Conditional Training with Bounding Map for Universal Lesion Detection

Li, Han; Chen, Long; Han, Hu; Cheng, Ying; Zhou, S. Kevin

doi:10.1007/978-3-030-87240-3_14

Cited by 13 publications

(11 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Five state-of-the-art ULD approaches [6,7,9,16,17] and one natural image [39] detection method are compared to evaluate SATr's effectiveness.…”

Section: Lesion Detection Performancementioning

confidence: 99%

“…Universal Lesion Detection (ULD) in computed tomography (CT) [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18], aiming to localize different types of lesions instead of identifying lesion types [19][20][21][22][23][24][25][26][27][28], plays an essential role in computer-aided diagnosis (CAD) [29,30]. ULD is a challenging task because different lesions have diverse shapes and sizes, easily leading to false positive and false negative detections.…”

Section: Introductionmentioning

confidence: 99%

“…Mainly inspired by the clinical fact that radiologists need several adjacent slices for locating and diagnosing lesions on one CT slice, most existing ULD methods take several adjacent 2D CT slices as the inputs to a 2D network architecture [3, 4, 6-10, 12, 15-18] or directly adopt 3D network designs [10] that take 3D volume as input to extract more 3D context information. While both 2D and 3D methods have yielded great A3D [16] A3D+SATr Input CT slices cBM [17] cBM+ SATr ULD performances, the multi-slice-input based 2D detection methods are much more popular than pure 3D fashion because 2D networks benefit from robust 2D models pretrained from large-scale data whereas publicly available 3D medical datasets are not large enough for robust 3D pretraining. While achieving success in ULD, the multi-slice-input based 2D approaches have inherent limitations: (i) Weak global context modeling within each slice.…”

Section: Introductionmentioning

confidence: 99%

“…Such convolution-based fusion methods are good at dealing with local features, but they unfortunately deteriorate in handling global features among different slices. To tackle this, some ULD approaches [9,16,17] propose to reshuffle feature channels among different slices which relieves this issue to some degree. But their ability in capturing rich global representations are still weak due to the use of pure convolutional operations.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

SATr: Slice Attention with Transformer for Universal Lesion Detection

Li¹,

Chen²,

Huang³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Universal Lesion Detection (ULD) in computed tomography plays an essential role in computer-aided diagnosis. Promising ULD results have been reported by multi-slice-input detection approaches which model 3D context from multiple adjacent CT slices, but such methods still experience difficulty in obtaining a global representation among different slices and within each individual slice since they only use convolutionbased fusion operations. In this paper, we propose a novel Slice Attention Transformer (SATr) block which can be easily plugged into convolutionbased ULD backbones to form hybrid network structures. Such newly formed hybrid backbones can better model long-distance feature dependency via the cascaded self-attention modules in the Transformer block while still holding a strong power of modeling local features with the convolutional operations in the original backbone. Experiments with five state-of-the-art methods show that the proposed SATr block can provide an almost free boost to lesion detection accuracy without extra hyperparameters or special network designs.

show abstract

“…Five state-of-the-art ULD approaches [6,7,9,16,17] and one natural image [39] detection method are compared to evaluate SATr's effectiveness.…”

Section: Lesion Detection Performancementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

SATr: Slice Attention with Transformer for Universal Lesion Detection

Li¹,

Chen²,

Huang³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…To alleviate this problem, many researchers [3,17,19,34] utilize labeled data together with unlabeled data in…”

Section: Introductionmentioning

confidence: 99%

Which images to label for few-shot medical landmark detection?

Quan¹,

Yao²,

Li³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

The success of deep learning methods relies on the availability of well-labeled large-scale datasets. However, for medical images, annotating such abundant training data often requires experienced radiologists and consumes their limited time. Few-shot learning is developed to alleviate this burden, which achieves competitive performances with only several labeled data. However, a crucial yet previously overlooked problem in few-shot learning is about the selection of template images for annotation before learning, which affects the final performance. We herein propose a novel Sample Choosing Policy (SCP) to select "the most worthy" images for annotation, in the context of few-shot medical landmark detection. SCP consists of three parts: 1) Self-supervised training for building a pre-trained deep model to extract features from radiological images, 2) Key Point Proposal for localizing informative patches, and 3) Representative Score Estimation for searching the most representative samples or templates. The advantage of SCP is demonstrated by various experiments on three widelyused public datasets. For one-shot medical landmark detection, its use reduces the mean radial errors on Cephalometric and HandXray datasets by 14.2% (from 3.595mm to 3.083mm) and 35.5% (4.114mm to 2.653mm), respectively.

show abstract

SATr: Slice Attention with Transformer for Universal Lesion Detection

Chen

Han

et al. 2022

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Conditional Training with Bounding Map for Universal Lesion Detection

Cited by 13 publications

References 30 publications

SATr: Slice Attention with Transformer for Universal Lesion Detection

SATr: Slice Attention with Transformer for Universal Lesion Detection

Which images to label for few-shot medical landmark detection?

SATr: Slice Attention with Transformer for Universal Lesion Detection

Contact Info

Product

Resources

About