An Interactive Editing Framework for Electron Microscopy Image Segmentation

Abstract

Abstract. There are various automated segmentation algorithms for medical images. However, 100% accuracy may be hard to achieve because medical images usually have low contrast and high noise content. These segmentation errors may require manual correction. In this paper, we present an interactive editing framework that allows the user to quickly correct segmentation errors produced by automated segmentation algorithms. The framework includes two editing methods: (1) editing through multiple choice and (2) interactive editing through graph cuts. The first method provides a set of alternative segmentations generated from a confidence map that carries valuable information from multiple cues, such as the probability of a boundary, an intervening contour cue, and a soft segmentation by a random walker. The user can then choose the most acceptable one from those segmentation alternatives. The second method introduces an interactive editing tool where the user can interactively connect or disconnect presegmented regions. The editing task is posed as an energy minimization problem: We incorporate a set of constraints into the energy function and obtain an optimal solution via graph cuts. The results show that the proposed editing framework provides a promising solution for the efficient correction of incorrect segmentation results.

IntellEditS: Intelligent Learning-Based Editor of Segmentations

“…While combining learning with energy minimization has proven successful in interactive segmentation, it has not been used to locally edit presegmentations. Other energy-minimization-based editors only employ volume data to calculate w ij [6,3,10,15]. As a result, these methods neglect the highly informative ensemble of local foreground and background voxel features that users implicitly specify during edits.…”

“…Apart from 2D-focused tools [10,15], many 3D editors function by propagating user corrections from an interaction plane to the larger 3D volume [6,3,8]. Propagation can be achieved by minimizing an energy term constrained by user input, the presegmentation, and the volumetric data [6].…”

A splice-guided data driven interactive editing

“…Algorithms that ignore presegmentation (eg. [4], [8]) do not perform editing that benefits from the current segmentation but rather it solves a new segmentation problem in a neighborhood defined by the user input. In [4], the segmentation problem is solved using a live-wire framework [1] and in [8], it is solved using graph cuts [2].…”

Interactive 3D segmentation repair with image-foresting transform, supervoxels and seed robustness

“…(a) El-Zehiri et al [14] (b) Harrison et al [15] (c) Jackowski et al [8] (d) Valenzuela et al [10] (e) Yang et al [11] (f) Karimov et al [17] (g) Miranda et al [12,13] El-Zehiri et al [14] and Grady and Funka-Lea [40] apply RW [25] to correct presegmented images, optimized with downsampling, which loses important and high frequency information, like small objects, negatively affecting the result. Harrison et al [15] join discriminative classification and energy minimization with RW for contour-based correction, using GPU training.…”

“…For both described problems, the history of presegmentation should be estimated, which is a reverse segmentation problem, as depicted in Figure 1 Although the segmentation problem has motivated the development of a variety of works, the problem of editing segmentations has not caught much attention. Parametric surfaces [7][8][9][10], energy minimization [11], low level editing [8][9][10], region-based segmentation [12,13], edgebased ones [14][15][16], graphs [7,[11][12][13]15] or Human-Computer Interface [7,17] have been applied. These methods differ in user effort degree, complexity, running time, flexibility and robustness of algorithm.…”