Date of Award

6-14-2019

Document Type

Thesis - SCU Access Only

Publisher

Santa Clara : Santa Clara University, 2019

Department

Computer Engineering

First Advisor

Tokunbo Ogunfunmi

Second Advisor

Silvia Figueira

Third Advisor

Yuling Yan

Abstract

Our senior design project was to develop an app that helps bring medical care to patients who don’t have accessible health care facilities; however, we only classify skin cancer for our senior design project. The goal of the app if it were to be implemented outside of the senior design project would be to let anyone with access to a phone use the app wherever they are by taking a picture of a potentially cancerous mole and receive a classification on the spot based on machine learning analysis clearly stating if the mole is cancerous or not. If the app diagnoses an image as malignant the patient then knows they should invest the time and money to get to the nearest doctor to have the mole checked and get the appropriate treatment. The final product we developed for our senior design project is only a prototype and proof-of-concept though as a health app would need to receive FDA approval before being okayed to be used by the general public. We would need to devote more attention to exploring machine learning models to increase the accuracy of the classification and run thorough tests as a part of the process to achieve FDA approval for an app like this. The product we developed is an app that users can take a picture of their mole with and then see the classification of the image they took, whether the mole is cancerous or not, on the app. We only classified Skin Cancer Images as cancerous or not, we did not classify if they were Melanoma or not. That required an extra layer of classification that we were not able to do within the time of our project because we wanted to complete all of our critical requirements. The classification accuracy we achieved was only about 82 percent so we recommend that the machine learning aspect of the project be further explored to find a model that achieves a higher accuracy. We used Flutter to develop our app and the random forest classifier for our machine learning model. We discuss the other technologies we used for our project in the following report, as well as our design plan going into the project, how we implemented the project, societal connections from the project, lessons learned, and future work for the project. IMPORTANT: PLEASE NOTE anywhere the word diagnose, diagnosis, diagnosed, or diagnoser is used in the paper and on the app which you’ll see screenshots of below for the sake of our senior design project only means a classification of an image as cancerous (Malignant) or non-cancerous (Benign), it does not imply an official diagnosis. Our completed work and submission for the senior design project is only a prototype and proof of concept of the idea, what we've developed for the senior design project is not for public use and classification results based on our project are not an official diagnosis.

Download

SCU Access Only

To access this paper, please log into or create an account in Scholar Commons using your scu.edu email address.

Share

COinS