Closing the “Last Mile Gap” in Access to Multimodality Imaging in Rural Settings: Design of The Imaging Core of The Risk Underlying Rural Areas Longitudinal Study

Abstract

Achieving optimal cardiovascular health in rural populations can be challenging for several reasons including a decreased access to care with limited availability of imaging modalities, specialist physicians, and other important healthcare team members. Therefore, innovative solutions are needed to optimize healthcare and address cardiovascular health disparities in rural areas. Mobile examination units (MEUs) can bring imaging technology to underserved or remote communities with limited access to healthcare services. MEUs can be equipped with a wide array of assessment tools and multiple imaging modalities such as computed tomography (CT) scanning and echocardiography. The detailed structural assessment of cardiovascular and lung pathology, as well as the detection of extra-cardiac pathology afforded by CT imaging combined with the functional and hemodynamic assessments acquired by echocardiography, yield a deep phenotyping of heart and lung disease for populations historically underrepresented in epidemiological studies. Moreover, by bringing the MEU to local communities, innovative approaches are now possible including engagement with local professionals to perform these imaging assessments, thereby augmenting local expertise and experience. However, several challenges exist before MEU-based examinations can be effectively integrated into the rural healthcare setting including standardizing acquisition protocols, maintaining consistent image quality and addressing ethical and privacy considerations. Herein, we discuss the potential importance of cardiac multimodality imaging to improve cardiovascular health in rural regions, outline the emerging experience in this field, highlight important current challenges, and offer solutions based on our experience in the Risk Underlying Rural Areas Longitudinal (RURAL) cohort study.

Introduction

Cardiovascular disease (CVD) continues to be the leading cause of morbidity and mortality in the United States (US) and globally, accounting for every 1 in 4 deaths.^1-2 Alarmingly, several studies have suggested that residents of rural communities in the US have experienced the least decline in cardiovascular mortality over the last two decades relative to their urban counterparts.³ This rural-urban disparity in CVD burden continues to widen and is associated with a 20% greater mortality in rural areas than in metropolitan areas – approximating 60,000 excess deaths annually.^4-5 Importantly, this excess burden of CVD mortality in rural areas has persisted across all age, sex and race/ethnicity groups, and type of CVDs.⁶

The rural-urban health disparities are known to be multifactorial, including limited access to healthcare, insufficient healthcare resources, and paucity of medical providers and support staff. Restricted access to high-quality healthcare is a major obstacle to timely CVD diagnosis and treatment.^7-8 About 60 million individuals, or 20% of the total US population, live in rural areas, while only 9% of US physicians practice in rural areas.⁹ In the coming years, this shortage of rural physicians, including cardiologists, is expected to worsen with the aging of the US population and increasing demand for healthcare services. Over 100 rural hospitals have closed since 2010, 43 of which have closed since 2020.^10-13 The majority of rural hospitals lack specialist physician coverage, care team members, and vital cardiac imaging modalities to aid decision-making for the diagnosis and management of most CVD, particularly heart failure, valvular heart disease, and coronary artery disease.^14-15

Bridging these rural-urban healthcare access gaps is a public health priority that can be partly addressed by enhancing access and proximity to cardiovascular imaging and specialists. Deploying multimodality cardiac and lung imaging in a mobile examination unit (MEU) that can travel between rural communities to increase access to image acquisition may be an innovative solution to address cardiovascular health disparities in rural areas. Herein, we describe the approach of a population-based epidemiological cohort study that has implemented a novel strategy to improving access to multimodality cardiac imaging in select counties in the rural South.

Synopsis of the RURAL Cohort Study

The Risk Underlying Rural Areas Longitudinal (RURAL) Cohort Study is an ongoing, NHLBI-funded research project that aims to identify the prevalence of risk factors for heart and lung diseases in select rural counties in four states, i.e., Alabama, Kentucky, Louisiana, and Mississippi (https://theruralstudy.org). Six rural counties within these states have been chosen to represent higher age adjusted CVD mortality rates than the rest of the country (top 10% highest age-adjusted mortality rates among the 50% poorest counties in the countries), while four counties that experience lower CVD mortality rates (these counties were chosen from the 33% counties with the lowest age-adjusted mortality rates of the 50% poorest counties) were selected to represent a comparison group.^16-18 These ten counties were matched (within state) for their degree of poverty, the proportion of marginalized people of color, and population size, in an ecologically-paired design. The RURAL Study will identify the interactions between psychosocial, economic, familial, and behavioral and standard risk factors and how these contribute to the development of health disorders including CVD and lung disease. The goal of the RURAL Study’s Imaging Core is to reach people living in the target rural counties and assess the prevalence, incidence and correlates of early and subclinical heart failure, and coronary atherosclerosis using state-of-the-art cardiac CT and echocardiography scanning in an MEU.

Unmet Needs: Imaging and Detecting Subclinical Coronary Atherosclerosis and Heart Failure in the Rural South

The data on subclinical CVD in rural populations in the Southern U.S. is sparse. The presence of subclinical coronary artery calcium (CAC) and other preclinical atherosclerotic markers is a strong predictor of future risk of overt coronary heart disease compared to traditional risk factors. The ability to identify and target significant subclinical disease is critically important for CVD prevention, given that CVD is the leading cause of mortality and morbidity in the US with an estimated $900 billion annual healthcare costs by 2030.^{1, 19-21} The CAC score has been shown to enhance CVD risk classification in intermediate-risk individuals in several population-based studies, including the Multi-Ethnic Study of Atherosclerosis (MESA), Atherosclerosis Risk in Communities (ARIC), Jackson Heart Study, the Rotterdam Study, the Heinz Nixdorf Recall Study, and many others. It has been shown previously that the presence of more traditional risk factors in rural central Appalachia correlated with higher elevated CAC scores.^22-23 Additionally, there are racial and ethnic differences in CAC score, notably a lower prevalence of coronary calcification, yet higher rates of CVD events in Black individuals compared to their White counterparts, which may inform CVD risk assessment and risk stratification.

The interplay between social determinants of health (SDoH) and genetic susceptibility to CVD may be associated with increasing CVD burden (Figure 1). It is possible that such genetic susceptibility for CVD also exists in the RURAL cohort, which samples rural individuals across the socioeconomic strata.²⁴ Further, the evaluation of these non-traditional risk factors (SDoH and genetic variation) and their relations with other subclinical CVD markers such as CAC score, has not been well studied in the rural south in an adequately powered large multi-ethnic cohort.

Figure 1. Genes, the Environment and Gene-Environment Interactions Along with Exposures and Social Determinants Accumulate Over the Life-Course.

Compared to those living in the urban areas, rural residents tend to suffer from a greater clustering of these risk factors across their life-course (allostatic load) resulting in a greater risk for heart, lung, blood and sleep disorders. In the RURAL Study during the mobile examination unit a self-administered social determinants of health survey includes questions regarding healthcare utilization, financial insecurity, food insecurity, housing stability, childhood and adulthood stressors, depression, resilience, residential history, income, implicit/explicit bias, systemic racism and occupation.

Pervasive biological and social stressors (unfavorable SDoH) pose risks to the cardiovascular system and can result in maladaptive changes. Echocardiographic left ventricular hypertrophy (LVH), increased LV mass index (LVMI), regional wall motion abnormalities, LV enlargement or LV systolic dysfunction often occur before the onset of clinical symptoms; the presence of any of these represents Stage B heart failure (HF).²⁵ Stage B HF is a high-risk phenotype for incident HF; it carries a nearly 5-fold increased risk of overt HF compared to having normal LV function, and confers a 1.5-2-fold risk for overall CV morbidity and mortality.^26-27 Cohort studies are optimal for studying the prevalence and determinants of Stage B HF in the community since the participants are generally asymptomatic, mitigating the referral bias inherent in hospital-based samples. A large study of individuals sampled from community health centers across 12 southern states has revealed there is an increased hazard of HF in rural areas relative to urban regions. Compared to their urban counterparts, southern rural residents had a 19% greater risk of incident HF.²⁸ This study also reported important race- and sex-related differences in HF risk; the greatest rurality-associated risk of incident HF was in Black men (34%), White women (22%), and Black women (18%) while White men had a 3% lower risk than the overall group. RURAL’s approach to incorporating echocardiography into the baseline examination provides a paradigm-shifting approach to assess the burden of subclinical and clinical cardiac dysfunction in select rural communities that have hitherto had very limited access to advanced cardiac imaging technology that is vital for CVD prevention.^{3,28, 29}

The advanced multi-modality imaging approach of RURAL goes beyond traditional risk factor screening, providing a nuanced understanding of the multifaceted factors contributing to health inequalities. By leveraging cutting-edge technology and a range of scans, the RURAL study can effectively discern and address the complex interactions shaping health outcomes in rural communities, surpassing the limitations of conventional public health approaches.

The RURAL Mobile Examination Unit (MEU)

RURAL utilizes an MEU within a semitrailer consisting of a laboratory, private examination rooms, and an imaging center including a CT scanner and echocardiography (Figure 2). This innovative approach allows for easy transport into rural communities that promotes engagement with local participants. The MEU is temperature-controlled, wheelchair accessible, offers personal privacy, and is equipped with an independent power supply, making the unit easily customizable for multi-purpose use. The MEU obviates the need for separate free-standing field centers in each state and offers standardized imaging at each location, thereby mitigating imaging equipment-related variability, improving overall image quality, while also being cost-effective. The MEU is (re)deployed, set up, and tested for operability at predetermined accessible locations in the target counties.

Figure 2. Schematic of the RURAL MEU.

A) The RURAL MEU shown in schematic identifying the imaging center which houses both the CT scanner and echocardiography. Numerous rooms are available on the MEU for laboratory space, interviewing participants and a comfortable reception area; B) A view of the MEU from the outside.

Comprehensive Assessment of Lung Diseases and Obesity in RURAL and Simultaneous Coronary Artery Calcium Scanning: A Single-Session Approach

Rural areas exhibit poor health outcomes not only in terms of CVD mortality but also in other health domains that impact overall life expectancy. For instance, the incidence of lung cancer in non-Hispanic white women residing in non-metropolitan regions continues to have a negative impact on their life expectancy, and so do other respiratory diseases relative to their male counterparts. Higher smoking rates are observed in rural communities in the US relative to urban regions.^30-31 The age-adjusted prevalence and mortality rates of chronic obstructive pulmonary disease (COPD) exhibit a positive correlation with the degree of rurality, with the highest Medicare hospitalization rates for COPD observed among individuals residing in more rural areas.³² The prevalence of obesity in non-metropolitan areas of Mississippi (40%), Louisiana (40%), Alabama (38.3%), and Kentucky (38%) are among the highest reported in the nation in 2022.

The RURAL study includes a single CT imaging session for evaluation of cardiovascular and respiratory diseases. As visceral adipose tissue is a good indicator of future cardiac risk, and mid-thigh muscle mass on CT is predictor of insulin resistance, diabetes, cardiovascular disease, our study implemented addition of single slice abdominal and mid-thigh CT to the coronary and lung scans to characterize these body fat depots (such as non-alcoholic fatty liver disease) and assess lean muscle mass (Table 1).

Table 1.

Cardiovascular Imaging in the RURAL Study

Modality	CT	AI-Echo
Scanning protocols	CAC	LV systolic and diastolic function
	Lung scan	GLS in multiple views
	Abdominal and thigh scan	Chamber sizes
		Valvular function
		Other cardiac and extra-cardiac structural and hemodynamic abnormalities
Return of all results to participants	Yes	Yes
Critical alerts notification	Yes	Yes

^*CT = computed tomography; AI = artificial intelligence; GLS= global longitudinal strain; CAC=coronary artery Calcium

CAC Scan and Image Acquisition

All RURAL Study participants undergo a gated, non-contrast chest CT using a dual source 256-slice MDCT scanner (GE, CardioGraphe^™)³³ which is a one-beat heart scanner known for its swift gantry rotation of 0.24 sec providing a high temporal resolution (120 msec). At the heart of its appeal is a small footprint (179 cm in width/ 278 cm in length), requiring a mere 4.4 m x 3.4 m room, that positions it as an ideal candidate for integration into a MEU compared to other conventional CT scanners. Due to its expansive field of view from a focused field of view of 25 cm for cardiovascular imaging to 45 cm for other organs, it emerges not only as a dedicated cardiovascular imaging for the RURAL MEU but a versatile workhorse, capable of seamlessly transitioning between cardiac and extracardiac imaging realms (for lung, abdomen and thigh scans), answering diverse needs of the RURAL study.^34-35 This adaptability aligns perfectly with the holistic approach of the RURAL Study. Summary of the coronary calcium scan, lung, abdomen and thigh scan protocols presented in Table 2. Certified technicians and staff sequentially travel through chosen counties, maintaining consistency and cost-effectiveness by avoiding separate field centers. Exam schedules are designed to be flexible, allowing participants to choose staggered start times based on preferences. Each exam lasts 3-5 hours with 4-8 participants a day (40/week). The MEU will travel across ten counties over the study period, stationed in each county for 16-32 weeks based on the county.

Table 2.

Heart, Lung, Abdomen and Thigh Scan Protocols

Parameter/protocol	Coronary Calcium scan	Lung scan	Abdomen	Thigh
Coverage	Cephalad: pulmonary artery bifurcation Caudad: Beyond apex into upper abdomen	Entire lung	L4-L5 interspace	Midpoint of Right femur (if right leg is not suitable then left leg will be used)
Slice thickness	2.5 mm	1.5 −2.5 mm	10 mm (single slice)	10 mm (single slice)
Tube current	320 mA (weight< 220lb) / 400 mA (weight > 220 lbs.)	200 mAs (inspiratory scan) / 50 mA (expiratory scan)	170 mA	170 mA
Tube Voltage	120 Kv	120 Kv	120 Kv	120 Kv
Rotation time	350 msec	1 sec	1 sec	1 sec
Collimation	16x2.5 mm	Lowest	Lowest	Lowest
FOV	32-35 cm (to include calibration phantom)	Fit to Lung	Fit to entire slice	Fit to entire slice
EKG gated	Yes (Triggers at 75% heart cycle)	No	No	No

^*FOV, field of view; EKG, electrocardiogram

Comprehensive Assessment of Cardiovascular Disease in RURAL: An Artificial Intelligence-based Echocardiography (AI Echo) Approach

The integration of AI in echocardiography encompasses two broad domains: image acquisition and image interpretation. AI-assisted image acquisition improves protocol selection, provides real-time guidance, and controls image quality. Automated analysis, anomaly identification, decision assistance, and process optimization are all features of AI-driven image interpretation. In AI-enhanced echocardiography, automated recognition sequences and algorithms perform roles traditionally held by sonographers in optimizing and acquiring cardiac images. The use of AI can potentially improve image quality, reduce scanning time, eliminate artifacts, minimize variability, and enhance diagnostic accuracy (Table 3). AI algorithms, therefore, assist healthcare providers in making accurate and timely diagnoses, regardless of their level of expertise (Figure 3).

Table 3.

Summary of studies highlighting the role of artificial intelligence (AI) enhanced echocardiography.

Author	Aim of Study	Sample Size	Method of AI	Results
Asch et al. 36	To evaluate AI algorithms for the detection of endocardial boundaries and measurement of LV volumes and function	99	ML algorithm	Auto-EF values showed high consistency (mean absolute deviation =2.9%) and excellent agreement with the reference values: r=0.95, bias=1.0%, with sensitivity 0.90 and specificity 0.92 for detection of EF ≤35%
Narang et al. 37	To test whether novice users could obtain 10-view TTE studies of diagnostic quality	240	DL algorithm	> 90% of diagnostic quality for key parameters such as LV size and function, RV size and function, and pericardial effusion
Schneider et al. 38	To test the algorithm by having 19 first-year medical students without previous ultrasound knowledge scan patients	57	ML model	Successfully acquired at least one of the three required views 91% of the time
Cheema et al. 39	To describe the use of AI technology that guides novice users to acquire high-quality cardiac ultrasound images	5	DL algorithm	Accurately interpreted ventricular dysfunctions and helped in guiding medical therapies
Leclerc et al. 40	To evaluate the DL models in assessing 2D echocardiographic images, i.e., segmenting cardiac structures and estimating clinical indices	500	DL algorithm	It provided more consistent data with less variability
Ouyang et al. 41	To evaluate a video-based DL algorithm for accurate assessment of cardiac function in echocardiogram videos	55	DL algorithm	Accurately and rapidly calculated the EF (absolute error 4.1%)
Jafari et al. 42	To evaluate computationally efficient mobile application with POCUS for accurate LVEF estimation	427	DL algorithm	Accurately calculated LVEF with a median absolute error of 6.2% compared to an expert cardiologist assessment
Asch et al 43	To test the accuracy of AI algorithms LV volume and function based on POCUS exams	166	ML algorithm	The agreement with the reference EF values was good (intraclass correlation, 0.86–0.95), with biases <2%. ML classification of LV function showed similar accuracy to that by physicians in most views.
Tromp et al 44	To compare a DL interpretation of 23 echocardiographic parameter with three repeated measurements by core lab sonographers	600	DL algorithm	The point estimates of individual equivalence coefficient were all <0 and the upper bound of the 95% confidence intervals below 0.25, indicating that the disagreement between the DL and human measures was lower than the disagreement among three core lab readers.
Akerman et al 45	To analyze a single apical 4-chamber transthoracic echocardiogram video clip to detect HFpEF.	6756	ML algorithm	Excellent discrimination (area under receiver-operating characteristic curve: 0.97 [95% CI: 0.96-0.97] and 0.95 [95% CI: 0.93-0.96] in training and validation, respectively)
Kelsey et al. 46	To evaluate the role of AI echo in assessing cardiac structure and function among rural populations	138	DL algorithm	Adequate image quality for visual EF determination in 97%, with LV dimensions measurable in 88% and LA diameter in 91%

AI, artificial intelligence; DL, deep learning; ML, machine learning; LA, left atrium; LV, left ventricle; RV, right ventricle; EF, ejection fraction; POCUS, point-of-care ultrasound; HFpEF, heart failure with preserved ejection fraction and TTE, transthoracic echocardiography

Figure 3: Current Challenges in Rural Areas and Possible Solutions with Artificial Intelligence Enhanced Echocardiography.

AI, Artificial Intelligence

AI Echo for Image Acquisition

The feasibility of deploying AI Echo in the RURAL MEU was rigorously tested using a prespecified checklist in partnership with the Duke Clinical Research Institute’s (DCRI) Imaging Core Laboratory (ICL) and industry partner Caption Health, which has now been acquired by GE Healthcare. The RURAL MEU is well suited for deploying AI Echo; it was specifically designed to enhance equipment ergonomics and workflow by optimizing the physical space available to the imaging unit (electrical outlets, lighting, and privacy, etc.), confirming the absence of electrical and sound interference, added equipment, confirming the cleaning and personal protective equipment (PPE) protocol and successfully testing image transfer to a remote server at the DCRI. Feasibility testing demonstrated high participant and scanner comfort across a range of body sizes.

Our approach to AI Echo in RURAL leverages the first FDA-authorized software that incorporates interconnected deep learning (DL) algorithms to provide real-time guidance and instructions to untrained operators in acquiring transthoracic echocardiographic (TTE) images.^47-48 The software relies solely on ultrasound images and does not require additional trackers or sensors. RURAL technicians in the MEU take advantage of AI-powered, real-time feedback and guidance during image acquisition to ensure optimal transducer positioning, probe angle, and equipment settings adjustments, resulting in more explicit and diagnostically valuable cardiac images. AI algorithms also improve quality control by analyzing live images and identifying potential issues or artifacts, enabling immediate adjustments to improve image quality.⁴⁹ By enhancing image acquisition, AI Echo improves overall diagnostic accuracy and efficiency, ensuring standardized and reliable image acquisition processes.

In a study by Narang et al., a DL algorithm was utilized to direct eight nurses with no previous ultrasonography experience in acquiring echocardiograms of 240 participants in two centers.³⁷ Expert echocardiographers, who blindly reviewed the scans, found scans to be >90% of diagnostic quality for key parameters such as LV size and function, right ventricular (RV) size and function, and pericardial effusion. The diagnostic accuracy for these parameters did not differ by body size from conventional scans performed on the patients by an expert sonographer. In this study, abnormal LV function was common (73%) but it did not impact image quality. The training program's success was highlighted by the fact that there was no discernible difference in the nurses' ability to obtain diagnostic scans over the course of the study period. These findings are supported by several other examples of untrained scanners providing high quality, interpretable images.^38-39

AI Echo for Image Interpretation

Machine learning (ML) techniques have enhanced automated image analysis, improving accuracy and speed. LVEF, a commonly used clinical measure, is usually determined by measuring the ratio of the LV end-diastolic to end-systolic volumes or using the Simpson biplane method. Several studies have shown how the use of AI Echo can minimize the known limitations of these approaches.^{40-42, 44,47,50,51}

An encoder-decoder-based convoluted neural network (CNN) DL software was employed by Leclerc et al. to analyze 500 echocardiogram images to measure LVEF using apical four and two-chamber views.⁴⁰ The software outperformed non-DL methods and was more consistent than the observed inter-observer variability. Ouyang et al. used AI to develop a video-based DL algorithm to rapidly and accurately assess LVEF using only apical four-chamber views and in one cardiac cycle. The CNN model, trained by using 10,030 apical four-chamber echocardiogram videos (including those with poor image quality), was able to accurately calculate the EF (absolute error 4.1%) and proved to be more rapid than measurements made by humans.⁴¹ Similarly, He et al. recently showed in a blinded randomized trial of 3,769 echocardiographic studies that the AI-guided workflow for cardiac function assessment saved time for both sonographers and cardiologists. Moreover, cardiologists were not able to distinguish between the initial assessments by AI versus the sonographer (blinding index of 0.088).⁵²

Jafari et al. developed mobile point-of-care ultrasound (POCUS) software to assess LVEF.⁴² This system employs DL to recognize, segment, and analyze LVEF using apical 4 and 2 chamber views. When applied to a sample of 427 patients, the resulting LVEF had a median absolute error of 6.2% compared to an expert cardiologist assessment. Asch et al. developed an ML algorithm that used CNN to calculate LVEF.³⁶ The system demonstrated that AI-powered Auto-EF was reliable (mean absolute deviation =2.9%), and the accuracy was comparable to conventional LV volume-based measurements (r=0.95, bias=1.0%). This automated ML algorithm was also applied specifically in POCUS exams showing excellent agreement with the reference EF values (intra-class correlation 0.86-0.95) with biases <2%.⁴³ Tromp et al. have more recently developed a fully automated DL-based workflow to automate the view classification, annotation, and interpretation of cardiac volumes, LVEF, and E/e' ratio for the interpretation of systolic and diastolic function.⁵³ This was also tested in a study among 602 echocardiographic studies from 600 people which showed that the disagreement between DL based workflow and human measures was lower than the disagreement seen among three core lab readers.⁴⁴ More recently, Akerman et al developed a three dimensional CNN to analyze a single apical 4 chamber view to detect heart failure with preserved ejection fraction (HFpEF).⁴⁵ Training and validation included 2,971 cases and 3,785 controls, and demonstrated excellent discrimination (area under receiver-operating characteristic curve: 0.97 [95% CI: 0.96-0.97] and 0.95 [95% CI: 0.93-0.96] in training and validation, respectively). This AI model based on a single routinely acquired echocardiographic video more often had better discrimination to detect HFpEF than HFpEF clinical scores.

In the RURAL study, we trained hitherto six technicians (non-sonographers) to acquire high quality standard images. In a preliminary analysis of 138 AI-analyzed echocardiograms, there was adequate image quality for visual EF assessment in 97%, with LV dimensions being measurable in 88% and left atrial (LA) diameter in 91%.⁴⁶ For most participants (96%), the Auto-EF corresponded with visual estimation. These results highlight the promising utility of AI Echo in limited healthcare and low-resource environments with high degrees of obesity (70% with BMI ≥ 30 kg/m²).

The RURAL study is also collecting other echocardiography measures that presage the development of CVD (Table 1). Global longitudinal strain (GLS), a metric that reflects the longitudinal strain of all myocardial segments in the apical views, is a robust marker of systolic function that has been extensively validated as a marker of early LV dysfunction and to guide therapy for HF.^54-56 Other echo measures that have been linked to poor prognosis in asymptomatic individuals include LA size/function/strain, cardiac remodeling, and LV diastolic function.^57-58 The inclusion of these measures in the RURAL study echocardiography protocol will allow us to detect HF risk even earlier along the disease progression spectrum than Stage B HF.

Two-dimensional echo images will be obtained in the parasternal long-axis and short-axis views, as well as apical 4-, 2- and 3-chamber views. Color Doppler will be acquired for evaluation of valvular regurgitation. GLS will be performed on apical views in the DCRI core lab and chamber dimensions will be measured according to American Society of Echocardiography (ASE) guidelines in the parasternal long and apical views. All acquired images will be reviewed for completeness and quality by a research sonographer. Once echoes are confirmed to be complete and query-free, then the studies will be processed and interpreted. All images are first read by a highly-experienced, research sonographer who performs all initial measurements and assessments. AI-calculated ejection fraction, manually calculated ejection fraction and the remainder of the core lab’s measurements and findings are recorded during the initial read by the sonographer – the only AI generated measurement is ejection fraction. An experienced cardiologist (level 3 training, echo board certified) then over-reads all quantitative and qualitative data. Final results are entered into a customized case report form. Prior to sending data for statistical analysis, all studies undergo a query and cleaning process to ensure a high-quality final imaging dataset.

Barriers and Challenges

The recruitment of research participants in rural areas presents a formidable challenge due to factors such as geographical remoteness, distinctive cultural and social attributes, and skepticism towards the intentions of scholarly investigators.^59-62 The implementation of the MEU, a clinic on wheel approach, is anticipated to streamline and simplify participation processes, minimizing associated inconveniences and complexities given the broad range of its imaging capabilities in the RURAL study. While significant progress has been achieved in optimizing AI for diagnostic echocardiography, the operation of the machine and optimal probe position remains a concern. Approaches, whereby trained sonographers remotely operate robotic echocardiography probes, are being explored.⁶³ Moreover, echocardiography can be affected by several patient-specific variables such as obesity. Hence, quality control and enhancement techniques play a crucial role in the clinical workflow, especially for less experienced users in rural areas. For example, it is well known that obesity (particularly severe obesity) is highly prevalent in rural areas, a condition in which excess subcutaneous tissue can impede the transmission of ultrasonic energy due to increased absorption and scattering, resulting in greater tissue attenuation and limited sonographic windows.^64-67 While studies have highlighted that the diagnostic accuracy of AI-Echo does not differ by body size, these findings need to be validated in populations such as RURAL where obesity is common. Additionally, variations in the physiological and anatomical structure of the heart between the two sexes and in different age groups must be accounted for. This implies that an equipment model with high accuracy using a specific imaging AI algorithm may not be applicable to every population. Maintaining consistent data quality and standardization across rural sites is imperative due to variations in equipment, operator expertise, and imaging protocols.

Return of Results (RoR) to the Community

An integral part of the RURAL Imaging Core process is timely return of imaging results to participants. Routine return of all imaging results occurs after the over-read within the core laboratories, with an expected return of results to participants of less than six weeks from when the scan was received at the core lab. We have established a RoR mechanism for alerting participants of important findings identified in imaging studies. The CAC score will be reported to each participant as an absolute CAC in Agatston units and as a percentile within the CAC score distribution corresponding to the participant’s age sex and race/ethnicity in comparison to values observed in the asymptomatic US population. For CAC scores above the 75^th percentile, the report will include a recommendation to participants to further discuss the result with their healthcare provider.^68-70 Alerts and other incidental non-coronary findings warranting further follow-up will also be communicated to the participants and their physicians (if consented) promptly. Major clinical alerts include the detection of aortic aneurysms (ascending aortic diameter >45 mm), dense aortic valve calcifications (calcification score >1000), lung masses (> 30 mm), lobar pneumonia, pneumothorax, non-calcified lung nodules >8 mm and <30 mm consistent with the highest threshold for a clinically actionable nodule, large pericardial effusions, and dense(non-cystic) lesions in liver, spleen, and kidneys.⁷¹

A priori qualitative echocardiographic findings are provided to each RURAL study participant. These predetermined findings include statements regarding right and left ventricular size, right and left ventricular function, aortic size, valvular morphology, presence/absence of pericardial fluid, and valvular regurgitation, even if the findings are normal. A summary letter to the participant includes a statement regarding the overall ventricular function, as well as the presence of any clinically meaningful qualitative findings.

During the over-reading process, an initial review is performed for any critical findings that warrant earlier notification. These critical findings are returned to the participant within three days of when the images are received at the core laboratory. Findings from imaging results are not given to participants unless they have been reviewed by a physician within the core laboratory. Those participants with high-risk findings (i.e., “critical alerts”) are referred to their clinicians of record or to federally qualified health centers that have agreed to collaborate with the RURAL study. Critical alert thresholds from echocardiograms are based on ASE thresholds and include: LV ejection fraction <50%, severely enlarged LV dimensions (left ventricular internal dimension in diastole >= 6.2 cm in women or >= 6.9 cm in men), severely reduced RV function (fractional area change <= 17% or qualitatively severely impaired), pericardial effusion >2cm, severe LVH or other findings suggestive of hypertrophic cardiomyopathy, ascending aortic aneurysm >= 4.5 cm, aortic dissection or any other clinically actionable critical alert as determined by the cardiologist reader at the DCRI. The MEU staff will be trained to expedite detection of abnormal echo findings, however final interpretation and referral for medical assessment lies with the MEU supervisor in consultation with the DCRI cardiologists. ICL staff and cardiologists will be available remotely on-call to support MEU staff and Recruitment and Retention Core (RRC).

All imaging results are returned to participants along with other examination results from the RURAL study (in both English and Spanish), with basic interpretations included based on current cardiovascular guidelines.

Future Directions

Engaging community personnel to perform multimodality imaging significantly enhances access to healthcare and our understanding of local CVD burden. A study in Australia has initiated bringing advanced cardiac imaging to remote communities in the outback region to alleviate the CVD burden.⁷² Aboriginal health practitioners in these communities use AI to perform echocardiography, and health professionals will take advantage of automated remote image processing to accurately characterize heart function and pathology. This study lays the groundwork for future initiatives in rural areas of other countries as well to improve the standard of healthcare delivery. As the RURAL study primarily focuses on improving healthcare in underserved and economically challenged areas, adding advanced cardiac imaging in future studies will provide early detection of CVD risk that can help reduce the cardiac disease burden and better understand the interplay between cardiac risk factors and social, economic, and geographical factors. Our design and findings could also be adopted for long term healthcare needs in rural regions.

Conclusion

Cardiac multimodality imaging has an important role in improving cardiovascular health among the rural population where resources are scarce and access to healthcare is limited. Mitigating the rural health penalty requires bridging gaps in healthcare access and expertise. The RURAL MEU approach is an innovative approach to enhance access to a comprehensive CT scanning protocol assessing cardiac, lung, muscle and adipose tissue. AI-enhanced echocardiography bridges the gap in access to specialized imaging supported by automatic image acquisition and interpretation. These efforts in the RURAL study ensure greater workflow efficiency, enhanced repeatability, and higher diagnostic accuracy while demonstrating an accountable epidemiologic approach by returning diagnostic information to rural participants. Our design and findings provide important support and direction for future studies in rural health.