RadFormer: Transformers with global–local attention for interpretable and accurate Gallbladder Cancer detection

Basu, Soumen; Gupta, Mona; Rana, Pratyaksha; Gupta, Pankaj; Arora, Chetan

doi:10.1016/j.media.2022.102676

Cited by 14 publications

(13 citation statements)

References 45 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As a result, the network is compelled to learn a more refined representation of GB malignancy while reconstructing the masked tokens. We report an accuracy of 96.4% using our approach as against 84% by the current SOTA of GBCNet [5] and Radformer [6] 1 .…”

Section: Introductionmentioning

confidence: 89%

“…al [8] later utilized unsupervised contrastive learning to learn malignancy representations. On the other hand, [6] exploits a transformer-based dual-branch architecture for accurate and explainable GBC detection. [21] investigates application of transformers for differentiation of GBC with xanthogranulomatous cholecystitis.…”

Section: Related Workmentioning

confidence: 99%

“…GBCNet [5] CNN 0.840±0.105 0.843±0.204 0.843±0.072 US-UCL [8] CNN 0.808±0.127 0.871±0.217 0.776±0.109 RadFormer (SOTA) [6] Transformer 0.840±0.105 0.776±0.162 0.877±0.088…”

Section: Trainingmentioning

confidence: 99%

“…Gallbladder Cancer (GBC). Lately, automated GBC detection has drawn an increased interest from the researchers [5,6,10,31]. GBC is difficult to detect at an early stage [27], and surgical resection becomes infeasible for most patients as the disease gets detected at a late stage.…”

Section: Introductionmentioning

confidence: 99%

“…Malignant cases, unlike non-malignant ones with clear anatomy, are difficult to detect due to the lack of a distinct GB boundary or shape and the presence of masses. While there are recent efforts to circumvent the challenges of US for accurate GBC detection [5,6,8], these techniques are primarily image-based. Due to the challenges discussed earlier, single images may lack unambiguous features for malignancy detection.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Surpassing the Human Accuracy: Detecting Gallbladder Cancer from USG Images with Curriculum Learning

Basu

Gupta

Rana

et al. 2022

2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

View full text Add to dashboard Cite

In recent years, automated Gallbladder Cancer (GBC) detection has gained the attention of researchers. Current state-of-the-art (SOTA) methodologies relying on ultrasound sonography (US) images exhibit limited generalization, emphasizing the need for transformative approaches. We observe that individual US frames may lack sufficient information to capture disease manifestation. This study advocates for a paradigm shift towards video-based GBC detection, leveraging the inherent advantages of spatiotemporal representations. Employing the Masked Autoencoder (MAE) for representation learning, we address shortcomings in conventional image-based methods. We propose a novel design called FocusMAE to systematically bias the selection of masking tokens from high-information regions, fostering a more refined representation of malignancy. Additionally, we contribute the most extensive US video dataset for GBC detection. We also note that, this is the first study on US video-based GBC detection. We validate the proposed methods on the curated dataset, and report a new SOTA accuracy of 96.4% for the GBC detection problem, against an accuracy of 84% by current Image-based SOTA -GBCNet and RadFormer, and 94.7% by Video-based SOTA -AdaMAE. We further demonstrate the generality of the proposed FocusMAE on a public CTbased Covid detection dataset, reporting an improvement in accuracy by 3.3% over current baselines. Project page with source code, trained models, and data is available at: https://gbc-iitd.github.io/focusmae.

show abstract

Section: Introductionmentioning

confidence: 89%

Section: Related Workmentioning

confidence: 99%

“…GBCNet [5] CNN 0.840±0.105 0.843±0.204 0.843±0.072 US-UCL [8] CNN 0.808±0.127 0.871±0.217 0.776±0.109 RadFormer (SOTA) [6] Transformer 0.840±0.105 0.776±0.162 0.877±0.088…”

Section: Trainingmentioning

confidence: 99%