@article{shwetar_lalush_zhang_mcanany_jeffrey_haendel_2025, title={A practical introduction to wavelet analysis in electroretinography}, url={https://doi.org/10.1007/s10633-025-10070-x}, DOI={10.1007/s10633-025-10070-x}, abstractNote={CWT and DWT provide complementary and objective insight into ERG responses. Open-source MATLAB toolkit and step-by-step tutorial provided herein lower technical barriers and enable use by the broader community.}, journal={Documenta Ophthalmologica}, author={Shwetar, Yousif J. and Lalush, David S. and Zhang, Alice Y. and McAnany, J. Jason and Jeffrey, Brett G. and Haendel, Melissa A.}, year={2025}, month={Dec} }
 @article{shwetar_brooks_jeffrey_solomon_haendel_2025, title={Advances in machine learning for ABCA4-related retinopathy: segmentation and phenotyping}, DOI={10.1007/s10792-025-03690-4}, abstractNote={ML techniques are well on their way to automate key steps in ABCA4R evaluation with excellent performance. These emerging methods have the potential to expedite therapeutic innovation and enhance our understanding of ABCA4R.}, journal={International Ophthalmology}, author={Shwetar, Yousif J. and Brooks, Brian P. and Jeffrey, Brett G. and Solomon, Benjamin D. and Haendel, Melissa A.}, year={2025}, month={Jul} }
 @article{shwetar_haendel_2025, title={Discrete Wavelet Transform Analysis of PERG Signal Energies for Differentiating Retinal Pathologies}, DOI={10.1109/embc58623.2025.11253207}, abstractNote={Pattern Electroretinography (PERG) can effectively assess both macular cone and retinal ganglion cell activity. However, heterogeneous disease presentations and its low signal-to-noise ratio has limited its application for diagnostic purposes. In this study, we perform the first exploratory analysis of PERG features outside typical time domain markers (P50, N95), and use time-frequency measures from Discrete Wavelet Transform (DWT) in normal controls as well as a collection of inherited retinal diseases (IRDs). For DWT analysis, the Haar wavelet was selected as the mother wavelet due to its ability to fully retain energy profiles of decomposed signals. Biologically relevant frequency bands were retained including detail levels 4-7 (D4-D7) and approximation level 7 (A7). Features were formulated that best characterized energies across each level including Mean Energy, Standard Deviation Energy, and Percentage Energy. A final dataset of 178 recordings from 52 normal controls and 370 recordings from 125 subjects with an IRD were formed respectively. Results revealed distinct energy profiles for each pathology, with Normal subjects exhibiting the highest energy levels and Cone-Rod Dystrophy showing markedly lower values across all features. A7 Energy Percentage yielded a perfect Area Under the Curve (AUC) and distance to top left corner (D) of 1.000 and 0.000 respectively when predicting Normal vs any of the IRDs. Using the A7 Mean Energy feature, we obtained an AUC of 0.703 when discriminating between Cone-Rod Dystrophy and Stargardt Disease. This work establishes the first reported trends in PERG based DWT derived features in some of the most prevalent IRDs. Future research should explore alternative wavelet families and apply these features for improved diagnostic purposes such as classification via machine learning modeling.Clinical Takeaway- Moderate discriminatory success is achieved (AUC 0.703 via ROC Thresholding) using level 7 approximation energy features between Cone-Rod Dystrophy and Stargardt Disease.}, author={Shwetar, Yousif and Haendel, Melissa}, year={2025}, month={Jul} }