FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3003898841> ?p ?o ?g. }

• W3003898841 abstract "Last decades have witnessed an unprecedented expansion of medical data in various largescale and complex systems. While achieving a lot of successes in many complex medical problems, there are still some challenges to deal with. Class imbalance is one of the common problems of medical image segmentation. It occurs mostly when there is a severely unequal class distribution, for instance, when the size of target foreground region is several orders of magnitude less that the background region size. In such problems, typical loss functions used for convolutional neural networks (CNN) segmentation fail to deliver good performances.Widely used losses,e.g., Dice or cross-entropy, are based on regional terms. They assume that all classes are equally distributed. Thus, they tend to favor the majority class and misclassify the target class. To address this issue, the main objective of this work is to build a boundary loss, a distance based measure on the space of contours and not regions. We argue that a boundary loss can mitigate the problems of regional losses via introducing a complementary distance-based information. Our loss is inspired by discrete (graph-based) optimization techniques for computing gradient flows of curve evolution.Following an integral approach for computing boundary variations, we express a non-symmetric L2 distance on the space of shapes as a regional integral, which avoids completely local differential computations. Our boundary loss is the sum of linear functions of the regional softmax probability outputs of the network. Therefore, it can easily be combined with standard regional losses and implemented with any existing deep network architecture for N-dimensional segmentation (N-D).Experiments were carried on three benchmark datasets corresponding to increasingly unbalanced segmentation problems: Multi modal brain tumor segmentation (BRATS17), the ischemic stroke lesion (ISLES) and white matter hyperintensities (WMH). Used in conjunction with the region-based generalized Dice loss (GDL), our boundary loss improves performance significantly compared to GDL alone, reaching up to 8% improvement in Dice score and 10% improvement in Hausdorff score. It also yielded a more stable learning process." @default.
• W3003898841 created "2020-02-07" @default.
• W3003898841 creator A5014776120 @default.
• W3003898841 date "2019-05-27" @default.
• W3003898841 modified "2023-09-27" @default.
• W3003898841 title "Surface loss for medical image segmentation" @default.
• W3003898841 cites W1498436455 @default.
• W3003898841 cites W1507175590 @default.
• W3003898841 cites W1546787007 @default.
• W3003898841 cites W1555845716 @default.
• W3003898841 cites W1677182931 @default.
• W3003898841 cites W1686810756 @default.
• W3003898841 cites W1783315696 @default.
• W3003898841 cites W1884191083 @default.
• W3003898841 cites W1901129140 @default.
• W3003898841 cites W1903029394 @default.
• W3003898841 cites W1910109443 @default.
• W3003898841 cites W1972626213 @default.
• W3003898841 cites W2030216529 @default.
• W3003898841 cites W2038218396 @default.
• W3003898841 cites W2043075535 @default.
• W3003898841 cites W2048387604 @default.
• W3003898841 cites W2054131729 @default.
• W3003898841 cites W2075752599 @default.
• W3003898841 cites W2080323815 @default.
• W3003898841 cites W2095049733 @default.
• W3003898841 cites W2097117768 @default.
• W3003898841 cites W2099290282 @default.
• W3003898841 cites W2101926813 @default.
• W3003898841 cites W2103870502 @default.
• W3003898841 cites W2112796928 @default.
• W3003898841 cites W2121193622 @default.
• W3003898841 cites W2121947440 @default.
• W3003898841 cites W2122034173 @default.
• W3003898841 cites W2127890285 @default.
• W3003898841 cites W2128806031 @default.
• W3003898841 cites W2135056195 @default.
• W3003898841 cites W2160754664 @default.
• W3003898841 cites W2161236525 @default.
• W3003898841 cites W2169551590 @default.
• W3003898841 cites W2226808373 @default.
• W3003898841 cites W2253429366 @default.
• W3003898841 cites W2274227799 @default.
• W3003898841 cites W2281949732 @default.
• W3003898841 cites W2284198383 @default.
• W3003898841 cites W2301358467 @default.
• W3003898841 cites W2310992461 @default.
• W3003898841 cites W2337429362 @default.
• W3003898841 cites W2338271170 @default.
• W3003898841 cites W2346062110 @default.
• W3003898841 cites W2396622801 @default.
• W3003898841 cites W2419448466 @default.
• W3003898841 cites W2524608787 @default.
• W3003898841 cites W2556177465 @default.
• W3003898841 cites W2567599812 @default.
• W3003898841 cites W2589409328 @default.
• W3003898841 cites W2589647984 @default.
• W3003898841 cites W2604790786 @default.
• W3003898841 cites W2608353599 @default.
• W3003898841 cites W2620296437 @default.
• W3003898841 cites W2734349601 @default.
• W3003898841 cites W274818618 @default.
• W3003898841 cites W2763160469 @default.
• W3003898841 cites W2767896621 @default.
• W3003898841 cites W2771088639 @default.
• W3003898841 cites W2793954249 @default.
• W3003898841 cites W2891118631 @default.
• W3003898841 cites W2891451067 @default.
• W3003898841 cites W2897772195 @default.
• W3003898841 cites W2899771611 @default.
• W3003898841 cites W2901462650 @default.
• W3003898841 cites W2902800533 @default.
• W3003898841 cites W2903617078 @default.
• W3003898841 cites W2962914239 @default.
• W3003898841 cites W2963446712 @default.
• W3003898841 cites W2963874375 @default.
• W3003898841 cites W2964098128 @default.
• W3003898841 cites W2964121744 @default.
• W3003898841 cites W2964288706 @default.
• W3003898841 cites W2964290592 @default.
• W3003898841 cites W3101414204 @default.
• W3003898841 hasPublicationYear "2019" @default.
• W3003898841 type Work @default.
• W3003898841 sameAs 3003898841 @default.
• W3003898841 citedByCount "0" @default.
• W3003898841 crossrefType "journal-article" @default.
• W3003898841 hasAuthorship W3003898841A5014776120 @default.
• W3003898841 hasConcept C11413529 @default.
• W3003898841 hasConcept C124504099 @default.
• W3003898841 hasConcept C126255220 @default.
• W3003898841 hasConcept C134306372 @default.
• W3003898841 hasConcept C154945302 @default.
• W3003898841 hasConcept C167981619 @default.
• W3003898841 hasConcept C188441871 @default.
• W3003898841 hasConcept C33923547 @default.
• W3003898841 hasConcept C41008148 @default.
• W3003898841 hasConcept C45374587 @default.
• W3003898841 hasConcept C62354387 @default.
• W3003898841 hasConcept C81363708 @default.
• W3003898841 hasConcept C89600930 @default.

• next