Ahmed Majid

Can machine learning algorithms predict the presence of lung cancer?

Started May 11, 2023

Abstract or project description

Introduction

Lung cancer, often referred to as the “silent killer,” frequently goes under the radar, avoiding detection until its advanced stages. Often time, lung tumors are missed simply because nobody is looking for them. The late detection of the tumors limits treatment options for the patients, and severely decreases the patient’s survival time. Applying an algorithm that would run on any chest MRI, CT-Scan, PET Scan, etc would help to solve this problem as it would ensure lung cancer would no longer go undetected.

Hypothesis

I hypothesize that a predictive machine learning model, will successfully identify lung cancer nodules in chest imaging.

Methods

First, the data will need to be explored, adjusted, and properly cleaned in order to ensure it is fit to be used for the experiment. The datasets will be sectioned off into two groups, training and testing. The training data will be used to efficiently train the two algorithms that will be used — Neural Network and K-Nearest Neighbor algorithms will be used. Following the training stage, the algorithms will be fed the testing data; from there, the accuracy of the two algorithms will be recorded and compared.

Anticipated Results

From prior research, I expect that the Neural Network algorithm will be more effective at predicting the presence of lung cancer. Neural Networks are known to demonstrate exceptional performance in image analysis tasks, whereas KNN algorithms may struggle with certain types of data — generally data with a large number of features or attributes.

Discussion

This experiment will address the question at hand as it will evaluate the effectiveness of two different types of algorithms. From there, one can make an educated judgement as to whether or not machine learning algorithms can truly assist in identifying the presence of lung cancer.