Adversarial Multiscale Feature Learning Framework for Overlapping Chromosome Segmentation

Abstract

Chromosome karyotype analysis is of great clinical importance in the diagnosis and treatment of diseases. Since manual analysis is highly time and effort consuming, computer-assisted automatic chromosome karyotype analysis based on images is routinely used to improve the efficiency and accuracy of the analysis. However, the strip-shaped chromosomes easily overlap each other when imaged, significantly affecting the accuracy of the subsequent analysis and hindering the development of chromosome analysis instruments. In this paper, we present an adversarial, multiscale feature learning framework to improve the accuracy and adaptability of overlapping chromosome segmentation. We first adopt the nested U-shaped network with dense skip connections as the generator to explore the optimal representation of the chromosome images by exploiting multiscale features. Then we use the conditional generative adversarial network (cGAN) to generate images similar to the original ones; the training stability of the network is enhanced by applying the least-square GAN objective. Finally, we replace the common cross-entropy loss with the advanced Lovász-Softmax loss to improve the model’s optimization and accelerate the model’s convergence. Comparing with the established algorithms, the performance of our framework is proven superior by using public datasets in eight evaluation criteria, showing its great potential in overlapping chromosome segmentation.

Global Semantic-Sense Aggregation Network for Salient Object Detection in Remote Sensing Images

“…Clear-edge structures contribute to accurately capturing object contours, providing visual clues for the localization of salient objects. Meanwhile, rich semantic features provide valuable contextual information for salient objects, assisting the model in comprehensively understanding the connotations and background of salient targets [ 26 ]. In shallow neural networks, feature maps have higher spatial resolution and carry richer spatial positional information.…”

Cross-Attention-Guided Feature Alignment Network for Road Crack Detection

“…Recent theoretical progress reveals that deep learning can proficiently tackle intricate problems by autonomously acquiring features at various hierarchical levels [15]. Deep Convolutional Neural Networks (DCNNs) are known for their ability to extract rich hierarchical features [16][17][18][19][20][21], and their end-to-end trainable framework of deep learning has demonstrated remarkable advancements in pixel-level semantic segmentation tasks [22][23][24]. As a result, several crack detection methods leveraging deep learning, such as object detection [25,26] and image block segmentation [27][28][29], have emerged.…”

DFA-Net: Multi-Scale Dense Feature-Aware Network via Integrated Attention for Unmanned Aerial Vehicle Infrared and Visible Image Fusion

“…Infrared image fusion methods are currently divided into traditional fusion methods [2][3][4] and deep-learning-based methods [5][6][7]. The traditional image fusion method can be categorized into three categories: multi-scale transformation methods (MST) [8][9][10][11][12][13][14], sparse representation methods (SR) [15][16][17][18], and hybrid methods [19][20][21][22]. Among them, multi-scale transformation-based methods are more commonly used, and the algorithms can be further divided into three categories: pyramid-transform-based image decomposition [23], wavelettransform-based image decomposition, and multi-scale geometric decomposition [24].…”