Short Note: Machine Learning in Radiology

Machine learning (ML) lets computers learn patterns from data and make predictions without being explicitly programmed.

Type	Use in radiology
Supervised	Classify tumour benign vs malignant, detect fracture (labelled data)
Unsupervised	Cluster patients or tissue types without labels
Reinforcement	Treatment planning and robotics (emerging)
Deep learning (CNN)	Automatic feature learning for detection and segmentation

Common algorithms: SVM, random forest, k-NN, logistic regression, neural networks.

Applications

Lung nodule and breast cancer detection, bone age estimation, brain tumour segmentation, image quality improvement, workflow prioritisation and radiomics.

Challenges

Need for good labelled data, overfitting, bias, explainability, privacy and regulatory approval.

Exam tip

Draw the train-validate-predict pipeline and name 3 ML types with examples.