-
-
Example of the program finding abnormalities and detecting cancer.
Inspiration
I recently got to know about someone who had lost their eye due to retinal cancer. they informed me that such a thing could have been prevented if detected early and it can be detected through the white eye reflections in flash images.
## What it does My project uses machine learning and AI along with Computer Vision to detect unusual "White eye reflections in images" which is scientifically called - leukocoria - this happens to suggest dangerous eye health problems like retinoblastoma (a type of eye cancer), cataracts, or retinal detachment.
How we built it
The app was built using Google’s Teachable Machine and TensorFlow.js to train a model on labeled images and run predictions directly in the browser.
Challenges we ran into
There were many challenges that i ran into such as i could not use public real people images to train the model due to ethical obligations so it was extremely tough to find public reports which gave access to their findings along with images to train the model . also i ran into a serious problem as due to an error the machine learning algorithm was not functioning properly. thankfully after hours of debugging i got it to work successfully.
## Accomplishments that we're proud of I am proud that even after many errors and enduring trouble finding training and testing data i managed to complete the project in time. (i only had a few days as i registered late)
What we learned
I learned how to preprocess images for machine learning, handle multiple image inputs, and troubleshoot model predictions in the browser. I also gained experience with AI ethics and working with sensitive medical data.
What's next for Cancer Detection App
Next I hope to include even more intricate and detailed analysis and symptom detection in detecting abnormalities. I hope to make sure that everyone can afford to scan and check their eyes for any health issues without spending thousands.
Log in or sign up for Devpost to join the conversation.