Effect of a flipped classroom course to foster medical students' AI literacy with a focus on medical imaging: a single group pre-and post-test study.

BACKGROUND: The use of artificial intelligence applications in medicine is becoming increasingly common. At the same time, however, there are few initiatives to teach this important and timely topic to medical students. One reason for this is the predetermined medical curriculum, which leaves very little room for new topics that were not included before. We present a flipped classroom course designed to give undergraduate medical students an elaborated first impression of AI and to increase their "AI readiness". METHODS: The course was tested and evaluated at Bonn Medical School in Germany with medical students in semester three or higher and consisted of a mixture of online self-study units and online classroom lessons. While the online content provided the theoretical underpinnings and demonstrated different perspectives on AI in medical imaging, the classroom sessions offered deeper insight into how "human" diagnostic decision-making differs from AI diagnoses. This was achieved through interactive exercises in which students first diagnosed medical image data themselves and then compared their results with the AI diagnoses. We adapted the "Medical Artificial Intelligence Scale for Medical Students" to evaluate differences in "AI readiness" before and after taking part in the course. These differences were measured by calculating the so called "comparative self-assessment gain" (CSA gain) which enables a valid and reliable representation of changes in behaviour, attitudes, or knowledge. RESULTS: We found a statistically significant increase in perceived AI readiness. While values of CSA gain were different across items and factors, the overall CSA gain regarding AI readiness was satisfactory. CONCLUSION: Attending a course developed to increase knowledge about AI in medical imaging can increase self-perceived AI readiness in medical students.

Renal artery stenosis: comparative assessment by unenhanced renal artery MRA versus contrast-enhanced MRA.

OBJECTIVES: To evaluate steady-state free precession (SSFP) non-contrast-enhanced MR angiography (Unenhanced-MRA) versus conventional contrast-enhanced MR angiography (CE-MRA) in the detection of renal artery stenosis (RAS). METHODS: Retrospective analysis of 70 consecutive patients referred for suspected RAS, examined by SSFP Unenhanced-MRA and CE-MRA. Image quality, quality of visible renal arterial segments, presence and grade of RAS were evaluated. The Unenhanced-MRA were compared against reference standard CE-MRA results. RESULTS: 149 renal arteries were assessed with 21 haemodynamically significant stenoses (≥ 50% stenosis) demonstrated by CE-MRA. Combined sensitivity and specificity for RAS detection by Unenhanced-MRA was 72.8% and 97.8% respectively. There is substantial correlation for RAS detection between Unenhanced-MRA and CE-MRA with kappa values of between 0.64 and 0.74. There was excellent inter-observer correlation for RAS on Unenhanced-MRA (kappa values 0.82-1.0). CONCLUSIONS: Our study has shown Unenhanced-MRA to be a viable alternative to CE-MRA, yielding images equal in quality without the requirement for gadolinium contrast agents. The sensitivity and specificity for the detection of haemodynamically significant stenoses are comparable to CE-MRA. Potentially, Unenhanced-MRA could be used as an initial investigation to avoid performing CE-MRA in patients with normal renal arteries, however we suggest that its real value will lie in being complementary to CE-MRA.

Use of advanced imaging techniques during visits to emergency departments--implications, costs, patient benefits/risks.

25 years ago, on a Friday evening at 9 pm, the emergency department (ED) was full of patients with a wide range of clinical problems. Their investigations included plain radiographs, but no other imaging was included until the next working day. At present, many patients are receiving advanced imaging such as ultrasound, CT and MRI, often delivered out of hours--an obvious advance for patients or sometimes an unnecessary development? In this article, we will consider how to assess patient benefits and whether increased use of advanced imaging is an overall advance for patients. We will address the general implications for healthcare services which come with greater use of advanced imaging. We will then address the effect of advanced imaging on individual fictional ED patients with a variety of complaints.

Midfoot and hindfoot bone marrow edema identified by magnetic resonance imaging in feet of subjects with diabetes and neuropathic ulceration is common but of unknown clinical significance.

OBJECTIVE: We conducted a retrospective cohort study assessing the prevalence and clinical and radiological outcome of remote areas of bone marrow edema on magnetic resonance imaging (MRI) in the feet of subjects with diabetes and neuropathic foot ulceration. RESEARCH DESIGN AND METHODS: MRIs performed over 6 years looking for osteomyelitis associated with neuropathic lesions were assessed for remote areas of signal change. RESULTS: Seventy MRI studies were assessed. Remote areas of signal change were present in 21 (30%) subjects, involved midfoot or hindfoot in 20 subjects, were associated with younger age and renal replacement therapy, and did not predict future Charcot neuroarthropathy or infection at that site. Repeat MRIs in 11 subjects with such areas found that none had progressed, six had improved, and two had resolved; in 29 subjects without such areas, five had developed new areas. CONCLUSIONS: Bone marrow edema in the midfoot and hindfoot of subjects with diabetes and neuropathic lesions is common, often transient, and of unknown significance.

Feasibility of abdomino-pelvic T1-weighted real-time thermal mapping of laser ablation.

PURPOSE: To prove the hypothesis that T1-weighted (T1W) thermal mapping is reliable and achievable in magnetic resonance (MR)-guided laser tumor ablation. MATERIALS AND METHODS: Near real-time gray and color-scale T1W thermal maps in 111 MR-guided laser thermal ablations (LTA) of liver, kidney, and uterine tumors were studied. After laser fiber placement, near real-time gray and color-scale thermal maps were produced. Previous work showed T1 signal is inversely proportional to temperature below 55 degrees C (the point of irreversible necrosis). RESULTS: Thermal mapping was successful in 84% of uterine, 74% of hepatic, and 20% of renal ablations. For hepatic and uterine tumors, size and conspicuity of thermal lesions were significantly greater on subtracted colorized images rather than gray-scale raw image mapping (P = 0.001, paired Student's t-test). Patient movement (N = 24), fiber charring N = 2), magnetic field distortion, and reconstruction errors (N = 2) caused mapping failure. For both renal and hepatic tumors, comparison of near real-time T1W colorized thermal map and follow up gadolinium-enhanced MR imaging revealed moderate correlation (Pearson correlation coefficient of 0.55 and 0.5, respectively). CONCLUSION: In hepatic, renal, and uterine thermal maps, the color scale produced significantly greater sized lesions with significantly greater conspicuity than the gray scale. T1W thermal mapping was reliable and successfully achieved in 73.7% of procedures, and predicted the ablated area of the tumor moderately well.