Probability-based Mask R-CNN for pulmonary embolism detection

Long, Kun; Tang, Lei; Pu, Xiaorong; Ren, Yazhou; Zheng, Mingxiu; Gao, Li; Song, Chunjiang; Su, Hailin; Zhou, Min; Deng, Fengbin

doi:10.1016/j.neucom.2020.10.022

Cited by 35 publications

(25 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Study [ 27 ] produced satisfactory results with higher mAP with only two classes, tumor and non-tumor. Similarly, with other studies in [ 28 , 29 , 31 ] the instance segmentation approach can segment the object with best mAP performance. However, they only use two classes, healthy and non-healthy lesion, whereas the instance segmentation is prepared for multi-classes and multi object segmentation.…”

Section: Resultssupporting

confidence: 65%

“…However, it does not label all of the image pixels, as it segments only the RoIs. From previous studies [ 17 , 25 , 31 ], the segmentation rate was unsatisfactory, producing mAP values around 0.5. This happens as RGB images differ with a large pixel variation; thus, they cannot follow a distribution in [ 17 ].…”

Section: Resultsmentioning

confidence: 98%

“…A different study produced a mAP of over 0.75; the super pixels of each object image were similar between the ground truth and the prediction for the pixel-level instance performed the same process [ 19 , 26 , 27 , 28 , 29 , 31 ]. However, in [ 19 , 26 ], they added data augmentation and other preprocessing techniques; data augmentation arises from the data bias, as the augmented data distribution can be quite different from the original one.…”

Section: Resultsmentioning

confidence: 99%

See 2 more Smart Citations

Deep Learning-Based Computer-Aided Fetal Echocardiography: Application to Heart Standard View Segmentation for Congenital Heart Defects Detection

Nurmaini

Rachmatullah

Sapitri

et al. 2021

Sensors

View full text Add to dashboard Cite

Accurate segmentation of fetal heart in echocardiography images is essential for detecting the structural abnormalities such as congenital heart defects (CHDs). Due to the wide variations attributed to different factors, such as maternal obesity, abdominal scars, amniotic fluid volume, and great vessel connections, this process is still a challenging problem. CHDs detection with expertise in general are substandard; the accuracy of measurements remains highly dependent on humans’ training, skills, and experience. To make such a process automatic, this study proposes deep learning-based computer-aided fetal heart echocardiography examinations with an instance segmentation approach, which inherently segments the four standard heart views and detects the defect simultaneously. We conducted several experiments with 1149 fetal heart images for predicting 24 objects, including four shapes of fetal heart standard views, 17 objects of heart-chambers in each view, and three cases of congenital heart defect. The result showed that the proposed model performed satisfactory performance for standard views segmentation, with a 79.97% intersection over union and 89.70% Dice coefficient similarity. It also performed well in the CHDs detection, with mean average precision around 98.30% for intra-patient variation and 82.42% for inter-patient variation. We believe that automatic segmentation and detection techniques could make an important contribution toward improving congenital heart disease diagnosis rates.

show abstract

Section: Resultssupporting

confidence: 65%

Section: Resultsmentioning

confidence: 98%

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Deep Learning-Based Computer-Aided Fetal Echocardiography: Application to Heart Standard View Segmentation for Congenital Heart Defects Detection

Nurmaini

Rachmatullah

Sapitri

et al. 2021

Sensors

View full text Add to dashboard Cite

show abstract

“…Outras redes de segmentação podem também servir como alternativa, como a Mask R-CNN. No trabalho de [Long et al 2021] o autor utiliza uma variação da Mask R-CNN, chamada de P Mask R-CNN, queé otimizada para encontrar objetos pequenos.…”

Section: Conclusãounclassified

Detecção de Tromboembolia Pulmonar utilizando Redes Neurais Convolucionais e Extração de Características

Olescki¹,

Andrade²,

Escuissato³

et al. 2021

Anais Do XXI Simpósio Brasileiro De Computação Aplicada À Saúde (SBCAS 2021)

View full text Add to dashboard Cite

Embolia pulmonar é uma das principais causas de morte relacionadas a doenças cardiovasculares no mundo, uma vez que é feito o diagnostico é necessária uma resposta rápida pela equipe médica para salvar o paciente. A principal forma de diagnóstico é pelo exame de tomograﬁa computadorizada e, devido a grande quantidade de dados que o exame gera, algoritmos de deep learning têm mostrado bons resultados em encontrar embolia pulmonar de maneira autônoma. O objetivo deste trabalho é desenvolver uma rede de deep learning capaz de encontrar embolia pulmonar em exames de tomograﬁa computadorizada. Até então, utilizando uma rede inspirada na U-net, o método segmentou trombos atingindo um Dice Score de 0.81 e um IoU de 0.79.

show abstract

“…For example, existing deep learning-based LDCT image denoising approaches usually calculate the perceptual loss (which can be used to measure the difference of feature space between the denoised result and corresponding ground-truth) on the entire feature map [10,29,43,44,51]. However, the lesions detection task, for example, prefers focusing on the local region-of-interest [28], which causes the mismatch of the objective between these two tasks. In this case, the calculation of perceptual loss on local feature maps is a better choice for upstream denoising task.…”

Section: Introductionmentioning

confidence: 99%

Lesion-Inspired Denoising Network: Connecting Medical Image Denoising and Lesion Detection

Chen

Long

Ren

et al. 2021

Proceedings of the 29th ACM International Conference on Multimedia

Self Cite

View full text Add to dashboard Cite

Deep learning has achieved notable performance in the denoising task of low-quality medical images and the detection task of lesions, respectively. However, existing low-quality medical image denoising approaches are disconnected from the detection task of lesions. Intuitively, the quality of denoised images will influence the lesion detection accuracy that in turn can be used to affect the denoising performance. To this end, we propose a play-andplug medical image denoising framework, namely Lesion-Inspired Denoising Network (LIDnet), to collaboratively improve both denoising performance and detection accuracy of denoised medical images. Specifically, we propose to insert the feedback of downstream detection task into existing denoising framework by jointly learning a multi-loss objective. Instead of using perceptual loss calculated on the entire feature map, a novel region-of-interest (ROI) perceptual loss induced by the lesion detection task is proposed to further connect these two tasks. To achieve better optimization for overall framework, we propose a customized collaborative training strategy for LIDnet. On consideration of clinical usability and imaging characteristics, three low-dose CT images datasets are used to evaluate the effectiveness of the proposed LIDnet. Experiments show that, by equipping with LIDnet, both of the denoising and lesion detection performance of baseline methods can be significantly improved. CCS CONCEPTS• Computing methodologies → Machine learning approaches; Neural networks.

show abstract

Probability-based Mask R-CNN for pulmonary embolism detection

Cited by 35 publications

References 17 publications

Deep Learning-Based Computer-Aided Fetal Echocardiography: Application to Heart Standard View Segmentation for Congenital Heart Defects Detection

Deep Learning-Based Computer-Aided Fetal Echocardiography: Application to Heart Standard View Segmentation for Congenital Heart Defects Detection

Detecção de Tromboembolia Pulmonar utilizando Redes Neurais Convolucionais e Extração de Características

Lesion-Inspired Denoising Network: Connecting Medical Image Denoising and Lesion Detection

Contact Info

Product

Resources

About