skinCAM: An Effective Solution towards Skin Disease Diagnosis via Novel Deep Learning Algorithm
CSEF · 2018 Mathematical Sciences
Overview
Objectives/Goals This project resolves a medical issue that many people face regarding dermatology. Skin diseases are common yet typically benign, but for an accurate diagnosis, patients have to visit a dermatologist, a costly endeavor. However, with the building of a convolutional neural network in this novel application of artificial intelligence, patients will be able to self-diagnose diseases that they may have. Methods/Materials Our neural network was created to train on a database of over 5000 distinct images. The chosen training sets came from trusted online databases that were validated by professional dermatologists (AAD, dermWeb, webMD). We built the model by creating five convolutional layers: each layer containing the convolutional filter for highlighting the features of the input, the ReLU layer as an activation function, and the pooling layer to minimize computation time. Then, we passed the output of the convolutional layers into a fully connected layer which used backpropagation and gradient descent to move the model towards optimal accuracy by updating the filters. The final output of the program was a probability value between 0 and 1, indicating the disease to the app. Results After training the neural network through 1000 iterations with a training step of 0.01, the prediction of the model produced an accuracy of 95.2% from 1000 tests. Each test ran the image through a separate distributable code that utilized the training data to confirm a valid prediction. Based off of the large testing sample size, the program can be statistically accepted since its margin of error is well within the pre-established design criteria. Conclusions/Discussion Our study demonstrates that an artificial intelligence approach can be utilized for an initial diagnosis, providing users with useful information about whether or not they should consult a dermatologist for further treatment. The app not only saves time and cuts expenses, but it is also equally as important by providing a unique tool for dermatologists around the world.
Summary statement
We created a convolutional neural network to accurately diagnose skin diseases while also implementing the algorithm into an app for patients to use.
Help received
The project was created entirely without outside help. However, we have recently reached out to Dr. Sicun Gao and Dr. Paravar for future studies in improving the accuracy of the model with powerful cloud-computing tools and with dermatologists through testing the app. We have also gotten help from
Competition history
- CSEF 2018
Resources
Related projects
CSEF · 2019
SkinSight: A Novel Implementation of a Convolutional Neural Network to Recognize Skin Diseases
ISEF · 2023
DermaSkan: Using Convolutional Neural Networks to Detect and Prevent Skin Cancer in Underrepresented Communities by Using an Android App
ISEF · 2017
Scin: An Application to Detect and Diagnose Plant and Skin Diseases Utilizing a Feed Foward Convolutional Neural Network
ISEF · 2023
Derma X
CSEF · 2014
Accurate Detection of Skin Cancer Using Multi Stage Neural Networks
CSEF · 2019
Developing a Novel, Accurate, and Rapid Machine Learning Based Skin Disease Diagnosis Algorithm and Mobile Application
CSEF · 2015
DermatoScan: Machine Vision, Analysis, Learning & Natural Computing Optimizations for the Early Detection of Skin Cancer
CSEF · 2016
Modeling Deep Learning Neural Net Based Image Recognition to Classify Melanoma Better than a Trained Professional
Closest projects by meaning, across every fair and year in the corpus.
Browse more like this
Source: California Science & Engineering Fair public projects