/pdf/a-transfer-learning-approach-with-a-convolutional-neural-3gm3aycx.pdf

A Transfer Learning Approach with a Convolutional Neural Network for the Classification of Lung Carcinoma

Deep learning has been widely used in medical image segmentation and other aspects. However, the performance of existing medical image segmentation models has been limited by the challenge of obtaining sufficient high-quality labeled data due to the prohibitive data annotation cost. To alleviate this limitation, we propose a new text-augmented medical image segmentation model LViT (Language meets Vision Transformer). In our LViT model, medical text annotation is incorporated to compensate for the quality deficiency in image data. In addition, the text information can guide to generate pseudo labels of improved quality in the semi-supervised learning. We also propose an Exponential Pseudo label Iteration mechanism (EPI) to help the Pixel-Level Attention Module (PLAM) preserve local image features in semi-supervised LViT setting. In our model, LV (Language-Vision) loss is designed to supervise the training of unlabeled images using text information directly. For evaluation, we construct three multimodal medical segmentation datasets (image + text) containing X-rays and CT images. Experimental results show that our proposed LViT has superior segmentation performance in both fully-supervised and semi-supervised setting. The code and datasets are available at https://github.com/HUANGLIZI/LViT.

LViT: Language meets Vision Transformer in Medical Image Segmentation

Retina vessel segmentation is a vital step in diagnosing ophthalmologic diseases. Traditionally, ophthalmologists segment retina vessels by hand, which is time-consuming and error-prone. Thus, more and more researchers are committed to the research of automatic segmentation algorithms. With the development of convolution neural networks (CNNs), many tasks can be solved by CNNs. In this paper, we propose an encoding-decoding network with a pyramid self-attention module (PSAM) to segment retinal vessels. The network follows a U shape structure, and it comprises stacked feature selection blocks (FSB) and a PSAM. The proposed FSB consists of two convolution blocks with the same weight and a channel-wise attention block. At the head of the network, we apply a PSAM consisting of three parallel self-attention modules to capture long-range dependence of different scales. Due to the power of PSAM and FSB, the performance of the network improves. We have evaluated our model on two public datasets: DRIVE and CHASE_DB1. The results show the performance of our model is better than other methods. The F1, Accuracy, and area under curve (AUC) are 82.21%/80.57%, 95.65%/97.02%, and 98.16%/98.46% on DRIVE and CHASEDB1, respectively.

Encoding-decoding Network With Pyramid Self-attention Module For Retinal Vessel Segmentation

Aortic dissection is a life‐threatening vascular disease characterized by abrupt formation of a new flow channel (false lumen) within the aortic wall. Survivors of the acute phase remain at high risk for late complications, such as aneurysm formation, rupture, and death. Morphologic features of aortic dissection determine not only treatment strategies in the acute phase (surgical vs. endovascular vs. medical), but also modulate the hemodynamics in the false lumen, ultimately responsible for late complications. Accurate description of the true and false lumen, any communications across the dissection membrane separating the two lumina, and blood supply from each lumen to aortic branch vessels is critical for risk prediction. Patient‐specific surface representations are also a prerequisite for hemodynamic simulations, but currently require time‐consuming manual segmentation of CT data. We present an aortic dissection cross‐sectional model that captures the varying aortic anatomy, allowing for reliable measurements and creation of high‐quality surface representations. In contrast to the traditional spline‐based cross‐sectional model, we employ elliptic Fourier descriptors, which allows users to control the accuracy of the cross‐sectional contour of a flow channel. We demonstrate (i) how our approach can solve the requirements for generating surface and wall representations of the flow channels, (ii) how any number of communications between flow channels can be specified in a consistent manner, and (iii) how well branches connected to the respective flow channels are handled. Finally, we discuss how our approach is a step forward to an automated generation of surface models for aortic dissections from raw 3D imaging segmentation masks.

Implicit Modeling of Patient-Specific Aortic Dissections with Elliptic Fourier Descriptors

Video surveillance using deep transfer learning and deep domain adaptation: Towards better generalization

Transfer learning for medical images analyses: A survey

Fast high-precision patient-specific vascular tissue and geometric structure reconstruction is an essential task for vascular tissue engineering and computer-aided minimally invasive vascular disease diagnosis and surgery. In this paper, we present an effective vascular geometry reconstruction technique by representing a highly complicated geometric structure of a vascular system as an implicit function. By implicit geometric modelling, we are able to reduce the complexity and level of difficulty of this geometric reconstruction task and turn it into a parallel process of reconstructing a set of simple short tubular-like vascular sections, thanks to the easy-blending nature of implicit geometries on combining implicitly modelled geometric forms. The basic idea behind our technique is to consider this extremely difficult task as a process of team exploration of an unknown environment like a cave. Based on this idea, we developed a parallel vascular modelling technique, called Skeleton Marching, for fast vascular geometric reconstruction. With the proposed technique, we first extract the vascular skeleton system from a given volumetric medical image. A set of sub-regions of a volumetric image containing a vascular segment is then identified by marching along the extracted skeleton tree. A localised segmentation method is then applied to each of these sub-image blocks to extract a point cloud from the surface of the short simple blood vessel segment contained in the image block. These small point clouds are then fitted with a set of implicit surfaces in a parallel manner. A high-precision geometric vascular tree is then reconstructed by blending together these simple tubular-shaped implicit surfaces using the shape-preserving blending operations. Experimental results show the time required for reconstructing a vascular system can be greatly reduced by the proposed parallel technique.

/pdf/skeleton-marching-based-parallel-vascular-geometry-54tcsfffum.pdf

Qing Qi Hong

Papers

Transfer learning for medical images analyses: A survey

Skeleton Marching-based Parallel Vascular Geometry Reconstruction Using Implicit Functions