README - Cosmos — ContextQMD

<h2> Cell Segmentation <a href="https://colab.research.google.com/github/Curovearth/Cell-Segmentation-and-Denoising/blob/main/Cell_Segmentation_.ipynb"></a></h2> <p>The study aims to determine a solution for the automatic segmentation and localization of cells. I have tried to utilise UNet++ architecture over UNet to detect cell nuclei and perform segmentation into individual objects.</p>

Image sets and experiment description

Dataset which was used was extracted from the mentioned repo data: <a href="https://github.com/Subham2901/Nuclei-Cell-segmentaion/tree/master/Data">Shubham2901/Nuclei Cell Segmentation</a>

Dataset	Images	Input Size	Modality	Provider
Cell Nuclei	670	96x96	microscopy	Kaggle Data Science Bowl 2018

<samp>Block diagram for the overall process</samp>

</p>

Results Obtained

<p>TP, FP, TN and FN are the numbers of true positive, false positive, true negative and false negative detections.</p>


<b>Precision Score</b>	0.91
<b>Recall Score</b>	0.85
<b>F-1 Score</b>	0.88
<b>Sensitivity</b>	0.85
<b>Specificity</b>	0.99
<b>IOU</b>	0.79
<b>AUC</b>	0.92

	Precision	Recall	F-1 Score	Support
False	0.98	0.99	0.98	5707370
True	0.91	0.85	0.88	846230
Accuracy			0.97	6553600
Macro Avg	0.95	0.92	0.93	6553600
Weighted Avg	0.97	0.97	0.97	6553600

Final Result

Future of Cell Segmentation

With the rise of size and complexity of cell images, the requirements for cells segmentation methods are also increasing. The basic image processing algo developed decades ago should not be the golden standards for dealing with these challenging cell segmentation problems any more.

On the contrary, development of more effective image processing algorithms is more promising for the progress of cell segmentation. In the meantime, comparison of these newly developed algorithms and teaching the biologists to use these newly developed algorithms are also very important. Hence, the open access and authoritative platforms are necessary for researchers all over the world to share, learn, and teach the data, codes, and algorithms.