Name
Abstract | ||
|---|---|---|
In this paper, we present a novel strategy for automatic segmentation of biomedical images acquired from bio-engineered nerve tissues exhibiting variable morphological characteristics. Automatic image segmentation is one step towards the end goal of automatic analysis of the impact of various neurotoxic drug treatments on these artificial nerve tissues. We propose a deep learning architecture to perform this task. Our proposed architecture can be seen as a variation of U-Net that helps deal with a small manually annotated training data set. We present promising preliminary results and our human expert analysis shows that in some cases the model is even more precise in detecting the relevant morphological characteristics of the tissue compared to the manually annotated data. In the future, our model can be adapted for end-to-end automatic analysis of treated tissues. Moreover, based on a very small set of annotated data, it provides a reasonable segmentation to be used by human annotators. This will reduce the time of manual annotation significantly and streamline the process of generating a larger manually annotated data set for training our final ideal segmentation model. |
Year | DOI | Venue |
|---|---|---|
2019 | 10.1109/EMBC.2019.8856378 | 2019 41ST ANNUAL INTERNATIONAL CONFERENCE OF THE IEEE ENGINEERING IN MEDICINE AND BIOLOGY SOCIETY (EMBC) |
Field | DocType | Volume |
Training set,Computer vision,Pattern recognition,Segmentation,Computer science,Manual annotation,Image segmentation,Artificial intelligence,Deep learning,Neural tissues | Conference | 2019 |
ISSN | Citations | PageRank |
1557-170X | 0 | 0.34 |
References | Authors | |
0 | 4 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Hamid R. Karimian | 1 | 0 | 0.34 |
| Kevin J. Pollard | 2 | 0 | 0.34 |
| Michael J. Moore | 3 | 0 | 0.34 |
| Parisa Kordjamshidi | 4 | 143 | 18.52 |