Skin Cancer Tumor Detection
Rohan Peddireddy
Lay Summary:
Skin cancer is a huge issue around the world as tumors can be mistaken for other abnormalities on the skin. To help stop skin tumors from being mistaken and the skin cancer from progressing to further stages I trained a model to detect tumors on the skin at multiple different stages, for a more efficient way to detect skin cancer as earlier stages when tumors aren't as prominent.
Abstract:
Skin cancer is a major and potentially deadly disease, making early and accurate detection crucial for quick help. My research focused on developing a neural network-based model to detect skin cancer tumors with precision. I programmed and trained a neural network using a dataset of labeled skin tumor images, which included both benign and malignant cases. The model was designed to differentiate between these cases by learning patterns in the image data. After testing, the neural network showed accuracy in identifying malignant tumors. This research gives the potential of neural networks in enhancing skin cancer detection, giving an efficient tool and mobile method for early diagnosis. My future work will aim to refine the model further, using larger datasets and more diagnostic features to improve its daily applicability.
Q&A:
Bios: Rohan Peddireddy
Program Track: Skills Development
GitHub Username:
rohanpeddireddy -Rohan Peddireddy
What was your favorite seminar? Why?
Virtual RNA inference work in colon and skin because I find RNA inference and Gene editing very interesting due to the abilities they have to change physical properties of any living thing.
-Rohan Peddireddy
If you were to summarize your summer internship experience in one sentence, what would it be?
A diverse research project allowing people to pick what they want to research about. -Rohan Peddireddy
Blog Post
Blog Post
** **Team 8 - Rohan Peddireddy
Skill Development Track
Skin Cancer Tumor Detection
Mentors: Suchir Paruchuri, Khang, Dr. Joshua Levy
Skin cancer is a huge issue around the world as tumors can be mistaken for other abnormalities on the skin. To help stop skin tumors from being mistaken and the skin cancer from progressing to further stages I trained a model to detect tumors on the skin at multiple different stages, for a more efficient way to detect skin cancer as earlier stages when tumors aren't as prominent.
#
With the help of my mentors and Dr. Levy, I found many datasets containing images of different types of skin tumors. I chose the Hams 10000 dataset as it was the most diverse and had 8 different types of skin abnormalities sorted out. From these abnormalities, I picked the 7 most common and used those to train my model. The biggest problem I had while training my neural networks was when I continuously overfit the model, to deal with this I did some research and used my code code to use data augmentation to prevent overfitting. My neural network has an accurate recognition of the various types of skin abnormalities and can differentiate between the multiple different types.
#
This summer I made a successful working neural network, something not many people can say they did at my age. I had to deal with complications and more problems that people have to face daily. I gained a further love and passion for AI and research and has further interested me in biomedical research. I have gained skills in Python and many more helpful skills. I was able to further reach out to pretty much anyone in the levy lab and they were all willing to help and motivate me. I'm really glad I was able to have the chance to participate this summer, I truly will not forget what I learned.
#
A special thank you to Doctor Levy for giving me this opportunity. And thanks to my mentors Suchir and Khang, seminar broadcasters, and my fellow team members.