Shape-Aware Organ Segmentation by Predicting Signed Distance Maps DOI Open Access
Yuan Xue, Hui Tang, Zhi Qiao

et al.

Proceedings of the AAAI Conference on Artificial Intelligence, Journal Year: 2020, Volume and Issue: 34(07), P. 12565 - 12572

Published: April 3, 2020

In this work, we propose to resolve the issue existing in current deep learning based organ segmentation systems that they often produce results do not capture overall shape of target and lack smoothness. Since there is a rigorous mapping between Signed Distance Map (SDM) calculated from object boundary contours binary map, exploit feasibility SDM directly medical scans. By converting task into predicting an SDM, show our proposed method retains superior performance has better smoothness continuity shape. To leverage complementary information traditional training, introduce approximated Heaviside function train model by SDMs maps simultaneously. We validate models conducting extensive experiments on hippocampus dataset public MICCAI 2015 Head Neck Auto Segmentation Challenge with multiple organs. While carefully designed backbone 3D network improves Dice coefficient more than 5% compared state-of-the-arts, produces smoother smaller Hausdorff distance average surface distance, thus proving effectiveness method.

Language: Английский

A Deep Learning System to Screen Novel Coronavirus Disease 2019 Pneumonia DOI Creative Commons
Xiaowei Xu,

Xiangao Jiang,

Chunlian Ma

et al.

Engineering, Journal Year: 2020, Volume and Issue: 6(10), P. 1122 - 1129

Published: June 27, 2020

The real-time reverse transcription-polymerase chain reaction (RT-PCR) detection of viral RNA from sputum or nasopharyngeal swab had a relatively low positive rate in the early stage coronavirus disease 2019 (COVID-19). Meanwhile, manifestations COVID-19 as seen through computed tomography (CT) imaging show individual characteristics that differ those other types pneumonia such influenza-A (IAVP). This study aimed to establish an screening model distinguish IAVP and healthy cases pulmonary CT images using deep learning techniques. A total 618 samples were collected: 219 110 patients with (mean age 50 years; 63 (57.3%) male patients); 224 61 156 (69.6%) 175 39 97 (55.4%) patients). All contributed three COVID-19-designated hospitals Zhejiang Province, China. First, candidate infection regions segmented out image set 3D model. These separated then categorized into COVID-19, IAVP, irrelevant (ITI) groups, together corresponding confidence scores, location-attention classification Finally, type overall score for each case calculated Noisy-OR Bayesian function. experimental result benchmark dataset showed accuracy was 86.7% terms all taken together. models established this effective demonstrated be promising supplementary diagnostic method frontline clinical doctors.

Language: Английский

Citations

726
Artificial Intelligence and COVID-19: Deep Learning Approaches for Diagnosis and Treatment DOI Creative Commons
Mohammad Jamshidi, Ali Lalbakhsh, Jakub Talla

et al.

IEEE Access, Journal Year: 2020, Volume and Issue: 8, P. 109581 - 109595

Published: Jan. 1, 2020

COVID-19 outbreak has put the whole world in an unprecedented difficult situation bringing life around to a frightening halt and claiming thousands of lives. Due COVID-19's spread 212 countries territories increasing numbers infected cases death tolls mounting 5,212,172 334,915 (as May 22 2020), it remains real threat public health system. This paper renders response combat virus through Artificial Intelligence (AI). Some Deep Learning (DL) methods have been illustrated reach this goal, including Generative Adversarial Networks (GANs), Extreme Machine (ELM), Long/Short Term Memory (LSTM). It delineates integrated bioinformatics approach which different aspects information from continuum structured unstructured data sources are together form user-friendly platforms for physicians researchers. The main advantage these AI-based is accelerate process diagnosis treatment disease. most recent related publications medical reports were investigated with purpose choosing inputs targets network that could facilitate reaching reliable Neural Network-based tool challenges associated COVID-19. Furthermore, there some specific each platform, various forms data, such as clinical imaging can improve performance introduced approaches toward best responses practical applications.

Language: Английский

Citations

514
CPFNet: Context Pyramid Fusion Network for Medical Image Segmentation DOI
Shuanglang Feng, Heming Zhao, Fei Shi

et al.

IEEE Transactions on Medical Imaging, Journal Year: 2020, Volume and Issue: 39(10), P. 3008 - 3018

Published: March 27, 2020

Accurate and automatic segmentation of medical images is a crucial step for clinical diagnosis analysis. The convolutional neural network (CNN) approaches based on the U-shape structure have achieved remarkable performances in many different image tasks. However, context information extraction capability single stage insufficient this structure, due to problems such as imbalanced class blurred boundary. In paper, we propose novel Context Pyramid Fusion Network (named CPFNet) by combining two pyramidal modules fuse global/multi-scale information. Based first design multiple global pyramid guidance (GPG) between encoder decoder, aiming at providing levels decoder reconstructing skip-connection. We further scale-aware fusion (SAPF) module dynamically multi-scale high-level features. These can exploit rich progressively. Experimental results show that our proposed method very competitive with other state-of-the-art methods four challenging tasks, including skin lesion segmentation, retinal linear multi-class thoracic organs risk edema lesions.

Language: Английский

Citations

465
Loss odyssey in medical image segmentation DOI
Jun Ma, Jianan Chen, Matthew Ng

et al.

Medical Image Analysis, Journal Year: 2021, Volume and Issue: 71, P. 102035 - 102035

Published: March 20, 2021

Language: Английский

Citations

418
Unified Focal loss: Generalising Dice and cross entropy-based losses to handle class imbalanced medical image segmentation DOI Creative Commons
Michael Yeung, Evis Sala, Carola‐Bibiane Schönlieb

et al.

Computerized Medical Imaging and Graphics, Journal Year: 2021, Volume and Issue: 95, P. 102026 - 102026

Published: Dec. 13, 2021

Automatic segmentation methods are an important advancement in medical image analysis. Machine learning techniques, and deep neural networks particular, the state-of-the-art for most tasks. Issues with class imbalance pose a significant challenge datasets, lesions often occupying considerably smaller volume relative to background. Loss functions used training of algorithms differ their robustness imbalance, direct consequences model convergence. The commonly loss based on either cross entropy loss, Dice or combination two. We propose Unified Focal new hierarchical framework that generalises entropy-based losses handling imbalance. evaluate our proposed function five publicly available, imbalanced imaging datasets: CVC-ClinicDB, Digital Retinal Images Vessel Extraction (DRIVE), Breast Ultrasound 2017 (BUS2017), Brain Tumour Segmentation 2020 (BraTS20) Kidney 2019 (KiTS19). compare performance against six functions, across 2D binary, 3D binary multiclass tasks, demonstrating is robust consistently outperforms other functions. Source code available at: https://github.com/mlyg/unified-focal-loss.

Language: Английский

Citations

377
Boundary loss for highly unbalanced segmentation DOI
Hoel Kervadec,

Jihene Bouchtiba,

Christian Desrosiers

et al.

Medical Image Analysis, Journal Year: 2020, Volume and Issue: 67, P. 101851 - 101851

Published: Oct. 6, 2020

Language: Английский

Citations

279
Artificial intelligence and machine learning for medical imaging: A technology review DOI Open Access
Ana María Barragán Montero, Umair Javaid, Gilmer Valdés

et al.

Physica Medica, Journal Year: 2021, Volume and Issue: 83, P. 242 - 256

Published: March 1, 2021

Language: Английский

Citations

270
Adaptive Radiotherapy for Anatomical Changes DOI
Jan‐Jakob Sonke, M.C. Aznar, C. Rasch

et al.

Seminars in Radiation Oncology, Journal Year: 2019, Volume and Issue: 29(3), P. 245 - 257

Published: April 23, 2019

Language: Английский

Citations

234
A survey on U-shaped networks in medical image segmentations DOI
Liangliang Liu, Jianhong Cheng, Quan Quan

et al.

Neurocomputing, Journal Year: 2020, Volume and Issue: 409, P. 244 - 258

Published: June 1, 2020

Language: Английский

Citations

227
Deep learning to achieve clinically applicable segmentation of head and neck anatomy for radiotherapy DOI Creative Commons
Stanislav Nikolov, Sam Blackwell, Alexei Zverovitch

et al.

arXiv (Cornell University), Journal Year: 2018, Volume and Issue: unknown

Published: Jan. 1, 2018

Over half a million individuals are diagnosed with head and neck cancer each year worldwide. Radiotherapy is an important curative treatment for this disease, but it requires manual time consuming delineation of radio-sensitive organs at risk (OARs). This planning process can delay treatment, while also introducing inter-operator variability resulting downstream radiation dose differences. While auto-segmentation algorithms offer potentially time-saving solution, the challenges in defining, quantifying achieving expert performance remain. Adopting deep learning approach, we demonstrate 3D U-Net architecture that achieves expert-level delineating 21 distinct OARs commonly segmented clinical practice. The model was trained on dataset 663 deidentified computed tomography (CT) scans acquired routine practice both segmentations taken from created by experienced radiographers as part research, all accordance consensus OAR definitions. We model's applicability assessing its test set CT practice, two independent experts. introduce surface Dice similarity coefficient (surface DSC), new metric comparison organ delineation, to quantify deviation between contours rather than volumes, better reflecting task correcting errors automated segmentations. generalisability then demonstrated open source datasets, different centres countries training. With appropriate validation studies regulatory approvals, system could improve efficiency, consistency, safety radiotherapy pathways.

Language: Английский

Citations

213