CIRA_ML

CIRA (Chromosomal Image Recognition Algorithm) Image Processing and ML

Uses Machine Learning to label cells as unhealthy or healthy.

Runs on red cell images.
Example Data:

More examples and results can be viewed in the Image Results folder.

Machine Learning Algorithms

Clustering Algorithms

The clustering algorithms were used as classifying algorithms by clustering the data with varying values for K then labeling each cluster with the class that occured the most. When a new cell needed to be classified it was placed in the closest cluster and given the same label as that cluster. Accuracy was calculated using this method and the original class labels.

K-Means

Accuracy: 0.791

Image 1. K-Means K=2 Graph: This shows the best clustering result among the 3 chosen methods K-Means, with K=2. The left side would be the unhealthy cluster, and the right side would be the healthy cluster.

Agglomerative Clustering and Ward

Accuracy: 0.610

Agglomerative Clustering with Complete Link

Accuracy: 0.712

Classification Algorithms

Support Vector Machine

Accuracy 0.4975

Logistic Regression

Accuracy: 0.87

Multi-Level Perceptron

Accuracy: 0.86025