An artificial intelligence algorithm for pulmonary embolism detection on polychromatic computed tomography: performance on virtual monochromatic images.

OBJECTIVES: Virtual monochromatic images (VMI) are increasingly used in clinical practice as they improve contrast-to-noise ratio. However, due to their different appearances, the performance of artificial intelligence (AI) trained on conventional CT images may worsen. The goal of this study was to assess the performance of an established AI algorithm trained on conventional polychromatic computed tomography (CT) images (CPI) to detect pulmonary embolism (PE) on VMI. METHODS: Paired 60 kiloelectron volt (keV) VMI and CPI of 114 consecutive patients suspected of PE, obtained with a detector-based spectral CT scanner, were retrospectively analyzed by an established AI algorithm. The CT pulmonary angiography (CTPA) were classified as positive or negative for PE on a per-patient level. The reference standard was established using a comprehensive method that combined the evaluation of the attending radiologist and three experienced cardiothoracic radiologists aided by two different detection tools. Sensitivity, specificity, positive and negative predictive values and likelihood ratios of the algorithm on VMI and CPI were compared. RESULTS: The prevalence of PE according to the reference standard was 35.1% (40 patients). None of the diagnostic accuracy measures of the algorithm showed a significant difference between CPI and VMI. Sensitivity was 77.5% (95% confidence interval (CI) 64.6-90.4%) and 85.0% (73.9-96.1%) (p = 0.08) on CPI and VMI respectively and specificity 96.0% (91.4-100.0%) and 94.6% (89.4-99.7%) (p = 0.32). CONCLUSIONS: Diagnostic performance of the AI algorithm that was trained on CPI did not drop on VMI, which is reassuring for its use in clinical practice. CLINICAL RELEVANCE STATEMENT: A commercially available AI algorithm, trained on conventional polychromatic CTPA, could be safely used on virtual monochromatic images. This supports the sustainability of AI-aided detection of PE on CT despite ongoing technological advances in medical imaging, although monitoring in daily practice will remain important. KEY POINTS: • Diagnostic accuracy of an AI algorithm trained on conventional polychromatic images to detect PE did not drop on virtual monochromatic images. • Our results are reassuring as innovations in hardware and reconstruction in CT are continuing, whilst commercial AI algorithms that are trained on older generation data enter healthcare.

Diagnostic accuracy of an artificial intelligence algorithm versus radiologists for fracture detection on cervical spine CT.

OBJECTIVES: To compare diagnostic accuracy of a deep learning artificial intelligence (AI) for cervical spine (C-spine) fracture detection on CT to attending radiologists and assess which undetected fractures were injuries in need of stabilising therapy (IST). METHODS: This single-centre, retrospective diagnostic accuracy study included consecutive patients (age ≥18 years; 2007-2014) screened for C-spine fractures with CT. To validate ground truth, one radiologist and three neurosurgeons independently examined scans positive for fracture. Negative scans were followed up until 2022 through patient files and two radiologists reviewed negative scans that were flagged positive by AI. The neurosurgeons determined which fractures were ISTs. Diagnostic accuracy of AI and attending radiologists (index tests) were compared using McNemar. RESULTS: Of the 2368 scans (median age, 48, interquartile range 30-65; 1441 men) analysed, 221 (9.3%) scans contained C-spine fractures with 133 IST. AI detected 158/221 scans with fractures (sensitivity 71.5%, 95% CI 65.5-77.4%) and 2118/2147 scans without fractures (specificity 98.6%, 95% CI 98.2-99.1). In comparison, attending radiologists detected 195/221 scans with fractures (sensitivity 88.2%, 95% CI 84.0-92.5%, p < 0.001) and 2130/2147 scans without fracture (specificity 99.2%, 95% CI 98.8-99.6, p = 0.07). Of the fractures undetected by AI 30/63 were ISTs versus 4/26 for radiologists. AI detected 22/26 fractures undetected by the radiologists, including 3/4 undetected ISTs. CONCLUSION: Compared to attending radiologists, the artificial intelligence has a lower sensitivity and a higher miss rate of fractures in need of stabilising therapy; however, it detected most fractures undetected by the radiologists, including fractures in need of stabilising therapy. Clinical relevance statement The artificial intelligence algorithm missed more cervical spine fractures on CT than attending radiologists, but detected 84.6% of fractures undetected by radiologists, including fractures in need of stabilising therapy. KEY POINTS: The impact of artificial intelligence for cervical spine fracture detection on CT on fracture management is unknown. The algorithm detected less fractures than attending radiologists, but detected most fractures undetected by the radiologists including almost all in need of stabilising therapy. The artificial intelligence algorithm shows potential as a concurrent reader.

Model-based Prediction of Critical Illness in Hospitalized Patients with COVID-19.

Background The prognosis of hospitalized patients with severe coronavirus disease 2019 (COVID-19) is difficult to predict, and the capacity of intensive care units was a limiting factor during the peak of the pandemic and is generally dependent on a country's clinical resources. Purpose To determine the value of chest radiographic findings together with patient history and laboratory markers at admission to predict critical illness in hospitalized patients with COVID-19. Materials and Methods In this retrospective study, which included patients from March 7, 2020, to April 24, 2020, a consecutive cohort of hospitalized patients with real-time reverse transcription polymerase chain reaction-confirmed COVID-19 from two large Dutch community hospitals was identified. After univariable analysis, a risk model to predict critical illness (ie, death and/or intensive care unit admission with invasive ventilation) was developed, using multivariable logistic regression including clinical, chest radiographic, and laboratory findings. Distribution and severity of lung involvement were visually assessed by using an eight-point scale (chest radiography score). Internal validation was performed by using bootstrapping. Performance is presented as an area under the receiver operating characteristic curve. Decision curve analysis was performed, and a risk calculator was derived. Results The cohort included 356 hospitalized patients (mean age, 69 years ± 12 [standard deviation]; 237 men) of whom 168 (47%) developed critical illness. The final risk model's variables included sex, chronic obstructive lung disease, symptom duration, neutrophil count, C-reactive protein level, lactate dehydrogenase level, distribution of lung disease, and chest radiography score at hospital presentation. The area under the receiver operating characteristic curve of the model was 0.77 (95% CI: 0.72, 0.81; P < .001). A risk calculator was derived for individual risk assessment: Dutch COVID-19 risk model. At an example threshold of 0.70, 71 of 356 patients would be predicted to develop critical illness, of which 59 (83%) would be true-positive results. Conclusion A risk model based on chest radiographic and laboratory findings obtained at admission was predictive of critical illness in hospitalized patients with coronavirus disease 2019. This risk calculator might be useful for triage of patients to the limited number of intensive care unit beds or facilities. © RSNA, 2020 Online supplemental material is available for this article.

Leukocyte Dynamics during the Evolution of Human Coronary Atherosclerosis: Conclusions from a Sevenfold, Chromogen-Based, Immunohistochemical Evaluation.

Atherosclerosis is a complex process with strong inflammatory component. We developed a straightforward sevenfold staining protocol for simultaneous assessment of dominant leukocyte classes, vascularization, and expression of the putative foam cell maker CD36. The method was applied on human coronaries covering the full spectrum of atherosclerotic disease. Results confirm the progressive association of macrophages and T cells with the process and a global presence of mast cells. B cells are exclusively present in adventitial follicles that accompany the process plaque destabilization (thin cap and ruptured lesions) and are otherwise absent. Neutrophils are only present as part of the hemorrhage that accompanies plaque rupture. This study does not classify CD36 as a key factor in foam cell formation. Observed macrophage accumulation in the neointima of stabilized fibrous calcified plaques is consistent with a process of neoatherosclerosis. This study on human coronaries shows a progressive association of macrophage and T-cell abundance with plaque progression. Follicle-like structures are transiently present during the process of plaque destabilization. Plaque healing is accompanied by cessation of the inflammatory response but followed by a new cycle of atherosclerosis.

Validating human and mouse tissues commonly used in atherosclerosis research with coronary and aortic reference tissue: similarities but profound differences in disease initiation and plaque stability.

Objective: Characterization of the atherosclerotic process fully relies on histological evaluation and staging through a consensus grading system. So far, a head-to-head comparison of atherosclerotic process in experimental models and tissue resources commonly applied in atherosclerosis research with the actual human atherosclerotic process is missing. Material and Methods: Aspects of the atherosclerotic process present in established murine atherosclerosis models and human carotid endarterectomy specimen were systematically graded using the modified American Heart Association histological classification (Virmani classification). Aspects were aligned with the atherosclerotic process observed in human coronary artery and aortic atherosclerosis reference tissues that were available through biobanks based on human tissue/organ donor material. Results: Apart from absent intraplaque hemorrhages in aortic lesions, the histological characteristics of the different stages of human coronary and aortic atherosclerosis are similar. Carotid endarterectomy samples all represent end-stage "fibrous calcified plaque" lesions, although secondary, progressive, and vulnerable lesions with gross morphologies similar to coronary/aortic lesions occasionally present along the primary lesions. For the murine lesions, clear histological parallels were observed for the intermediate lesion types ("pathological intimal thickening," and "early fibroatheroma"). However, none of the murine lesions studied progressed to an equivalent of late fibroatheroma or beyond. Notable contrasts were observed for disease initiation: whereas disease initiation in humans is characterized by a mesenchymal cell influx in the intima, the earliest murine lesions are exclusively intimal, with subendothelial accumulation foam cells. A mesenchymal (and medial) response are absent. In fact, it is concluded that the stage of "adaptive intimal thickening" is absent in all mouse models included in this study. Conclusions: The Virmani classification for coronary atherosclerosis can be applied for systematically grading experimental and clinical atherosclerosis. Application of this histological grading tool shows clear parallels for intermediate human and murine atherosclerotic lesions. However, clear contrasts are observed for disease initiation, and late stage atherosclerotic lesions. Carotid endarterectomy all represent end-stage fibrous calcified plaque lesions, although secondary earlier lesions may present in a subset of samples.

Differential expression of oxidation-specific epitopes and apolipoprotein(a) in progressing and ruptured human coronary and carotid atherosclerotic lesions.

The relationships between oxidation-specific epitopes (OSE) and lipoprotein (a) [Lp(a)] and progressive atherosclerosis and plaque rupture have not been determined. Coronary artery sections from sudden death victims and carotid endarterectomy specimens were immunostained for apoB-100, oxidized phospholipids (OxPL), apo(a), malondialdehyde-lysine (MDA), and MDA-related epitopes detected by antibody IK17 and macrophage markers. The presence of OxPL captured in carotid and saphenous vein graft distal protection devices was determined with LC-MS/MS. In coronary arteries, OSE and apo(a) were absent in normal coronary arteries and minimally present in early lesions. As lesions progressed, apoB and MDA epitopes did not increase, whereas macrophage, apo(a), OxPL, and IK17 epitopes increased proportionally, but they differed according to plaque type and plaque components. Apo(a) epitopes were present throughout early and late lesions, especially in macrophages and the necrotic core. IK17 and OxPL epitopes were strongest in late lesions in macrophage-rich areas, lipid pools, and the necrotic core, and they were most specifically associated with unstable and ruptured plaques. Specific OxPL were present in distal protection devices. Human atherosclerotic lesions manifest a differential expression of OSEs and apo(a) as they progress, rupture, and become clinically symptomatic. These findings provide a rationale for targeting OSE for biotheranostic applications in humans.

Activator protein-1 (AP-1) signalling in human atherosclerosis: results of a systematic evaluation and intervention study.

Animal studies implicate the AP-1 (activator protein-1) pro-inflammatory pathway as a promising target in the treatment of atherosclerotic disease. It is, however, unclear whether these observations apply to human atherosclerosis. Therefore we evaluated the profile of AP-1 activation through histological analysis and tested the potential benefit of AP-1 inhibition in a clinical trial. AP-1 activation was quantified by phospho-c-Jun nuclear translocation (immunohistochemistry) on a biobank of aortic wall samples from organ donors. The effect of AP-1 inhibition on vascular parameters was tested through a double blind placebo-controlled cross-over study of 28 days doxycycline or placebo in patients with symptomatic peripheral artery disease. Vascular function was assessed by brachial dilation as well as by plasma samples analysed for hs-CRP (high-sensitivity C-reactive protein), IL-6 (interleukin-6), IL-8, ICAM-1 (intercellular adhesion molecule-1), vWF (von Willebrand factor), MCP-1 (monocyte chemoattractant protein-1), PAI-1 (plasminogen activator inhibitor-1) and fibrinogen. Histological evaluation of human atherosclerosis showed minimal AP-1 activation in non-diseased arterial wall (i.e. vessel wall without any signs of atherosclerotic disease). A gradual increase of AP-1 activation was found in non-progressive and progressive phases of atherosclerosis respectively (P<0.044). No significant difference was found between progressive and vulnerable lesions. The expression of phospho-c-Jun diminished as the lesion stabilized (P<0.016) and does not significantly differ from the normal aortic wall (P<0.33). Evaluation of the doxycycline intervention only revealed a borderline-significant reduction of circulating hs-CRP levels (-0.51 µg/ml, P=0.05) and did not affect any of the other markers of systemic inflammation and vascular function. Our studies do not characterize AP-1 as a therapeutic target for progressive human atherosclerotic disease.

Effect of Physical Therapy vs Arthroscopic Partial Meniscectomy in People With Degenerative Meniscal Tears: Five-Year Follow-up of the ESCAPE Randomized Clinical Trial.

Importance: There is a paucity of high-quality evidence about the long-term effects (ie, 3-5 years and beyond) of arthroscopic partial meniscectomy vs exercise-based physical therapy for patients with degenerative meniscal tears. Objectives: To compare the 5-year effectiveness of arthroscopic partial meniscectomy and exercise-based physical therapy on patient-reported knee function and progression of knee osteoarthritis in patients with a degenerative meniscal tear. Design, Setting, and Participants: A noninferiority, multicenter randomized clinical trial was conducted in the orthopedic departments of 9 hospitals in the Netherlands. A total of 321 patients aged 45 to 70 years with a degenerative meniscal tear participated. Data collection took place between July 12, 2013, and December 4, 2020. Interventions: Patients were randomly allocated to arthroscopic partial meniscectomy or 16 sessions of exercise-based physical therapy. Main Outcomes and Measures: The primary outcome was patient-reported knee function (International Knee Documentation Committee Subjective Knee Form (range, 0 [worst] to 100 [best]) during 5 years of follow-up based on the intention-to-treat principle, with a noninferiority threshold of 11 points. The secondary outcome was progression in knee osteoarthritis shown on radiographic images in both treatment groups. Results: Of 321 patients (mean [SD] age, 58 [6.6] years; 161 women [50.2%]), 278 patients (87.1%) completed the 5-year follow-up with a mean follow-up time of 61.8 months (range, 58.8-69.5 months). From baseline to 5-year follow-up, the mean (SD) improvement was 29.6 (18.7) points in the surgery group and 25.1 (17.8) points in the physical therapy group. The crude between-group difference was 3.5 points (95% CI, 0.7-6.3 points; P < .001 for noninferiority). The 95% CI did not exceed the noninferiority threshold of 11 points. Comparable rates of progression of radiographic-demonstrated knee osteoarthritis were noted between both treatments. Conclusions and Relevance: In this noninferiority randomized clinical trial after 5 years, exercise-based physical therapy remained noninferior to arthroscopic partial meniscectomy for patient-reported knee function. Physical therapy should therefore be the preferred treatment over surgery for degenerative meniscal tears. These results can assist in the development and updating of current guideline recommendations about treatment for patients with a degenerative meniscal tear. Trial Registration: ClinicalTrials.gov Identifier: NCT01850719.

Qualitative evaluation of coronary atherosclerosis in a large cohort of young and middle-aged Dutch tissue donors implies that coronary thrombo-embolic manifestations are stochastic.

BACKGROUND AND AIMS: With the intention to gain support for the hypothesis that incident ischemic complications of atherosclerotic disease involve a stochastic aspect, we performed a histological, qualitative evaluation of the epidemiology of coronary atherosclerotic disease in a cohort of aortic valve donors. PATIENTS AND METHODS: Donors (n = 695, median age 54, range 11-65 years) were dichotomized into a non-cardiovascular (non-CVD) and a cardiovascular disease death (CVD) group. Consecutive 5 mm proximal left coronary artery segments were Movat stained, and the atherosclerotic burden for each segment was graded (revised AHA-classification). RESULTS: Non-CVD and CVD groups showed steep increase of atherosclerosis severity beyond the age of 40, resulting in an endemic presence of advanced atherosclerosis in men over 40 and women over 50 years. In fact, only 19% of the non-CVD and 6% of the CVD donors over 40 years were classified with a normal LCA or a so called non-progressive lesion type. Fibrous calcified plaques (FCP), the consolidated remnants of earlier ruptured lesions, dominated in both non-CVD and CVD donors. Estimates of the atherosclerosis burden (i.e. average lesion grade, proportion of FCPs, and average number of FCPs per cross-section) were all higher in the CVD group (p<1.10-16, p<0.0001, and p<0.05, respectively). CONCLUSIONS: Dominance of consolidated FCP lesions in males over 40 and females over 50 years, show that plaque ruptures in the left coronary artery are common. However, the majority of these ruptures remain asymptomatic. This implies that the atherosclerotic process is repetitive. A relative difference in disease burden between CVD and non-CVD donors supports the concept that complications of atherosclerotic disease involve a stochastic element.

Systematic Evaluation of the Cellular Innate Immune Response During the Process of Human Atherosclerosis.

BACKGROUND: The concept of innate immunity is well recognized within the spectrum of atherosclerosis, which is primarily dictated by macrophages. Although current insights to this process are largely based on murine models, there are fundamental differences in the atherosclerotic microenvironment and associated inflammatory response relative to humans. In this light, we characterized the cellular aspects of innate immune response in normal, nonprogressive, and progressive human atherosclerotic plaques. METHODS AND RESULTS: A systematic analysis of innate immune response was performed on 110 well-characterized human perirenal aortic plaques with immunostaining for specific macrophage subtypes (M1 and M2 lineage) and their activation markers, neopterin and human leukocyte antigen-antigen D related (HLA-DR), together with dendritic cells (DCs), natural killer (NK) cells, mast cells, neutrophils, and eosinophils. Normal aortae were devoid of low-density lipoprotein, macrophages, DCs, NK cells, mast cells, eosinophils, and neutrophils. Early, atherosclerotic lesions exhibited heterogeneous populations of (CD68(+)) macrophages, whereby 25% were double positive "M1" (CD68(+)/ inducible nitric oxide synthase [iNOS](+)/CD163(-)), 13% "M2" double positive (CD68(+)/iNOS(-)/CD163(+)), and 17% triple positive for (M1) iNOS (M2)/CD163 and CD68, with the remaining (≈40%) only stained for CD68. Progressive fibroatheromatous lesions, including vulnerable plaques, showed increasing numbers of NK cells and fascin-positive cells mainly localized to the media and adventitia whereas the M1/M2 ratio and level of macrophage activation (HLA-DR and neopterin) remained unchanged. On the contrary, stabilized (fibrotic) plaques showed a marked reduction in macrophages and cell activation with a concomitant decrease in NK cells, DCs, and neutrophils. CONCLUSIONS: Macrophage "M1" and "M2" subsets, together with fascin-positive DCs, are strongly associated with progressive and vulnerable atherosclerotic disease of human aorta. The observations here support a more complex theory of macrophage heterogeneity than the existing paradigm predicated on murine data and further indicate the involvement of (poorly defined) macrophage subtypes or greater dynamic range of macrophage plasticity than previously considered.

Histological evaluation disqualifies IMT and calcification scores as surrogates for grading coronary and aortic atherosclerosis.

BACKGROUND/OBJECTIVES: Carotid intimal media thickness (IMT) and coronary calcium scores (CCS) are thought to reflect atherosclerotic burden. The validity of this assumption for IMT is challenged by recent meta-analyses; for CCS by absence of a relationship between negative scores, and freedom of future events. As such, we considered evaluation of the relationship between tissue IMT and CCS, and extend of atherosclerotic disease relevant. METHODS: Analyses were performed on donor aortas obtained during renal graft procurement, and on coronary arteries collected during heart valve procurement for tissue donation. Movat pentachrome and Hematoxylin staining was performed, and the degree of atherosclerosis histologically graded. IMT and presence of calcium deposits were quantified on graded tissue sections. RESULTS: 304 aortas and 185 coronary arteries covering the full atherosclerotic spectrum were evaluated. Aortas and coronaries showed similar relationships between tissue IMT and degree of atherosclerosis, with gradual increase in tissue IMT during earlier phases of atherosclerosis (r=0.68 and r=0.30, P<0.00001 for aorta and coronaries respectively), followed by plateauing of the curve in intermediate and advanced stages. Results for tissue IMT reveal high variability, resulting in wide confidence intervals. Results for CCS are similar for aorta and coronaries, with calcium depositions limited to advanced lesions. CONCLUSIONS: Histological IMT measurements for the aorta and coronaries show large variations around the trend and plateauing of, and possibly reductions in IMT in late stage atherosclerotic disease. These observations for the aorta and coronaries may (partly) explain the limited benefit of including carotid IMT in risk prediction algorithms.