Skin Cancer Sentinel
Our solution takes in the image of skin as an input and processes it to return the score whether the person is suffering from Skin cancer or not and its type and connect users to the nearest doctors.
Created on 3rd November 2020
•
Skin Cancer Sentinel
Our solution takes in the image of skin as an input and processes it to return the score whether the person is suffering from Skin cancer or not and its type and connect users to the nearest doctors.
The problem Skin Cancer Sentinel solves
‘Truly said ‘Once Cancer happens, it changes the way you live for the rest of your life’
Skin cancer is a dangerous and widespread disease, and early detection increases the survival rate. It is found that a skilled dermatologist usually follows a series of steps, starting with the naked-eye observation of suspected lesions, then dermoscopy followed by biopsy. This would consume time and the disease may advance to later stages. Moreover, accurate diagnosis is subjective, depending on the skill of the clinician. In order to diagnose skin cancer speedily at the earliest stage, we need extensive research solutions by developing computer image analysis algorithms. We have come up with an AI Based Solution Compliance for predictive modelling and diagnosis of Skin Cancer disease.
The entire solution is divided into four parts:
- Trained Neural Network to predict disease and its stage ‘Benign or Malignant’.
- Web based frontend framework UI for users (patient).
- Geotagging of potential patients and referring them to the nearest doctor along with other information.
- Chatbot to assist users with diagnosis from the app and answer skin cancer related queries.
Challenges we ran into
The biggest challenge we ran into was to train our Neural Network model to achieve an accuracy of 85%. We trained our model on an efficient dataset provided by Internation Skin Imaging Collaboration- Malanoma Project ISIC. The data included images around 10, 000. We used the technique of Transfer Learning for Training our model to classify the disease as Melignant or Benign. We used weight layers from pre-trained model ResNet50 for this purpose and modified the last layer according to our classification. It is possible that user uploads an image that is not that of skin but is a random one having text image, sketch, or scenery.
Technologies used