A unified uncertainty network for tumor segmentation using uncertainty cross entropy loss and prototype similarity

Diao, Zhaoshuo; Jiang, Huiyan; Shi, Tianyu

doi:10.1016/j.knosys.2022.108739

Cited by 17 publications

(9 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For uncertainty estimation the variance or the entropy of predictions for a given image are calculated. In Diao et al (2022) and Rosvoll Groendahl et al (2021) entropy-based loss functions are used to create output uncertainty maps. In Diao et al (2022), the concept of pixel-level uncertainty is explored in the tumor segmentation task.…”

Section: Discussionmentioning

confidence: 99%

“…In Diao et al (2022) and Rosvoll Groendahl et al (2021) entropy-based loss functions are used to create output uncertainty maps. In Diao et al (2022), the concept of pixel-level uncertainty is explored in the tumor segmentation task. They designed a method which outputs the segmentation output, the pixel uncertainty maps and the case level uncertainty used for out-of-distribution detection.…”

Section: Discussionmentioning

confidence: 99%

“…Tumor segmentation, in fact, is more challenging, since tumors have variable presentation, i.e. they are different in location, size, and shape (see figure 1) (Diao et al 2022).…”

Section: Introductionmentioning

confidence: 99%

“…For the introduction in the clinic of DL auto-segmentation models and to gain the trust of the radiation oncologists in those methods, it would be beneficial to provide them with information about how reliable segmentations are and how confident the network is in predicting results (Wang et al 2019, Diao et al 2022. Uncertainty estimation of segmentation results, in fact, gained great importance in recent years.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Deep learning aided oropharyngeal cancer segmentation with adaptive thresholding for predicted tumor probability in FDG PET and CT images

et al. 2023

View full text Add to dashboard Cite

Tumor segmentation is a fundamental step for radiotherapy treatment planning. To define an accurate segmentation of the primary tumor (GTVp) of oropharyngeal cancer patients (OPC) each image volume is explored slice-by-slice from different orientations on different image modalities. However, the manual fixed boundary of segmentation neglects the spatial uncertainty known to occur in tumor delineation. This study proposes a novel deep learning-based method that generates probability maps which capture the model uncertainty in the segmentation task. We included 138 OPC patients treated with (chemo)radiation in our institute. Sequences of 3 consecutive 2D slices of concatenated FDG-PET/CT images and GTVp contours were used as input. Our framework exploits inter and intra-slice context using attention mechanisms and bi-directional long short term memory (Bi-LSTM). Each slice resulted in three predictions that were averaged. A 3-fold cross validation was performed on sequences extracted from the axial, sagittal, and coronal plane. 3D volumes were reconstructed and single- and multi-view ensembling was performed to obtain final results. The output is a tumor probability map determined by averaging multiple predictions. Model performance was assessed on 25 patients at different probability thresholds. Predictions were the closest to the GTVp at a threshold of 0.9 (mean surface DSC of 0.81, median HD95 of 3.906 mm). The promising results of the proposed method show that is it possible to offer the probability maps to radiation oncologists to guide them in a in a slice-by-slice adaptive GTVp segmentation.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

“…Tumor segmentation, in fact, is more challenging, since tumors have variable presentation, i.e. they are different in location, size, and shape (see figure 1) (Diao et al 2022).…”

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Deep learning aided oropharyngeal cancer segmentation with adaptive thresholding for predicted tumor probability in FDG PET and CT images

et al. 2023

View full text Add to dashboard Cite

show abstract

“…Various terms such as trust, certainty, and uncertainty play an essential role in artificial intelligence applications. In the real world, these terms have a great deal of significance (Abdar, Fahami, Rundo et al, 2022; Abdar, Khosravi, et al, 2022; Abdar, Pourpanah, et al, 2021; Diao et al, 2022). In medical studies, predicting AI methods with different uncertainties can be used to measure trust in AI, with the implied assumption that trust in AI can be measured in different ways.…”

Section: Uncertainty Quantification In Hsi and Medical Sectormentioning

confidence: 99%

Hyperspectral imaging for early diagnosis of diseases: A review

et al. 2023

View full text Add to dashboard Cite

Hyperspectral Imaging (HSI) has grown to be one of the most crucial optical imaging modalities with applications in numerous industries. The non‐invasive nature of HSI has led to widening its horizon to the medical domain, especially in areas like early diagnosis of various diseases. HSI combines both imaging and spectroscopy properties, thereby exploiting spectral and spatial dimensions of images captured, providing quick and accurate interpretation of data. The current study aims to give an exhaustive overview of HSI's applications pertaining to the medical industry for fast detection of diseases and aiding in surgical procedures. The survey focuses on hyperspectral imaging combined with various approaches ‐ machine learning, deep learning, genetic algorithms, and anomaly detection for the treatment of disorders. In addition, the survey highlights accompanying pre‐processing approaches, performance metrics, inferences, and future prospects of HSI in the medical domain. The current study can gauge computer vision specialists, researchers in machine and deep learning domain, doctors, and scientists by giving them a platform for improving existing treatment methods for the betterment of society.

show abstract