In this issue, vol. 27, issue 4, April 2023, 11 papers are published related to the to the topic Imaging Informatics. Please click here to view them, with link in IEEE XPLORE.
MSHT: Multi-stage Hybrid Transformer for the ROSE Image Analysis of Pancreatic Cancer.
Zhang, Tianyi; Feng, Yunlu; Zhao, Yu; Fan, Guangda; Yang, Aiming; Lyu, Shangqing; Zhang, Peng; Song, Fan; Ma, Chenbin; Sun, Yangyang; Feng, Youdan; Zhang, Guanglei.
Pancreatic cancer is a highly virulent malignancy associated with a high mortality rate. The rapid on-site evaluation (ROSE) method can expedite the diagnostic process for pancreatic cancer by enabling the immediate analysis of fast-stained cytopathological images with on-site pathologists. However, the widespread implementation of ROSE diagnosis has been impeded by the shortage of pathologists. Deep learning offers the prospect of automating the classification of ROSE images for diagnostic purposes. Nonetheless, it poses a challenge to accurately model the complex local and global image features. The conventional convolutional neural network (CNN) structure is proficient in extracting spatial features, yet it tends to overlook global features when prominent local features are misleading with bias. Conversely, the Transformer structure excels at capturing global features and long-range relationships, but has limited capacity to incorporate local features. To address this challenge, we propose a multi-stage hybrid Transformer (MSHT) that leverages the strengths of both methods by using spatial attention to enhance global understanding. The MSHT features a CNN backbone that extracts multi-stage local features at different scales to serve as attention guidance, while the Transformer encodes these features during its sophisticated global modeling. The MSHT achieves high accuracy and promising interpretability in the classification of cancerous and normal pancreatic cells in ROSE samples. The superior results obtained by MSHT compared to state-of-the-art models make it a possible approach for general cytopathological image analysis. This highlights the potential for AI-based computer-aided diagnosis (CAD) systems to boost the diagnostic process and alleviate the demand for on-site pathologists. The application of such systems can benefit broader populations where on-site pathologists are scarce. The code and records of MSHT are available at https://github.com/sagizty/Multi-Stage-Hybrid-Transformer.
Toxicity Prediction in Pelvic Radiotherapy Using Multiple Instance Learning and Cascaded Attention Layers.
Elhaminia, Behnaz; Gilbert, Alexandra; Lilley, John; abdar, moloud; Frangi, Alejandro; Scarsbrook, Andrew; Appelt, Ane; Gooya, Ali.
Multi-task Learning for Pulmonary Arterial Hypertension Prognosis Prediction via Memory Drift and Prior Prompt Learning on 3D Chest CT.
Yang, Guanyu; He, Yuting; Lv, Yang; chen, yang; Coatrieux, Jean-Louis; Sun, Xiaoxuan; Wang, Qiang; Wei, Yongyue; Li, Shuo; Zhu, Yinsu.
Self-supervised Tumor Segmentation with Sim2Real Adaptation.
Zhang, Xiaoman; Xie, Weidi; Huang, Chaoqin; Zhang, Ya; Chen, Xin; Tian, Qi; Wang, Yanfeng
TP-Net: Two-Path Network for Retinal Vessel Segmentation.
Qu, Zhiwei; Zhuo, Li; Cao, Jie; Li, Xiaoguang; Yin, Hongxia; Wang, Zhenchang
Medical Image Classification Using Light-weight CNN with Spiking Cortical Model Based Attention Module.
Zhou, Quan; Huang, Zhiwen; Ding, Mingyue; Zhang, Xuming.
Unsupervised Visual Representation Learning Based on Segmentation of Geometric Pseudo-shapes for Transformer-based Medical Tasks.
Viriyasaranon, Thanaporn; Woo, Sang Myung; Choi, Jang-Hwan.
Annotation Cost Minimization for Ultrasound Image Segmentation using Cross-domain Transfer Learning.
Monkam, Patrice; Jin, Songbai; Lu, Wenkai.
PCCT: Progressive Class-Center Triplet Loss for Imbalanced Medical Image Classification.
Chen, Kanghao; Lei, Weixian; Hu, Ping; Zhao, Shen; Zheng, Wei-shi; Wang, Ruixuan.
Knowledge Distillation in Histology Landscape by Multi-Layer Features Supervision.
Javed, Sajid; Mahmood, Arif; Qaiser, Talha; Werghi, Naoufel.
Reconstruction of Quantitative Susceptibility Mapping from Total Field Maps with Local Field Maps Guided UU-Net.
Li, Zheng; Ying, Shihui; Wang, Jun; He, Hongjian; Shi, Jun.