MAJOR PROJECT WORK

on
“ENHANCING SKIN LESION SEGMENTATION USING BIDIRECTIONAL CONVLSTM U-NET
WITH DENSELY CONNECTED CONVOLUTIONS FOR IMPROVED PRECISION ”

Supervisor & MRG-Coordinator : Presented by :

Dr. M. Sujatha A. Akhila (B21CS186L)
Assistant Professor, Dept. of CSE. Hasnain Akbar (B20CS007)
V. Saikrishna (B20CS035)
Coordinator :
V. Naga Thrisha Sri Reddy (B20CS033)
Dr. K. Vinay Kumar
Assistant Professor, Dept. of CSE.
CONTENTS
• Introduction
• Abstract
• Methodology
• Software and Hardware Requirements
• Dataset
• Implementation
• Outcomes
• References
INTRODUCTION
• Skin Lesion Segmentation is used to identify the regions of Interest within the skin and distinguish between
normal skin and areas that may exhibit anomalies such as lesions.

• Accurate Segmentation of Skin Lesions is essential for identifying and isolating potential anomalies at an
early stage and facilitating timely treatment.

• Bidirectional ConvLSTM: It is a kind of recurrent neural network (RNN) that can help capture
information from both past and future inputs by using this network's ability to process input data in both
forward and backward directions.

• U-Net: U-Net is also a neural network architecture that consists of a contracting path (encoder) and an
expansive path (decoder). It can capture high-resolution information through skip connections between the
encoder and decoder.

• Densely Connected Convolutions: This refers to densely connected blocks in which each layer receives
input from all previous layers in a block, promoting feature reuse. Dense connections help to encourage
feature propagation.
ABSTRACT

• Using the Bidirectional ConvLSTM we combine the feature maps extracted from the corresponding encoding
path and the previous decoding up-convolutional layer in a non-linear way.

• To strengthen feature propagation and encourage feature reuse, we use densely connected convolutions in the
last convolutional layer of the encoding path.

• Finally, we can accelerate the convergence speed of the proposed network by employing batch normalization
(BN).

• The Proposed Model is evaluated on the dataset called “ISIC(International Skin Imaging Collaboration) 2018”.
METHODOLOGY
MINIMUM SYSTEM REQUIREMENTS
• Operating System: Windows 7 or later, macOS 10.9 or later, or Linux (64-bit).

• Processor: Intel Core i5 or equivalent.

• GPU: NVIDIA GPU with CUDA support (e.g., NVIDIA GeForce RTX 30 series or NVIDIA A100).

• RAM: At least 16GB of RAM, preferably 32GB or more.

• Storage: SSD storage with sufficient capacity for datasets

• Python: Python 3.7 or later installed in your system.

• Libraries: Numpy, Matplotlib, Scikit learn

 DATASET

DERMOSCOPY (SKIN)
IMPLEMENTATION
IMPLEMENTATION
OUTCOMES
REFERENCES
[1] M. Z. Alom, M. Hasan, C. Yakopcic, T. M. Taha, and V. K. Asari. Recurrent residual convolutional neural
network based on u-net (r2u-net) for medical image segmentation. arXiv preprint arXiv:1802.06955, 2018.

[2] J. Staal, M. D. Abramoff, M. Niemeijer, M. A. Viergever, ` and B. Van Ginneken. Ridge-based vessel
segmentation in color images of the retina. IEEE transactions on medical imaging, 23(4):501–509, 2004.

[3] X. Zhou, T. Ito, R. Takayama, S. Wang, T. Hara, and H. Fujita. Three-dimensional ct image segmentation by
combining 2d fully convolutional network with 3d majority voting. In Deep Learning and Data Labeling for
Medical Applications, pages 111–120. Springer, 2016.
THANK YOU