Global Health Initiatives: International Physician-to-Physician Consultation Programs.

PURPOSE: The goal of this article is to provide technical and operational blueprints for two successful global telehealth programs. METHODS: The authors designed a physician-to-physician consultation program to provide subspecialty expertise to physicians in war-torn Ukraine. Leveraging secure web applications, telehealth platforms, and image-sharing platforms, the authors repeatedly iterated upon infrastructure and workflows, which in turn facilitated the development of a parallel international program for US Department of State (DOS) employees and families. The authors provide descriptive statistics and metrics of both programs' successes and failures and detail iterative improvements with workflow visuals. To measure the added value of subspecialty imaging consultation, two radiologists performed a retrospective comparative review of the DOS program imaging reports, comparing the initial report to the consult report in consensus, measuring diagnostic report agreement, and rating the clinical impact of identified discrepancies on a three-point scale (mild, moderate, or major). Bivariate analyses using χ2 tests were conducted to assess associations between diagnostic discrepancies and patient or imaging factors. P values <.05 were considered to indicate statistical significance. RESULTS: The Ukraine program (May 2022 to August 2023) provided 114 patient consultations with 77 subspecialty radiology consults, >50 WhatsApp chats, and >1,000 messages exchanged, with a 92% overall consult request response rate. The DOS program (November 2022 to July 2023) provided 275 consultations with 70 subspecialty radiology consults and a 36% to 38% rate of alternative diagnoses, with 20% rated as incurring moderate or major clinical impact. Bivariate analyses demonstrated no significant patient or imaging association with diagnostic disagreements (P > .05 for all). CONCLUSIONS: Global telehealth infrastructure and multiple applications and platforms can be optimized in a workflow to provide efficient, high-level clinical and imaging consultation services across the globe.

A radiographic, deep transfer learning framework, adapted to estimate lung opacities from chest x-rays.

Chest radiographs (CXRs) are the most widely available radiographic imaging modality used to detect respiratory diseases that result in lung opacities. CXR reports often use non-standardized language that result in subjective, qualitative, and non-reproducible opacity estimates. Our goal was to develop a robust deep transfer learning framework and adapt it to estimate the degree of lung opacity from CXRs. Following CXR data selection based on exclusion criteria, segmentation schemes were used for ROI (Region Of Interest) extraction, and all combinations of segmentation, data balancing, and classification methods were tested to pick the top performing models. Multifold cross validation was used to determine the best model from the initial selected top models, based on appropriate performance metrics, as well as a novel Macro-Averaged Heatmap Concordance Score (MA HCS). Performance of the best model is compared against that of expert physician annotators, and heatmaps were produced. Finally, model performance sensitivity analysis across patient populations of interest was performed. The proposed framework was adapted to the specific use case of estimation of degree of CXR lung opacity using ordinal multiclass classification. Acquired between March 24, 2020, and May 22, 2020, 38,365 prospectively annotated CXRs from 17,418 patients were used. We tested three neural network architectures (ResNet-50, VGG-16, and ChexNet), three segmentation schemes (no segmentation, lung segmentation, and lateral segmentation based on spine detection), and three data balancing strategies (undersampling, double-stage sampling, and synthetic minority oversampling) using 38,079 CXR images for training, and validation with 286 images as the out-of-the-box dataset that underwent expert radiologist adjudication. Based on the results of these experiments, the ResNet-50 model with undersampling and no ROI segmentation is recommended for lung opacity classification, based on optimal values for the MAE metric and HCS (Heatmap Concordance Score). The degree of agreement between the opacity scores predicted by this model with respect to the two sets of radiologist scores (OR or Original Reader and OOBTR or Out Of Box Reader) in terms of performance metrics is superior to the inter-radiologist opacity score agreement.

Incidence and pathology of repeat CT abdomen and pelvis in an adult emergency department population.

INTRODUCTION: Our objective was to identify the incidence of adult patients who undergo more than one computed tomography (CT) abdomen and pelvis within 1 year and detect the incidence of significant pathology on these repeat scans. METHODS: All adults with an initial CT within 12 months and then during an emergency department visit were retrospectively identified. RESULTS: A percentage of 21.1 of the repeat CT scans were positive. Approximately 20% of positive repeat CT scans occurred within the first month and nearly 70% within 6 months of the initial CT scan. CONCLUSIONS: Many patients undergo multiple CT scans within a 1 year time frame with significant pathology identified.

Does a normal screening ultrasound of the abdominal aorta reduce the likelihood of rupture in emergency department patients?

INTRODUCTION: Abdominal aortic aneurysm (AAA) development is a multifactorial process that is more prevalent among people ≥65years of age. Major risk factors are obesity, male sex, history of smoking (at least 100 cigarettes in a person's lifetime), and history of AAA in a first-degree relative. The United States Preventative Task Force has recommended a one-time ultrasound screening for men aged 65-75years. Based on studies, negative results on a single ultrasound examination around the age of 65years appear to virtually exclude the risk for future AAA rupture or death. While ultrasonography (US) is the confirmatory study of choice, computed tomography (CT) can also be used in the diagnosis of AAA. The goal of this study is to determine if AAA rupture can reliably be excluded in individuals with abdominal pain who have had a normal caliber aorta on CT or US after the age of 65years. MATERIALS AND METHODS: A retrospective study (approved by institutional review board) of emergency department (ED) patients in an urban academic center was performed. Subjects were included if they met the following criteria: age ≥65years; an initial CT or US as an ED patient, inpatient, or outpatient for any indication, which identified an abdominal aorta <3cm; and a second CT or US during an ED visit. The incidence of ruptured AAA on the second CT or US with a history of normal aortic caliber was identified. RESULTS: During the study period, 606 subjects were enrolled. Demographic data are listed in Table 1. Three subjects (0.5%) exhibited an abnormal-sized aorta on ED evaluation. None of these three subjects had an AAA intervention. The average size of the abnormal aorta in these three subjects was 3.3cm (S.D. 0.17). CONCLUSION: Based on these results, it appears that AAA and rupture may reliably be excluded in ED patients with abdominal pain who have previously had a normal caliber aorta on CT or US after the age of 65years. A prospective, multicenter study would help validate these findings.

Preventing errors in laterality.

An error in laterality is the reporting of a finding that is present on the right side as on the left or vice versa. While different medical and surgical specialties have implemented protocols to help prevent such errors, very few studies have been published that describe these errors in radiology reports and ways to prevent them. We devised a system that allows the radiologist to view reports in a separate window, displayed in a simple font and with all terms of laterality highlighted in separate colors. This allows the radiologist to correlate all detected laterality terms of the report with the images open in PACS and correct them before the report is finalized. The system is monitored every time an error in laterality was detected. The system detected 32 errors in laterality over a 7-month period (rate of 0.0007 %), with CT containing the highest error detection rate of all modalities. Significantly, more errors were detected in male patients compared with female patients. In conclusion, our study demonstrated that with our system, laterality errors can be detected and corrected prior to finalizing reports.

PACS displays: how to select the right display technology.

The medical imaging display is a precision instrument with many features not found in commercial-grade displays. The more one understands what these features are and their corresponding clinical value, the better one can make a purchase decision. None of these displays maintain themselves for 5 years or more without some degree of automatic or manual performance testing. Routine calibration conformance checks are beginning to be mandated by the departments of health of many states. Most manufacturers provide mechanisms to perform these checks and keep track of their results, some more easily than others. A consistent display brightness of about 400 cd/m(2) and close conformance to the DICOM curve are the key components of a successful check. Displays are typically characterized by the number of pixels they contain, usually 2, 3, or 5 megapixels, but this is the least useful determinant of image quality. What matters most is the size of the pixels and the size of the whole display, which should be selected on the basis of the typical viewing distance. The farther one's eyes are from the display, the larger the pixels and the overall display size can be while still feeding the eye as much information as it can see. Care should be taken to use the appropriate display in a given setting for the clinical purpose at hand.

Use of mobile devices for medical imaging.

Mobile devices have fundamentally changed personal computing, with many people forgoing the desktop and even laptop computer altogether in favor of a smaller, lighter, and cheaper device with a touch screen. Doctors and patients are beginning to expect medical images to be available on these devices for consultative viewing, if not actual diagnosis. However, this raises serious concerns with regard to the ability of existing mobile devices and networks to quickly and securely move these images. Medical images often come in large sets, which can bog down a network if not conveyed in an intelligent manner, and downloaded data on a mobile device are highly vulnerable to a breach of patient confidentiality should that device become lost or stolen. Some degree of regulation is needed to ensure that the software used to view these images allows all relevant medical information to be visible and manipulated in a clinically acceptable manner. There also needs to be a quality control mechanism to ensure that a device's display accurately conveys the image content without loss of contrast detail. Furthermore, not all mobile displays are appropriate for all types of images. The smaller displays of smart phones, for example, are not well suited for viewing entire chest radiographs, no matter how small and numerous the pixels of the display may be. All of these factors should be taken into account when deciding where, when, and how to use mobile devices for the display of medical images.

Incidence and pathology of repeat computed tomography of the abdomen and pelvis in a pediatric emergency department population.

OBJECTIVES: Computed tomographic (CT) scanning is increasingly prevalent in emergency departments (EDs). It is a moderate- to high-radiation diagnostic technique that exposes more than 1 million children per year to unnecessary radiation. Repeat CT of the abdomen and pelvis (CTAP) among pediatric patients who return to the ED within 1 year may be an example of unnecessary pediatric radiation exposure. The objectives of this study were to identify the incidence of pediatric patients who undergo more than 1 CTAP within 1 year and to detect the incidence of significant pathology on these repeat scans. METHODS: This was a retrospective review of subjects younger than 18 years with an initial CTAP as an ED patient, inpatient, or outpatient and a second CTAP within 12 months and during an ED visit. RESULTS: During the observation period, 172 pediatric ED admissions had at least 1 repeat admission involving an abdominal CT scan. Thirty-seven of the CT scans (19.3%) were positive. Sixty percent of the positive cases (n = 22) were attributable to the 3 most prevalent diagnoses: appendicitis in 8 cases (21.6%), kidney stone in 8 cases (21.6%), and colitis in 6 cases (16.2%). Approximately, one third of repeat CT scans occurred within the first month of the initial CT scan, and two thirds occurred within 6 months of the initial CT scan. CONCLUSIONS: A substantial percentage of pediatric patients undergo more than 1 CTAP within a 1-year time frame. Among these patients, a large portion were diagnosed significant pathology.

Filmless in New Jersey: the New Jersey Medical School PACS Project.

Transitioning to a filmless department is no easy task, especially at a large academic medical center. At the University of Medicine and Dentistry of New Jersey-New Jersey Medical School, a phased modality integration schedule was implemented to allow the technical and clinical staff to gradually absorb all of the changes to workflow. One-on-one training sessions were designed to prepare radiologists and referring clinicians to access and navigate the in-house picture archiving and communication system (PACS) workstations as well as to view images over the Internet via the PACS Web server. An interdepartmental steering committee was formed to plan deployment of the in-house workstations. A planning committee met on a weekly basis to outline placement of workstations within the Radiology Department, and to redesign the reading room. A user group was created to discuss specific user problems. Of particular interest was the challenge of outfitting a dozen conference rooms with projection systems capable of displaying radiologic images. We distinguished between regular and working conferences. At regular conferences only a few cases are reviewed over the course of an hour and only after the diagnosis has been made at a PACS workstation. In contrast, the surgical and medical intensive care units conduct daily working conferences. At those sessions the images of 20 to 30 patients are reviewed, many of them for the first time, and for each case a definitive diagnosis is expected. During the implementation process, a range of issues came up that limited access of certain studies to radiologists and referring clinicians alike. Even after the initial PACS installation, many studies went unread because of a lack of worklists. Other problems included image ordering for head computed tomography and magnetic resonance imaging. A few of our modalities were not DICOM compliant and needed image capture devices in order to be integrated with the PACS. To our dismay, this was also true of one of our modalities that was supposed to be DICOM compliant. These problems, and the solutions we discovered, are discussed in this paper.