Deep learning accelerates assessment of cardiac aging and disease in Drosophila as a model for human disease – News

Machine Learning

Using high-speed video microscopy and artificial intelligence, statistics such as diastolic and systolic diameters, cardiac output, fractional shortening, and ejection fraction are calculated.

Mountain stream fruit flyUsing high-speed video microscopy and artificial intelligence, calculated statistics such as diastolic and systolic diameters, cardiac output, fractional shortening, and ejection fraction are obtained.Drosophila melanogaster (commonly known as the fruit fly) is a valuable model for human cardiac pathophysiology, including cardiac ageing and cardiomyopathies. However, a challenge in assessing the fruit fly heart is that human intervention is required to measure the moment when the heart maximally expands or contracts. This measurement allows the calculation of cardiac dynamics.

Researchers at the University of Alabama at Birmingham have now shown how using deep learning and high-speed video microscopy to study every beat of a fly's heart, they can drastically reduce the time needed for analysis while utilizing more of the heart area.

“Our machine learning method is not just faster; it minimizes human error by eliminating the need to manually mark each heart wall during systole and diastole,” said Girish Melkani, PhD, associate professor in the Department of Molecular and Cytopathology in the UAB School of Pathology. “Furthermore, we can run the analysis of hundreds of hearts and view the results once all the hearts have been analyzed.”

This will expand our ability to test how different environmental and genetic factors affect cardiac aging and pathology. Merkani envisions using deep learning-assisted research to investigate cardiac mutation models as well as other small animal models such as zebrafish and mice. “Furthermore, our technique can be adapted to human heart models, providing valuable insights into cardiac health and disease. By incorporating uncertainty quantification methods, we can further increase the confidence of our analysis. Moreover, our machine learning approach can predict cardiac aging with a high degree of accuracy.”

Mercani says that Drosophila models are already very powerful in understanding the pathophysiological basis of several human cardiovascular diseases, which remain among the leading causes of death and disability in the United States.

Mercani and his UAB colleagues evaluated the trained models on cardiac performance in both Drosophila cardiac aging and in a Drosophila model of dilated cardiomyopathy caused by knockdown of oxoglutarate dehydrogenase, a key enzyme in the TCA cycle. These automated assessments were then validated against existing experimental datasets. For example, for aging at 1 and 5 weeks of age, roughly half the fruit fly's lifespan, the UAB team used 54 hearts for model training and then validated the measurements against an experimental aging model containing 177 hearts. The trained models were able to reconstruct expected trends in cardiac parameters with aging.

RS50056 Girish Melkani 2 scr 1Dr. Girish MelkaniMercani says the team's model can be applied to readily available consumer hardware, and their code can provide calculated statistics such as diastolic and systolic diameters/intervals, shortening fraction, ejection fraction, cardiac cycle/rate, and quantified cardiac arrhythmias.

“To the best of our knowledge, this innovative platform for deep learning-based segmentation is the first to be applied to standard high-resolution high-speed light microscopy of the Drosophila heart and to quantify all relevant parameters,” said Mercani.

“Automating the process and providing detailed cardiac statistics paves the way for more accurate, efficient and comprehensive studies of Drosophila cardiac function. This method has great potential not only for understanding aging and disease in Drosophila, but also for applying these insights to human cardiovascular research.”

The first authors of the study, “Automated assessment of cardiac dynamics in Drosophila models of aging and dilated cardiomyopathy using machine learning,” published in Communications Biology, are Yash Melkani and Aniket Pant of the UAB Department of Pathology. Yiming Guo, also of the UAB Department of Pathology, is an additional author, and Girish Melkani is the corresponding author.

Support was provided by National Institutes of Health grant AG065992, a UAB Marnix E. Heersink School of Medicine AMC21 grant, and UAB Pathology startup funds.

Girish Melkani develops and uses clinically relevant Drosophila models to study the pathophysiological basis of human circadian/metabolic disorders related to cardiometabolic diseases, myofibrillar myopathies, proteinopathies, neurological disorders, sleep disorders, and aging disorders, as well as how lifestyle and genetic factors act to maintain the structural integrity of cells, tissues, and organs, which in turn dictate the physiology of the organism.

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *