https://github.com/pramod-zillella/skin-lesion-segmentation

Fully automatic skin lesion segmentation using the Berkeley wavelet transform and UNet algorithm.
https://github.com/pramod-zillella/skin-lesion-segmentation

deep-learning isic-challenge lesion-segmentation skin-lesion unet-segmentation

Last synced: 13 days ago
JSON representation

Fully automatic skin lesion segmentation using the Berkeley wavelet transform and UNet algorithm.

• Host: GitHub
• URL: https://github.com/pramod-zillella/skin-lesion-segmentation
• Owner: pramod-zillella
• Created: 2021-05-17T17:38:29.000Z (over 3 years ago)
• Default Branch: main
• Last Pushed: 2021-06-01T07:07:01.000Z (over 3 years ago)
• Last Synced: 2024-08-16T17:31:54.216Z (4 months ago)
• Topics: deep-learning, isic-challenge, lesion-segmentation, skin-lesion, unet-segmentation
• Language: Jupyter Notebook
• Homepage:
• Size: 4.45 MB
• Stars: 11
• Watchers: 1
• Forks: 2
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
## Skin-Lesion-Segmentation

### Background

Automatic skin lesion segmentation on dermoscopic images is an essential step in computer-aided diagnosis of melanoma. However, this task is challenging due to significant

variations of lesion appearances across different patients.Clinical treatment of skin lesion is primarily dependent on timely detection and delimitation

of lesion boundaries for accurate cancerous region localization.

For the accessing the interactive notebooks:

[![Binder](https://mybinder.org/badge_logo.svg)](https://mybinder.org/v2/gh/Pramod04121999/Skin-Lesion-Segmentation/HEAD)

### Introduction

We present a fully automated architecture for accurately detecting border and segmenting the skin lesion by coupling a deep learning model with the Berkeley wavelet transform map derived from the specific kernel filters. Our proposed method effectively combines the Berkeley wavelet transform feature maps into a deep learning U-Net model trained in an end-to-end manner, which results in the reduction of the number of trainable parameters. The model was trained on [HAM10000](https://www.nature.com/articles/s41591-020-0942-0) and [ISIC training data](https://ieeexplore.ieee.org/document/8363547) and test on [PH2](https://ieeexplore.ieee.org/document/6610779) and ISIC Validation data.

### Preprocessing

The dermoscopic images are pre-processed using [Berkeley wavelet transform](https://direct.mit.edu/neco/article/20/6/1537/7305/The-Berkeley-Wavelet-Transform-A-Biologically) by applying the 8 mother wavelets on the dermoscopic images to form the berkeley wavelet decomposed image. 



             



### Network architecture

![Network](https://user-images.githubusercontent.com/63542593/118666798-13bc6280-b811-11eb-93b7-697e7237b031.png)

### Quantitative Segmentation Results



  

| Dataset  | Sensitivity | Accuracy | Dice | Jaccard Similarity | 

| :---: | :---: | :---: | :---: | :---: |

|    [PH2](https://ieeexplore.ieee.org/document/6610779)    | 96.44  | 96.89  | 96.42  | 93.11  |

| [ISIC 2017](https://challenge.isic-archive.com/landing/2017) | 92.46  | 95.64  | 88.20  | 78.89  |



### Segmentation Visualization Results

######       Dermoscopic Image             BWT Feature Map                 Ground Truth             BWT+UNET Predicted Mask    Border Detected Lesion



             





             





             





             





             





             





             





             





             



### Conclusion

The Use of inexpensive Berkeley wavelet transform helps in enhancing the minutiae details of the skin lesion that helped in achieving an improved algorithm for accurate and automatic skin lesion segmentation. Furthermore, the model is computationally efficient with 1.2 Million parameters, and only takes 0.0625 seconds to segment the lesion from the dermoscopic image, making it extremely useful in clinical settings.