Medical Image Analysis is used to process medical images like X-rays(here),CT Scans etc. Here, we will classify whether or not that person suffers from a disease or not. Moreover, it can also be used for plant disease prediction as well.
The dataset used is the Chest X-Ray Images (Pneumonia)(https://www.kaggle.com/paultimothymooney/chest-xray-pneumonia) from Kaggle.
Task : Classification
The 2 class labels are:
1. Pneumonia: Given X-Ray is of a person suffering from Pneumonia.
2. Normal: Given X-Ray is of a normal person.
Various models were trained with the grayscale X-Rays like :
- Binary Logistic Regression.
- K- Nearest Neighbours (KNN).
- Support Vector Machines (SVM).
- Multilayer Perceptron (MLP).
- Neural Network.
Binary Logistic Regression is a Supervised Learning model in which the sigmoid function is used to predict a categorical dependent variable (here, X-Ray type) based on a given set of independent variables.
K-Nearest Neighbours is a Supervised Learing model in which is used for both classification and regression tasks. Here, we use KNN for classifying the type of Lung X-Ray. This algorithm classifies test values based k-nearest train values usually based on Euclidean distance metric.
Support Vector Machines is a Supervised Learning model which can be used for both classification and regression problems. The main idea behind this algorithm is estimate the best possible decision boundary to segreggate a n-dimensional space called hyperplanes. For a new set of independent variables it classifies the dependent variable based on the distance boundary or hyperplane.
Multilayer Perceptron is a Supervised Learning model which is majorly used for classification tasks. It is type feedforward Artificial Neural Network that generates a set of outputs from a set of inputs. Diagrammatically, it is a directed graph with the outermost layers as the input and output layers and the intermediate layers referred to as inner layers. The model is trained using back propagation technique.
The project is documented as a medium article.
To apply much better image processing techniques and training deep learning models.
Gaurav Pendharkar https://github.com/Gaurav0502.
Prasham Titiya https://github.com/prasham1515.