Cardiovascular Imaging Databases: Building Machine Learning Algorithms for Regenerative Medicine

• Published in: Current stem cell reports Vol. 8; no. 4; pp. 164 - 173
• Main Authors: Mehta, Chirag, Shah, Rohan, Yanamala, Naveena, Sengupta, Partho P.
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.12.2022; Springer Nature B.V
• Subjects: Artificial Tissues (A Atala and J G Hunsberger; Automation; Biomedical and Life Sciences; Biomedical Engineering/Biotechnology; Biomedicine; Cell Biology; Cell therapy; Clinical trials; Gene Therapy; Heart transplantation; Immunology; Learning algorithms; Life Sciences; Machine learning; Patients; Quality control; Radiomics; Regenerative medicine; Section Editors; Segmentation; Stem Cells; Topical Collection on Artificial Tissues; Ventricle; Deep neural networks; Regenerative medicine; Fully convolutional neural networks; U-Net; Machine learning; Convolutional neural networks; Radiomics
• ISSN: 2198-7866; 2198-7866
• DOI: 10.1007/s40778-022-00216-x

Purpose of Review Myocardial regeneration is a promising alternative to heart transplantation, but the ideal stem cell type remains unknown due to conflicting results in clinical trials. Trial discrepancies may be addressed by standardizing cell handling protocols, broadening clinical endpoints, and selecting patients likely to benefit from cell therapy. Machine learning can potentially assist with these tasks. Recent Findings We introduce machine learning and review literature with the most efficacious results translatable to regenerative cardiology, such as in quality control systems during cell culturing, automated segmentation, and myocardial tissue characterization. Investigators are then cautioned on potential pitfalls and offered solutions to minimize model biasing. Summary Standardizing imaging with automated segmentation can improve the quantification of left ventricular endpoints. Additionally, myocardial textural analysis has significant potential to uncover hidden biomarkers, which may address the need for novel clinical endpoints. Lastly, phenogrouping through radiomics signatures can assist in appropriating patients likely to respond to stem cell therapy.