Reliability of the BI-RADS Final Assessment Categories and Management Recommendations in a Telemammography Context

Salazar, Antonio J.; Romero, Javier; Bernal, Óscar; Moreno, Ángela; Velasco, Sofia

doi:10.1016/j.jacr.2016.08.004

Cited by 6 publications

(2 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Another challenge of health risk level estimation stems from the label quality. For instance, the agreement rate of the radiologists for malignancy is usually less than 80%, resulting in a noisy labeled dataset [38,47]. Despite the often-uncleared distinction between adjacent labels, it is more possible that a well-trained annotator will mislabel a Severe DR (3) sample to Moderate DR (2) rather than No DR (0).…”

Section: Introductionmentioning

confidence: 99%

Unimodal-Uniform Constrained Wasserstein Training for Medical Diagnosis

Liu

Han

Qiao

et al. 2019

2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW)

View full text Add to dashboard Cite

The labels in medical diagnosis task are usually discrete and successively distributed. For example, the Diabetic Retinopathy Diagnosis (DR) involves five health risk levels: no DR (0), mild DR (1), moderate DR (2), severe DR (3) and proliferative DR (4). This labeling system is common for medical disease. Previous methods usually construct a multi-binary-classification task or propose some reparameter schemes in the output unit. In this paper, we target on this task from the perspective of loss function. More specifically, the Wasserstein distance is utilized as an alternative, explicitly incorporating the inter-class correlations by pre-defining its ground metric. Then, the ground metric which serves as a linear, convex or concave increasing function w.r.t. the Euclidean distance in a line is explored from an optimization perspective.Meanwhile, this paper also proposes of constructing the smoothed target labels that model the inlier and outlier noises by using a unimodal-uniform mixture distribution. Different from the one-hot setting, the smoothed label endues the computation of Wasserstein distance with more challenging features. With either one-hot or smoothed target label, this paper systematically concludes the practical closed-form solution. We evaluate our method on several medical diagnosis tasks (e.g., Diabetic Retinopathy and Ultrasound Breast dataset) and achieve state-of-the-art performance.

show abstract

Section: Introductionmentioning

confidence: 99%

Unimodal-Uniform Constrained Wasserstein Training for Medical Diagnosis

Liu

Han

Qiao

et al. 2019

2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW)

View full text Add to dashboard Cite

show abstract

“…• The distinction between adjacent labels is often unclear and annotator-dependent, leading to a situation where the same input may be marked with different (although probably adjacent) labels by different practitioners (or even by the same one). Radiologist usually agree on malignancy of less than 80% cases ( [8]), whereas agreement rate may be even lower on predicting individual BIRADS labels ( [10]).…”

Section: Introductionmentioning

confidence: 99%

Learning multiple non-mutually-exclusive tasks for improved classification of inherently ordered labels

Ratner,

Shoshan,

Kachman

2018

Preprint

View full text Add to dashboard Cite

Medical image classification involves thresholding of labels that represent malignancy risk levels. Usually, a task defines a single threshold, and when developing computer-aided diagnosis tools, a single network is trained per such threshold, e.g. as screening out healthy (very low risk) patients to leave possibly sick ones for further analysis (low threshold), or trying to find malignant cases among those marked as non-risk by the radiologist ("second reading", high threshold). We propose a way to rephrase the classification problem in a manner that yields several problems (corresponding to different thresholds) to be solved simultaneously. This allows the use of Multiple Task Learning (MTL) methods, significantly improving the performance of the original classifier, by facilitating effective extraction of information from existing data.

show abstract