A Machine Learning Based Approach to Decrease the Lung Cancer Malignancy Detection Threshold
CSEF · 2018 Computational Systems & Analysis
Overview
Objectives/Goals When creating our Machine Learning model, our objective was to be able to correctly identify the malignancy of the Lung presented based on the CT scan images. This was achieved through three main criteria. These included maximizing the classification accuracy of the model, minimizing the model's training time, and minimizing the model's overall run-time. We also wanted to measure the accuracy of our model in comparison to Google's Inception model. Our hope is that our model can be utilized in doctor's offices to forego the need to do biopsies or any other stressful surgeries. Methods/Materials Equipment: PC equipped with Nvidia GeForce GTX 980 GPU, Laptop with Nvidia GeForce GTX 1050. Software Components: Tensorflow - Google's Machine Learning and Artificial Intelligence Library for Python, CT scan image database acquired from a cancer imaging database, Google's Inception Deep Learning model Results After testing our model, we achieved around a 96% accuracy when classifying the cancer the patient had as being Benign, Malignant, or Metastatic, based on the testing data of CT scan images. The data that we used was sectioned off into training and testing sets to insure that the images in the testing sets haven't been seen by the model during training. This would result in an unbiased accuracy output. When compared to Google's Inception Deep Learning model, we consistently achieved around a 20 to 30% higher accuracy. These results were especially surprising, as we were able to achieve very high accuracy from a simplistic model in contrast to Inception which is extremely complex. Conclusions/Discussion Based on the accuracy we achieved with our model, we want to refine it so that doctors, especially radiologists and pulmonologists, can utilize this software to make an accurate diagnosis of the patient's condition without having to perform any biopsies. This will get rid of any stress or anxiety that is often attached to such procedures. By getting an accurate diagnosis of malignancy of the cancer early on, respective action can be taken without any delay, giving the patient peace of mind.
Summary statement
By creating a simple Convolutional Neural Network, we were able to achieve a surprisingly high accuracy when diagnosing the malignancy of the patient's Lung Cancer.
Help received
I received help from my dad in understanding some of the fundamentals of Machine Learning.
Competition history
- CSEF 2018
Resources
Related projects
ISEF · 2023
Detection of Malignant Lung Tumors From CT Scans Through Deep Learning-Based Artificial Intelligence Algorithms
AJAS · 2020
Lung Cancer Detection Using Nodule Based Methods and Machine Learning Algorithms
ISEF · 2019
A Lung Cancer Prediction and Detection System Using Nodule Based Methods and Machine Learning Algorithms
ISEF · 2016
Lung Cancer Decision Support System: Novel Automated Noninvasive Tumor Malignancy and Patient Outcomes Prediction Modelling through Radiomics Phenotype Feature Quantification
ISEF · 2017
Predicting Lung Cancer Onset Using Segmentation and Classification
ISEF · 2022
Detection of Benign and Malignant Lung Nodules in 3D Volumes Generated From Thoracic Computed Tomography Scans Leveraging Artificial Intelligence, Year 2
CSEF · 2019
Machine Learning Ensemble Model for Improved Personalized Lung Cancer Risk Assessment and Malignant Nodule Detection
ISEF · 2015
It's a Matter of Life and Breath: A Novel Computer Diagnosis Tool for Early Detection of Lung Cancer SPNs towards Big Data Analysis
Closest projects by meaning, across every fair and year in the corpus.
Browse more like this
Source: California Science & Engineering Fair public projects