A Review of the Clinical Applications of Artificial Intelligence in Abdominal Imaging.

Artificial intelligence (AI) has been a topic of substantial interest for radiologists in recent years. Although many of the first clinical applications were in the neuro, cardiothoracic, and breast imaging subspecialties, the number of investigated and real-world applications of body imaging has been increasing, with more than 30 FDA-approved algorithms now available for applications in the abdomen and pelvis. In this manuscript, we explore some of the fundamentals of artificial intelligence and machine learning, review major functions that AI algorithms may perform, introduce current and potential future applications of AI in abdominal imaging, provide a basic understanding of the pathways by which AI algorithms can receive FDA approval, and explore some of the challenges with the implementation of AI in clinical practice.

Impact of Artificial Intelligence-Assisted Indication Selection on Appropriateness Order Scoring for Imaging Clinical Decision Support.

PURPOSE: The aim of this study was to assess appropriateness scoring and structured order entry after the implementation of an artificial intelligence (AI) tool for analysis of free-text indications. METHODS: Advanced outpatient imaging orders in a multicenter health care system were recorded 7 months before (March 1, 2020, to September 21, 2020) and after (October 20, 2020, to May 13, 2021) the implementation of an AI tool targeting free-text indications. Clinical decision support score (not appropriate, may be appropriate, appropriate, or unscored) and indication type (structured, free-text, both, or none) were assessed. The χ2 and multivariate logistic regression adjusting for covariables with bootstrapping were used. RESULTS: In total, 115,079 orders before and 150,950 orders after AI tool deployment were analyzed. The mean patient age was 59.3 ± 15.5 years, and 146,035 (54.9%) were women; 49.9% of orders were for CT, 38.8% for MR, 5.9% for nuclear medicine, and 5.4% for PET. After deployment, scored orders increased to 52% from 30% (P < .001). Orders with structured indications increased to 67.3% from 34.6% (P < .001). On multivariate analysis, orders were more likely to be scored after tool deployment (odds ratio [OR], 2.7, 95% CI, 2.63-2.78; P < .001). Compared with physicians, orders placed by nonphysician providers were less likely to be scored (OR, 0.80; 95% CI, 0.78-0.83; P < .001). MR (OR, 0.84; 95% CI, 0.82-0.87) and PET (OR, 0.12; 95% CI, 0.10-0.13) were less likely to be scored than CT (; P < .001). After AI tool deployment, 72,083 orders (47.8%) remained unscored, 45,186 (62.7%) with free-text-only indications. CONCLUSIONS: Embedding AI assistance within imaging clinical decision support was associated with increased structured indication orders and independently predicted a higher likelihood of scored orders. However, 48% of orders remained unscored, driven by both provider behavior and infrastructure-related barriers.

AUR-RRA Review: Logistics of Academic-Industry Partnerships in Artificial Intelligence.

The Radiology Research Alliance (RRA) of the Association of University Radiologists (AUR) convenes Task Forces to address current topics in radiology. In this article, the AUR-RRA Task Force on Academic-Industry Partnerships for Artificial Intelligence, considered issues of importance to academic radiology departments contemplating industry partnerships in artificial intelligence (AI) development, testing and evaluation. Our goal was to create a framework encompassing the domains of clinical, technical, regulatory, legal and financial considerations that impact the arrangement and success of such partnerships.

Lessons From the Free-Text Epidemic: Opportunities to Optimize Deployment of Imaging Clinical Decision Support.

OBJECTIVE: The Protecting Access to Medicare Act of 2014 requires clinicians to consult Appropriate Use Criteria (AUC) when ordering advanced imaging procedures. Free-text order indications are available when there is no applicable structured indication but are unscored by the AUC. We determined the proportion of free-text indications among all advanced imaging orders and the proportion of free-text indications that could be mapped to a single structured indication. METHODS: All outpatient advanced diagnostic imaging orders placed in a large multisite health system were recorded after initial AUC deployment (November 20, 2017, to December 19, 2017). Clinicians were prompted upon order entry to select a structured indication or enter a free-text indication. We manually reviewed the two imaging examinations with the highest rate of free-text indications: enhanced CT abdomen/pelvis and unenhanced CT head. Regression analysis examined differences in patient-, imaging-, context-, and provider-level characteristics between scored and unscored examinations. RESULTS: Among all 39,533 orders for advanced imaging procedures, 59% (23,267 of 39,533) were unscored by the system. The regression model c-statistic (0.50-0.55) demonstrated poor model fit to evaluate for differences between scored and unscored examinations. Free-text indications were found in 71% (16,440 of 23,267) of unscored examinations and 42% (16,440 of 39,533) of all examinations. Manual review of all 1,693 CT abdomen/pelvis and 1,527 CT head examinations with free-text indications revealed that 3,132 free-text indications (97%) could be mapped to a single existing structured indication. DISCUSSION: Of all initially placed outpatient advanced imaging procedure orders, 42% included free-text indications and 97% of manually reviewed free-text indications could be mapped to a single structured indication.

Renal Imaging: Core Curriculum 2019.

Renal imaging has become a fundamental part of clinical care for patients with kidney disease. Imaging strategies for the kidney have been evolving during the past hundred years and have been even more rapidly changing during the past couple of decades due to the development of modern computed tomographic techniques, magnetic resonance imaging, and more sophisticated ultrasonographic techniques, such as contrast-enhanced ultrasonography. Applying the correct radiologic study for the clinical situation maximizes the diagnostic accuracy of the imaging, and a judicious choice between techniques helps limit radiation dose and potential adverse events. This Core Curriculum outlines the imaging modalities currently in use in radiology departments and is divided into 3 sections: (1) a review of the development of renal imaging and an outline of modalities available to the nephrologist, (2) imaging strategies for select clinical situations, and (3) a discussion of some potential adverse events from imaging, including effects of iodinated contrast on kidney function, risks of gadolinium-based contrast agents in kidney failure, and potential risks of imaging techniques that use ionizing radiation.

Changes in Primary Care Health Care Utilization after Inclusion of Epidemiologic Data in Lumbar Spine MR Imaging Reports for Uncomplicated Low Back Pain.

Purpose To determine whether inclusion of an epidemiologic statement in radiology reports of lumbar magnetic resonance (MR) imaging influences downstream health care utilization in the primary care population. Materials and Methods Beginning July 1, 2013, a validated epidemiologic statement regarding prevalence of common findings in asymptomatic patients was included in all lumbar MR imaging reports at a tertiary academic medical center. Data were collected from July 1, 2012, through June 30, 2014, and retrospective analysis was completed in September 2016. The electronic medical record was reviewed to capture health care utilization rates in patients for 1 year after index MR imaging. Of 4527 eligible adult patients with low back pain referred for lumbar spine MR imaging during the study period, 375 patients had their studies ordered by in-network primary care providers, did not have findings other than degenerative disease, and had at least one follow-up encounter within the system within 1 year of index MR imaging. In the before-and-after study design, a pre-statement-implementation cohort was compared with a post-statement-implementation cohort by using univariate and multivariate statistical models to evaluate treatment utilization rates in these groups. Results Patients in the statement group were 12% less likely to be referred to a spine specialist (137 of 187 [73%] vs 159 of 188 [85%]; P = .007) and were 7% less likely to undergo repeat imaging (seven of 187 [4%] vs 20 of 188 [11%]; P = .01) compared with patients in the nonstatement group. The intervention was not associated with any change in narcotic prescription (53 of 188 [28%] vs 54 of 187 [29%]; P = .88) or with the rate of low back surgery (24 of 188 [13%] vs 16 of 187 [9%]; P = .19). Conclusion In this study, inclusion of a simple epidemiologic statement in lumbar MR imaging reports was associated with decreased utilization in high-cost domains of low back pain management. © RSNA, 2018.

Symptomatic Lumbar Facet Synovial Cysts: Clinical Outcomes Following Percutaneous CT-Guided Cyst Rupture with Intra-articular Steroid Injection.

PURPOSE: To evaluate clinical outcomes following percutaneous rupture of symptomatic lumbar facet synovial cysts (LFSCs) with intra-articular steroid injection. MATERIALS AND METHODS: In this retrospective review, 44 consecutive patients with symptomatic LFSCs received primary treatment with CT-guided synovial cyst rupture with intra-articular steroid injection. Outcomes questionnaires were obtained before and 1, 4, 26, and 52 weeks after LFSC rupture. Assessment included pain medication use and numeric rating scale (NRS), Oswestry Disability Index (ODI), and 12-item short form health survey (SF-12) physical and mental composite scores (PCS and MCS). Clinical endpoint was 52-week survey response or surgery. RESULTS: LFSC rupture was technically successful in 84% (37/44) of cases. Clinical endpoint was reached in 68% (30/44) of patients with 82% overall 1-year follow-up. Lumbar spine surgery was performed in 25% (11/44) of patients within 1 year after procedure. Mean NRS, ODI, and SF-12 PCS demonstrated significant improvement at all follow-up time points (P < .001). At 52-week follow-up, NRS decreased from 8.1 to 3.7 (P < .001), ODI improved from 35 to 24 (P = .006), and SF-12 PCS improved from 31 to 42 (P < .001). Daily pain medication decreased from 71% (31/44) of patients before procedure to 29% (9/26) at 52-week follow-up (P = .012). History of prior lumbar intervention was associated with poorer LFSC rupture success (P = .025) and ODI (P = .047). CONCLUSIONS: NRS, ODI, and SF-12 PCS indices improved and pain medication use decreased significantly at all time points over 1-year follow-up after percutaneous rupture of symptomatic LFSCs with intra-articular steroid injection.

Outpatient follow-up of nonoperative cerebral contusion and traumatic subarachnoid hemorrhage: does repeat head CT alter clinical decision-making?

OBJECTIVES: Many neurosurgeons obtain repeat head CT at the first clinic follow-up visit for nonoperative cerebral contusion and traumatic subarachnoid hemorrhage (tSAH). The authors undertook a single-center, retrospective study to determine whether outpatient CT altered clinical decision-making. METHODS: The authors evaluated 173 consecutive adult patients admitted to their institution from April 2006 to August 2012 with an admission diagnosis of cerebral contusion or tSAH and at least 1 clinic follow-up visit with CT. Patients with epidural, subdural, aneurysmal subarachnoid, or intraventricular hemorrhage, and those who underwent craniotomy, were excluded. Patient charts were reviewed for new CT findings, new patient symptoms, and changes in treatment plan. Patients were stratified by neurological symptoms into 3 groups: 1) asymptomatic; 2) mild, nonspecific symptoms; and 3) significant symptoms. Mild, nonspecific symptoms included minor headaches, vertigo, fatigue, and mild difficulties with concentration, short-term memory, or sleep; significant symptoms included moderate to severe headaches, nausea, vomiting, focal neurological complaints, impaired consciousness, or new cognitive impairment evident on routine clinical examination. RESULTS: One hundred seventy-three patients met inclusion criteria, with initial clinic follow-up obtained within approximately 6 weeks. Of the 173 patients, 104 (60.1%) were asymptomatic, 68 patients (39.3%) had mild, nonspecific neurological symptoms, and 1 patient (1.0%) had significant neurological symptoms. Of the asymptomatic patients, 3 patients (2.9%) had new CT findings and 1 of these patients (1.0%) underwent a change in treatment plan because of these findings. This change involved an additional clinic appointment and CT to monitor a 12-mm chronic subdural hematoma that ultimately resolved without treatment. Of the patients with mild, nonspecific neurological symptoms, 6 patients (8.8%) had new CT findings and 3 of these patients (4.4%) underwent a change in treatment plan because of these findings; none of these patients required surgical intervention. The single patient with significant neurological symptoms did not have any new CT findings. CONCLUSIONS: Repeat outpatient CT of asymptomatic patients after nonoperative cerebral contusion and tSAH is very unlikely to demonstrate significant new pathology. Given the cost and radiation exposure associated with CT, imaging should be reserved for patients with significant symptoms or focal findings on neurological examination.