Colored multi-neuron image processing for segmenting and tracing neural circuits

Shao, Hao-Chiang; Cheng, Wei‐Yun; Chen, Yung‐Chang; Hwang, Wen-Liang

doi:10.1109/icip.2012.6467287

Cited by 8 publications

(2 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…By integrating the 3D alpha matting technique into the context-aware co-segmentation framework [23], the proposed PET-CT co-segmentation method eases the design of cost functions for segmentation, and significantly outperforms the state-of-theart PET-CT segmentation approach [23]. Note that although 3D matting has been used as post-processing for the refinement of segmentation [27,28], no previous work has been done using it for cost function design in the segmentation framework based on graph algorithms.…”

Section: Introductionmentioning

confidence: 99%

3D Alpha Matting Based Co-segmentation of Tumors on PET-CT Images

Zhi

Kim

Buatti

et al. 2017

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Positron emission tomography-computed tomography (PET-CT) has been widely used in modern cancer imaging. Accurate tumor delineation from PET and CT plays an important role in radiation therapy. The PET-CT co-segmentation technique, which makes use of advantages of both modalities, has achieved impressive performance for tumor delineation. In this work, we propose a novel 3D image matting based semi-automated co-segmentation method for tumor delineation on dual PET-CT scans. The "matte" values generated by 3D image matting are employed to compute the region costs for the graph based co-segmentation. Compared to previous PET-CT cosegmentation methods, our method is completely data-driven in the design of cost functions, thus using much less hyper-parameters in our segmentation model. Comparative experiments on 54 PET-CT scans of lung cancer patients demonstrated the effectiveness of our method.

show abstract

Section: Introductionmentioning

confidence: 99%

3D Alpha Matting Based Co-segmentation of Tumors on PET-CT Images

Zhi

Kim

Buatti

et al. 2017

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…The ability to trace in all 3D spatial dimensions simultaneously is also critical when handling axonal projections, in which fibers running principally along the z dimension could be very hard to detect in 2D hyper-stacks. Fully automated image segmentation is the ultimate goal, and this has been approached based upon applying spatio-color relations to link voxels from individual cells ( Bas and Erdogmus, 2010 ; Shao et al, 2012 ; Sumbul et al, 2016 ; Duan et al, 2021 ). However, while such automated image segmentation is at the technology development, rather than application stage, and the success of current iterations is constrained by variations in color within single neurons, The current alternative to multi-color labeling approaches is monochromatic sparse labeling, which can resolve the anatomy of individual neurons, but only when processes are not overlapping to ensure accurate single cell reconstruction.…”

Section: Discussionmentioning

confidence: 99%

Quantitative characterisation of ipRGCs in retinal degeneration using a computation platform for extracting and reconstructing single neurons in 3D from a multi-colour labeled population

Procyk

Rodgers²,

Zindy³

et al. 2022

Front. Cell. Neurosci.

View full text Add to dashboard Cite

Light has a profound impact on mammalian physiology and behavior. Intrinsically photosensitive retinal ganglion cells (ipRGCs) express the photopigment melanopsin, rendering them sensitive to light, and are involved in both image-forming vision and non-image forming responses to light such as circadian photo-entrainment and the pupillary light reflex. Following outer photoreceptor degeneration, the death of rod and cone photoreceptors results in global re-modeling of the remnant neural retina. Although ipRGCs can continue signaling light information to the brain even in advanced stages of degeneration, it is unknown if all six morphologically distinct subtypes survive, or how their dendritic architecture may be affected. To answer these questions, we generated a computational platform−BRIAN (Brainbow Analysis of individual Neurons) to analyze Brainbow labeled tissues by allowing objective identification of voxels clusters in Principal Component Space, and their subsequent extraction to produce 3D images of single neurons suitable for analysis with existing tracing technology. We show that BRIAN can efficiently recreate single neurons or individual axonal projections from densely labeled tissue with sufficient anatomical resolution for subtype quantitative classification. We apply this tool to generate quantitative morphological information about ipRGCs in the degenerate retina including soma size, dendritic field size, dendritic complexity, and stratification. Using this information, we were able to identify cells whose characteristics match those reported for all six defined subtypes of ipRGC in the wildtype mouse retina (M1−M6), including the rare and complex M3 and M6 subtypes. This indicates that ipRGCs survive outer retinal degeneration with broadly normal morphology. We additionally describe one cell in the degenerate retina which matches the description of the Gigantic M1 cell in Humans which has not been previously identified in rodent.

show abstract