Role:

Deep Learning Developer

Project Duration:

Mar 2021- May 2021

What did I do?

Literature Survey

Identifying Datasets

Defining the workflow

Coding the module(s)

Co-authoring the paper

Tools:

Visual Studio Code

Google Colaboratory

GitHub

WhiteBoard

Pen and Paper

Team Members:

Mrs. R Angeline (PhD)

Nithish Kanna Sivakumar

Ashwath Bhaskar

This paper was published in the Springer Artificial Intelligence in Healthcare on October 30 2021.

Research Paper

Identifying Malignancy of Lung Cancer Using Deep Learning Concepts

Problem Statement

Identifying if a tumor present in the lung is benign, 'unsure', or malignant using Deep Learning concepts on a dataset with CT, PET scans of the lungs of patients

What is the project about?

Our focus through this study is to work around detecting the type of tumor and improve the efficiency in identifying unsure or pre-malignant tumors using VGG-16 neural frameworks to train the LIDC-IDRI dataset.

Objectives

• Identify viable and suitable dataset
• Perform exploratory data analysis
• Perform image processing on the CT scans to identify pre-malignant tumors
• Improve the efficiency in identifying the pre-malignant tumors

Target Audience

• Doctors
• Radiologists
• Lab Technicians

Dataset Analysis

Analysis of the LIDC-IDRI dataset for lung cancer scans.

Results

                  **TRAINING SET VS VALIDATION SET**

TRAINING SET VS VALIDATION SET- CROSS ENTROPY

               **DETECTION OF EARLY STAGE LUNG NODULES**

                 **DETECTION OF LATER STAGE LUNG NODULES**

From our implementation, we found that 3% of every 100 cases are unsure pre-malignant nodules which correspond to appropriately 168 nodules are unsure since they have factors such as genetic lesions that predispose to malignant transformation or mutation abnormali-ties that cause the formation of asymptomatic nodules categorized as unsure or pre-malignant tumors