A novel dual-network architecture for mixed-supervised medical image segmentation

Wang, Duo; Li, Ming; Ben‐Shlomo, Nir; Corrales, C. Eduardo; Zhang, Tao; Jayender, Jagadeesan

doi:10.1016/j.compmedimag.2020.101841

Cited by 8 publications

(2 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…By integrating the pyramid and grab cut mechanism, a collaborative and concerted segmentation architecture was developed for lung nodule, 45 which utilized the Gibbs energy function to draw out the nodule in an absolute proportion. The MSDN 46 is a U‐Net network comprising of unchanging stages in the extraction and contraction path with just squeeze and excitation blocks added after every stage. It demonstrated a constraint wherein segmentation on images with more than tandem annotations could not be performed.…”

Section: Literature Reviewmentioning

confidence: 99%

SAtUNet: Series atrous convolution enhanced U‐Net for lung nodule segmentation

Selvadass,

Bruntha,

Sagayam

et al. 2023

Int J Imaging Syst Tech

View full text Add to dashboard Cite

Precise and unambiguous segmentation of pulmonary nodules from the CT images is imperative for a CAD framework implementation delineated for the prognosis of lung cancer. Lung nodule segmentation is an appealing research discipline for accurate dismemberment of lung cancer but the irregularity in shades, contours, and compositions, and the affinity between the tumors and the neighboring regions makes it an arduous task. This paper proffers a series atrous convolution enhanced U‐Net which uses a series of concatenated dilated convolution blocks after every stage in the encoder and decoder path. Our approach helps in obtaining the quintessential components from the feature maps, in addition to the absolute convergence of the model. It is largely assessed on the publicly accessible LIDC‐IDRI dataset. The average Dice Similarity Coefficient (DSC) obtained is 81.10% with an Intersection over Union (IoU/ Jaccard Index) of 72.24%. Exploratory outcomes prove that our architecture achieves ameliorate performance.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

SAtUNet: Series atrous convolution enhanced U‐Net for lung nodule segmentation

Selvadass,

Bruntha,

Sagayam

et al. 2023

Int J Imaging Syst Tech

View full text Add to dashboard Cite

show abstract

“…Due to the pooling layers and up-sampling process, spatial information may be lost during the prediction, so there are some inaccuracies in the predicted segmentation map, especially in sharp regions such as boundaries. Therefore, a hopping architecture in FCN has been proposed to solve this problem [7]. Following this idea, Ronneberger et al (2015) proposed a standard and popular medical image segmentation architecture called U-Net, which includes a symmetric encoder and decoder.…”

Section: Introductionmentioning

confidence: 99%

Upper and lower extremity bone segmentation with Mask R-CNN

Aydın,

Özcan

2024

Bitlis Eren Üniversitesi Fen Bilimleri Dergisi

View full text Add to dashboard Cite

Most medical image processing studies use medical images to detect and measure the structure of organs and bones. The segmentation of image data is of great importance for the determination of the area to be studied and for the reduction of the size of the data to be studied. Working with image data creates an exponentially increasing workload depending on the size and number of images and requires high computing power using machine learning methods. Our study aims to achieve high success in bone segmentation, the first step in medical object detection studies. In many situations and cases, such as fractures and age estimation, the humerus and radius of the upper extremity and the femur and tibia of the lower extremity of the human skeleton provide data. In our bone segmentation study on X-RAY images, 160 images from one hundred patients were collected using data compiled from accessible databases. A segmentation result with an average accuracy of 0.981 was obtained using the Mask R-CNN method with the resnet50 architecture.

show abstract