A Machine Learning Approach To Improving Diagnostics and Care For Heart Disease Patients 

A Summer With Stanford's Cardiovascular Institute

by Gabrielle Montenegro

Background 

The mammalian heart depends on tightly controlled contraction and relaxation cycles, which are coordinated from the molecular to organ level. Cardiomyocytes, or cardiac muscle cells, produce the contracting force of the heart. If you were to isolate live cardiomyocytes in a petri dish and view them under a microscope, you would see the cells "beat" as one. Sarcomeres within cardiomyocytes are important to this "beating" function. 





Image Credit: Tharp et al., 2020

Clip Credit: Elara Systems, 2016

Heart Disease Is A Health Disparity

Heart disease is the most prevalent disease across the human race and is unfortunately one of the leading health disparities in the United States. In fact, it is the leading morbidity and mortality worldwide (Javed et al., 2022). Heart disease entails a broad range of conditions of the heart, including infarction, arrhythmia, ectopy, hypertrophy, high blood pressure, congenital heart disease, and more. The CDC reports that cardiovascular diseases alone are responsible for 17.9 million mortalities annually (Cardiovascular Diseases, 2023). Mortality is significantly higher across low income, underrepresented populations for whom access to heart healthy foods, environmentally safe living conditions, quality healthcare, and clean air can be difficult to come by. 


Unfortunately, the color of one’s skin determines risk for heart disease (Javed et al., 2022). For many, the risk factors for heart disease are silent. My own experience has shown me that living in a low income, poorly maintained neighborhood is linked to chronic medical conditions and these, in turn, are connected to why someone who looks like me is more susceptible to heart disease. The low income family living in a Section 8 neighborhood does not consider the environmental toxins they may be exposed to by the industrial warehouse a few blocks down or the toxins in the paint coating the walls of their home or the fact that the nearest grocery store with fresh produce is miles away. These factors pose risk for heart disease. The risk for mortality is even higher when the nearest medical center is miles away. The low income family rather sees their Section 8 housing as a safe place to call home – a place of security rather than a root in health disparities like heart disease. Along with dismantling a historical societal structure of racism and discrimination towards minority populations, it is imperative that science and the healthcare system push research forward to improve early diagnosis and effective treatment of heart disease in order to close the gap in morbidity and mortality from heart disease globally. 


My Research In the Chiu Lab & Stanford MAvERICS

During my summer with Stanford's Cardiovascular Institute, I attended various workshops by Stanford Medicine faculty to educate myself on the fundamentals of the biological structure of the heart and inequities across cardiovascular research. As part of Stanford MaVERICS, I had the pleasure of participating in a meta-research study on preclinical cardiovascular research within the last decade. I helped screened hundreds of preclinical cardiovascular publications for methodological rigor and inclusion. I learned that many preclinical cardiovascular studies are often done using mammals of the male sex, and studies using human participants often include cohorts of primarily white males– a non-representative model . This in itself is a cause for concern in our aim to improve cardiovascular care and close the disparity in heart disease. I am currently in the works of drafting a manuscript to report to the scientific community and the general public on preclinical cardiovascular study rigor. 

In addition to meta research, I also had the honor of participating in structural biology research in the Chiu Lab. I worked under the mentorship of Dr. Rahel Woldeyes to investigate cardiomyocyte sarcomere using cryo-electron tomography and machine learning. Using rat and human iPSC cardiomyocytes, I used a supervised deep learning algorithm to train a neural network to recognize and accurately label sarcomeric proteins within tomograms. This involved annotating various human and rat iPSC cardiomyocytes, and Bash & Python scripting, to provide the network information on the biological structure of cardiomyocytes and their intracellular components. We found that optimization of parameters and an increased number of manual annotations resulted in better labeling performance by convolutional neural networks. Using these approaches, we expect to discover structural differences between healthy and diseased cardiomyocytes, which can ultimately improve our understanding of the heart and improve patient care. 





Thick & Thin Filament

Plots Provide Information on Training Performance

I-Band Of Sarcomere

Membrane

Mitochondria

The Knowledge I'll Take To Medical School

Before my experience with Stanford's Cardiovascular Institute, I had not given much thought to what could be further done in science and healthcare to help address the disproportionate effects of heart disease on underrepresented minority populations. Given its high global morbidity and mortality, I imagined heart disease and cardiovascular research would be well defined and advanced enough to consider the effects of the social determinants of health on population health and patient care. There is still work to be done to improve diversity, equity, and inclusion in cardiovascular research and care. 

With the advancement of technology, machine learning has the ability to contribute to closing the gap in mortality from heart disease between racial and ethnic groups. The color of one’s skin should not determine their risk for heart disease. By furthering the development of computational systems for phenotypic scaling and biological imaging, the problem of late diagnosis of heart disease can be mitigated and outcomes for cardiovascular care can be improved. The skills and knowledge I learned at Stanford's Cardiovascular Institute have better prepared me to enter the healthcare field where, as a future physician of color, I will bring my expertise to my community. I aspire to pursue a career in medicine to be a physician of color who considers lifestyle factors and medicine in patient care– a physician my grandmother feels comfortable receiving care from.

References

Gabrielle Montenegro

Hi! My name is Gabrielle Montenegro and I am from Berkeley, California. My interest in medicine and public health has led me to major in Biology (Biochemistry) with a minor in Chemistry and concentration in Community & Global Health. My passion also fueled my engagement in volunteer community health roles at Regions Hospital, Wellstone Elementary School, Hidden River Middle School, and ESTHER Homes MN. At Macalester, I also play violin as part of the Symphony Orchestra! Post-graduation, I plan to further immerse myself in cardiology research at UCSF, all while applying to medical school to become a future physician.