This repository contains the final coursework and supporting code for the Digital Biomarkers course, covering both 1D signal processing (e.g. ECG, respiration, neural signals) and 2D medical imaging (breast ultrasound classification).
final_1d.pdf: Solutions and annotated analysis for time-series signal processing tasks.- ECG & respiration correlation
- Moving average and IIR filters
- Spike detection in neural data
final_2d.pdf: Feature engineering and classification workflow for breast ultrasound imaging.- Shape and texture feature extraction
- Classification using logistic regression, SVM, random forest, and MLP
- R-peak detection and RR interval analysis (ECG)
- Respiratory sinus arrhythmia correlations
- Signal filtering (FIR, IIR, Notch)
- Neural spike detection via likelihood ratio thresholding
- Image preprocessing and mask alignment
- Shape-based biomarkers (area, perimeter, circularity, regularity)
- Texture analysis using GLCM (contrast, entropy, homogeneity)
- Machine learning classifiers for tumor type prediction
- Python (NumPy, SciPy, Matplotlib, scikit-learn, scikit-image)
- TPOT (automated ML)
- Signal processing techniques (filtfilt, lfilter, MA filter)
- Classical ML models (SVM, Random Forest, MLP)
This work was submitted as the final assessment for the Digital Biomarkers course (Spring 2025). It integrates signal processing, image analysis, and machine learning to explore how physiological signals and imaging can be transformed into meaningful digital biomarkers for clinical decision-making.
Sarah Meyer
Veterinarian & MSc Medical Informatics
GitHub | LinkedIn