging for U-Net, and its state-of-the-art variants, such as CE-Net and DoubleU-Net, to effectively model the higher-level output feature maps of the convolutional units of the network mostly due to the presence of various scales of the region of interest, intricacy of context environments, ambiguous boundaries, and multiformity of textures in medical images. In this paper, we exploit multi-contextual features and several attention strategies to increase networks' ability to model discriminative feature representation for more accurate medical image segmentation, and we present a novel dual U-Net-based architecture named DoubleU-NetPlus. The DoubleU-NetPlus incorporates several architectural modifications. In particular, we integrate EfficientNetB7 as the feature encoder module, a newly designed multi-kernel residual convolution module, and an adaptive feature re-calibrating attention-based atrous spatial pyramid pooling module to progressively and precisely accumulate discriminative multi-scale high-level contextual feature maps and emphasize the salient regions. In addition, we introduce a novel triple attention gate module and a hybrid triple attention module to encourage selective modeling of relevant medical image features. Moreover, to mitigate the gradient vanishing issue and incorporate high-resolution features with deeper spatial details, the standard convolution operation is replaced with the attention-guided residual convolution operations, ...
△ Less
Submitted 25 November, 2022;
originally announced November 2022.
-
MCFFA-Net: Multi-Contextual Feature Fusion and Attention Guided Network for Apple Foliar Disease Classification
Authors:
Md. Rayhan Ahmed,
Adnan Ferdous Ashrafi,
Raihan Uddin Ahmed,
Tanveer Ahmed
Abstract:
Numerous diseases cause severe economic loss in the apple production-based industry. Early disease identification in apple leaves can help to stop the spread of infections and provide better productivity. Therefore, it is crucial to study the identification and classification of different apple foliar diseases. Various traditional machine learning and deep learning methods have addressed and inves…
▽ More
Numerous diseases cause severe economic loss in the apple production-based industry. Early disease identification in apple leaves can help to stop the spread of infections and provide better productivity. Therefore, it is crucial to study the identification and classification of different apple foliar diseases. Various traditional machine learning and deep learning methods have addressed and investigated this issue. However, it is still challenging to classify these diseases because of their complex background, variation in the diseased spot in the images, and the presence of several symptoms of multiple diseases on the same leaf. This paper proposes a novel transfer learning-based stacked ensemble architecture named MCFFA-Net, which is composed of three pre-trained architectures named MobileNetV2, DenseNet201, and InceptionResNetV2 as backbone networks. We also propose a novel multi-scale dilated residual convolution module to capture multi-scale contextual information with several dilated receptive fields from the extracted features. Channel-based attention mechanism is provided through squeeze and excitation networks to make the MCFFA-Net focused on the relevant information in the multi-receptive fields. The proposed MCFFA-Net achieves a classification accuracy of 90.86%.
△ Less
Submitted 25 November, 2022;
originally announced November 2022.
-
Prediction of a Gene Regulatory Network from Gene Expression Profiles With Linear Regression and Pearson Correlation Coefficient
Authors:
Md Mehedi Hassan Onik,
Shakhawat Ahmmed Nobin,
Adnan Ferdous Ashrafi,
Tareque Mohmud Chowdhury
Abstract:
Reconstruction of gene regulatory networks is the process of identifying gene dependency from gene expression profile through some computation techniques. In our human body, though all cells pose similar genetic material but the activation state may vary. This variation in the activation of genes helps researchers to understand more about the function of the cells. Researchers get insight about di…
▽ More
Reconstruction of gene regulatory networks is the process of identifying gene dependency from gene expression profile through some computation techniques. In our human body, though all cells pose similar genetic material but the activation state may vary. This variation in the activation of genes helps researchers to understand more about the function of the cells. Researchers get insight about diseases like mental illness, infectious disease, cancer disease and heart disease from microarray technology, etc. In this study, a cancer-specific gene regulatory network has been constructed using a simple and novel machine learning approach. In First Step, linear regression algorithm provided us the significant genes those expressed themselves differently. Next, regulatory relationships between the identified genes has been computed using Pearson correlation coefficient. Finally, the obtained results have been validated with the available databases and literatures. We can identify the hub genes and can be targeted for the cancer diagnosis.
△ Less
Submitted 1 May, 2018;
originally announced May 2018.