Publications

COVID-19, which is caused by SARS-CoV-2, can result in acute respiratory distress syndrome and multiple organ failure1-4, but little is known about its pathophysiology. Here we generated single-cell atlases of 24 lung, 16 kidney, 16 liver and 19 heart autopsy tissue samples and spatial atlases of 14 lung samples from donors who died of COVID-19. Integrated computational analysis uncovered substantial remodelling in the lung epithelial, immune and stromal compartments, with evidence of multiple paths of failed tissue regeneration, including defective alveolar type 2 differentiation and expansion of fibroblasts and putative TP63+ intrapulmonary basal-like progenitor cells. Viral RNAs were enriched in mononuclear phagocytic and endothelial lung cells, which induced specific host programs. Spatial analysis in lung distinguished inflammatory host responses in lung regions with and without viral RNA. Analysis of the other tissue atlases showed transcriptional alterations in multiple cell types in heart tissue from donors with COVID-19, and mapped cell types and genes implicated with disease severity based on COVID-19 genome-wide association studies. Our foundational dataset elucidates the biological effect of severe SARS-CoV-2 infection across the body, a key step towards new treatments.

PURPOSE: Non-contrast CT ASPECTS (NCCTasp) has an established role in determining eligibility for mechanical thrombectomy in centers without ready access to perfusion or DWI. Moreover, it has been suggested that CTA source ASPECTS (CTAasp) may be superior to NCCTasp in predicting final infarct volume (FIV). In this study, we hypothesized that CTA maximum intensity projection ASPECTS (MIPSasp) would be superior compared to both NCCTasp and CTAasp in predicting FIV as measured by DWI.

MATERIALS AND METHODS: In 41 consecutive patients with MCA territory infarcts, NCCTasp, CTAasp and MIPSasp were visually assessed by 2 neuroradiologists. Disagreements were adjudicated by a third neuroradiologist, and the reconciled data used for all further analysis. MR-DWI was used as the standard for FIV determination. Receiver operating characteristic curve analysis was used to compare the area under the curve for all three CT-based methods in predicting FIV ≥70 ml.

RESULTS: MIPSasp (AUC: 0.98, CI: 0.88-1.00) were statistically better than NCCTasp (AUC: 0.87, 95% CI: 0.72-0.95; p=0.01) in predicting FIV ≥70 ml. MIPSasp were also superior to CTAasp (AUC: 0.9, CI: 0.79-.98; p˂0.05). Optimal test performance for predicting FIV ≥70 ml for MIPSasp was ≤6 (sensitivity=100%, specificity=91.4%; Youden's J=0.98).

CONCLUSION: Our preliminary study suggests that a novel CTA-MIPS derived ASPECTS better predicts large MCA territory infarcts compared to CTA source and non-contrast ASPECTS. Thus, MIPSasp may be a promising technique for future studies aimed at improving ischemic stroke treatment in centers using ASPECTS for stroke management.

PURPOSE: To evaluate whether the recalculation of lung shunt fraction (LSF) is necessary prior to next-stage or same lobe repeat radioembolization.

MATERIALS AND METHODS: Retrospective chart review was performed for patients who underwent radioembolization between February 2008 and December 2018. Eighty of 312 patients had repeat mapping angiograms and LSF calculations. A total of 160 LSF calculations were made using planar imaging (155, [97%]) and single-photon emission computed tomography (5 [3%]) technetium-99m macroaggregated albumin hepatic arterial injection imaging. The mean patient age was 61.8 years ± 12.7; 69 (86%) patients had metastatic disease and 11 (14%) had hepatocellular carcinoma.

RESULTS: Patients had a median LSF of 5% (interquartile range [IQR] 3%-9%) with a median absolute difference of 1.25 (IQR 0.65-3.4) and a median of 76 days (IQR 42.5-120 days) between repeat LSF calculations. There was a median change in LSF of 0.2% between mapping studies (P = .11). There was no statistical significance between the repeat LSFs regardless of the arterial distribution (P = .79) or between tumor types (P = .75). No patients exceeded lung dose limits using actual or predicted prescribed dose amounts. The actual median lung dose was 2.6 Gy (IQR 1.8-4.4 Gy, maximum = 20.5) for the first radioembolization and 2.0 Gy (IQR 1.3-3.7 Gy, maximum = 10.1) for the second radioembolization.

CONCLUSIONS: No significant difference in LSF was identified between different time points and arterial distributions within the same patient undergoing repeat radioembolization. In patients who receive well under 30-Gy lung dose for the initial treatment and a 50-Gy cumulative lung dose, repeat radioembolization treatments in the same patient may not require a repeat LSF calculation.

Primary aldosteronism is an underdiagnosed cause of hypertension. Although inadequate screening is one reason for underdiagnosis, another important contributor is that clinicians may inappropriately exclude the diagnosis when screening aldosterone concentrations fall below traditionally established thresholds. We evaluated the intraindividual variability in screening aldosterone concentrations and aldosterone-to-renin ratios, and how this variability could impact case detection, among 51 patients with confirmed primary aldosteronism who had 2 or more screening measurements of renin and aldosterone on different days. There were a total of 137 screening measurements with a mean of 3 (range 2-6) per patient. The mean intraindividual variability, expressed as coefficients of variation, was 31% for aldosterone and 45% for the aldosterone-to-renin ratio. Aldosterone concentrations ranged from 4.9 to 51 ng/dL; 49% of patients had at least one aldosterone measurement below 15 ng/dL, 29% had at least 2 aldosterone measurements below 15 ng/dL, and 29% had at least one measurement below 10 ng/dL. Individual aldosterone-to-renin ratios ranged from 8.2 to 427 ng/dL per ng/mL·hour; 57% had at least one ratio below 30 ng/dL per ng/mL·hour, 27% had at least 2 ratios below 30 ng/dL per ng/mL·hour, and 24% had at least one ratio below 20 ng/dL per ng/mL·hour. Aldosterone concentrations and aldosterone-to-renin ratios are highly variable in patients with primary aldosteronism, with many screening values falling below conventionally accepted diagnostic thresholds. The diagnostic yield for primary aldosteronism may be substantially increased by recalibrating the definition of a positive screen to include more liberal thresholds for aldosterone and the aldosterone-to-renin ratio.

PURPOSE: To determine if the presence of a dark cortical rim around the ovary on magnetic resonance imaging (MRI) is associated with polycystic ovarian syndrome (PCOS).

MATERIALS AND METHODS: This retrospective study included 52 PCOS patients with 98 total ovaries and 52 age-matched controls with 104 total ovaries. The ovaries were evaluated on MRI with at least two orthogonal views on T2-weighted sequences. Ovarian volume and follicular count per ovary were measured. Each ovary was also assessed for a dark cortical rim around the ovary on T2-weighted imaging which involved equal to or more than 50% of the ovarian circumference. The degree of rim continuity was classified as continuous if the rim involved greater than 75% of the ovarian circumference, discontinuous if 50-75% of the ovarian circumference was covered, or absent if less than 50% of the ovarian circumference was involved. The rim thickness was measured if present. T test and χ2 tests were performed to compare continuous and categorical variables, correspondingly, between cases and controls. ROC curves and area under the curve (AUC) were used to assess predictive performance and DeLong's paired test was used to compare AUCs.

RESULTS: A higher percentage of PCOS patients exhibited a continuous cortical rim about the ovary (71%) and a lower percentage of an absent cortical rim (8%) compared to controls (25% and 37%, respectively) (p < 0.001). A continuous cortical ovarian rim has a sensitivity and specificity of 71% and 75%, respectively, for diagnosing PCOS. Mean cortical rim thickness is significantly higher in the PCOS group (1.4 mm) compared with controls (0.8 mm) (p < 0.001). Cortical rim thickness and presence of a continuous cortical rim are strongly correlated. Cortical rim thickness of 1.2 mm provides a sensitivity and specificity of 75% and 60%, correspondingly, for a diagnosis of PCOS. Cortical rim thickness combined with cortical rim continuity has an AUC of 0.77 for diagnosing PCOS, which is similar to conventional imaging features of ovarian volume and follicular count combined.

CONCLUSION: A dark cortical rim around the ovary is an MRI feature that can be used to support a diagnosis of PCOS.

OBJECTIVES: To determine the feasibility and safety of ultrasound-guided minimally invasive autopsy in COVID-19 patients.

METHODS: 60 patients who expired between 04/22/2020-05/06/2020 due to COVID-19 were considered for inclusion in the study, based on availability of study staff. Minimally invasive ultrasound-guided autopsy was performed with 14G core biopsies through a 13G coaxial needle. The protocol required 20 cores of the liver, 30 of lung, 12 of spleen, 20 of heart, 20 of kidney, 4 of breast, 4 of testis, 2 of skeletal muscle, and 4 of fat with total of 112 cores per patient. Quality of the samples was evaluated by number, size, histology, immunohistochemistry, and in situ hybridization for COVID-19 and PCR-measured viral loads for SARS-CoV-2.

RESULTS: Five (5/60, 8%) patients were included. All approached families gave their consent for the minimally invasive autopsy. All organs for biopsy were successfully targeted with ultrasound guidance obtaining all required samples, apart from 2 patients where renal samples were not obtained due to atrophic kidneys. The number, size, and weight of the tissue cores met expectation of the research group and tissue histology quality was excellent. Pathology findings were concordant with previously reported autopsy findings for COVID-19. Highest SARS-CoV-2 viral load was detected in the lung, liver, and spleen that had small to moderate amount, and low viral load in was detected in the heart in 2/5 (40%). No virus was detected in the kidney (0/3, 0%).

CONCLUSIONS: Ultrasound-guided percutaneous post-mortem core biopsies can safely provide adequate tissue. Highest SARS-CoV-2 viral load was seen in the lung, followed by liver and spleen with small amount in the myocardium.

OBJECTIVE: To review the technical feasibility of resin microsphere (SIR-Spheres®) yttrium-90 radioembolization prescribed using the medical internal radiation dose (MIRD) model.

METHODS: All radioembolization procedures for hepatic malignancies using resin microspheres with MIRD model between November 2015 and February 2019 were included in this IRB-approved study (n = 60). Student's T test was used to compare prescribed activity based on MIRD and BSA models. Adverse events were assessed immediately, 30 days, and 6 months post-treatment.

RESULTS: Sixty radioembolizations were performed in 54 patients (age 68 ± 9 years, 48-87 years, 35% female). Mean prescribed activity calculated by the MIRD model (target absorbed dose 120-200 Gy for primary and 80-200 Gy for metastatic liver cancers) was 1.7 GBq (0.3-6.4) compared with 0.6 GBq (0.12-2.1) if BSA had been used (p < 0.0001). The prescribed activity was successfully delivered in 93% (56/60) treatments. Prophylactic embolization and anti-reflux catheters were used in 20% (12/60) and 5% (3/60) treatments, respectively. No immediate post-procedural complications occurred. Abdominal pain was the most common clinical Grade 3 CTCAE in 30 days (10%) and 6 months (12%). Radiation pneumonitis occurred in 3 (5%) patients but no radiation-induced gastric ulcer or radiation-induced liver disease occurred.

CONCLUSION: MIRD dosimetry results in higher prescribed activity compared with BSA dosimetry with resin microspheres. MIRD prescribed activity with target absorbed doses up to 200 Gy can be successfully administered without prophylactic embolization in selected patients.

KEY POINTS: •MIRD dosimetry results in higher prescribed activity compared with BSA dosimetry for radioembolization. •MIRD dosimetry can be used for yttrium-90 resin microsphere radioembolization with acceptable safety profile.

OBJECTIVE. Diagnostic accuracy of core needle biopsy (CNB) for adipocytic tumors can be low because of sampling error from these often large, heterogeneous lesions. The purpose of this study was to evaluate the diagnostic accuracy of image-guided CNB for various adipocytic tumors in comparison with excisional pathology. MATERIALS AND METHODS. Adipocytic tumors (n = 77) of all adult patients undergoing image-guided CNB and subsequent surgical excision of an adipocytic tumor at a tertiary referral center between 2005 and 2019 were studied. To determine concordance, we compared pathologic diagnoses based on CNB to the reference standard of pathologic diagnoses after surgical excision. Tumors were divided into three categories (benign lipomatous tumors [lipoma, lipoma variants, hibernomas], atypical lipomatous tumors [ALTs] or well-differentiated liposarcomas [WDLs], and higher grade liposarcomas [myxoid, dedifferentiated, pleomorphic]), and diagnostic accuracy was calculated for each category. RESULTS. In 73 of 77 adipocytic tumors (95%), diagnosis at CNB and diagnosis after excision were concordant. Accuracy of diagnosis was poorer for ALTs and WDLs than for the other two categories, and the difference was statistically significant (p < .002). For the 29 benign lipomatous tumors and the 27 higher-grade liposarcomas, diagnoses at CNB and after excision were concordant in all cases (100%). Seventeen of the 21 tumors (81%) diagnosed as ALTs or WDLs at CNB had a concordant diagnosis after excision; four of the 21 were upgraded (dedifferentiated liposarcoma, n = 3; myxoid liposarcoma, n = 1). CONCLUSION. CNB provides high diagnostic accuracy for adipocytic tumors, particularly for benign lipomatous tumors and higher grade liposarcomas. However, though still high at 81%, diagnostic accuracy of CNB is not as high for tumors diagnosed as ALTs or WDLs. Awareness of this limitation is important when determining management, particularly of cases of ALT or WDL for which surgery is not planned.

PURPOSE: While systemic tumor-stimulating effects can occur following ablation of normal liver linked to the IL-6/HGF/VEGF cytokinetic pathway, the potential for tumor cells themselves to produce these unwanted effects is currently unknown. Here, we study whether partially treated tumors induce increased tumor growth post-radiofrequency thermal ablation (RFA).

METHODS: Tumor growth was measured in three immunocompetent, syngeneic tumor models following partial RFA of the target tumor (in subcutaneous CT26 and MC38 mouse colorectal adenocarcinoma, N = 14 each); and in a distant untreated tumor following partial RFA of target subcutaneous R3230 rat breast adenocarcinoma (N = 12). Tumor cell proliferation (ki-67) and microvascular density (CD34) was assessed. In R3230 tumors, in vivo mechanism of action was assessed following partial RFA by measuring IL-6, HGF, and VEGF expression (ELISA) and c-Met protein (Western blot). Finally, RFA was performed in R3230 tumors with adjuvant c-Met kinase inhibitor or VEGF receptor inhibitor (at 3 days post-RFA, N = 3/arm, total N = 12).

RESULTS: RFA stimulated tumor growth in vivo in residual, incompletely treated surrounding CT26 and MC38 tumor at 3-6 days (p < 0.01). In R3230, RFA increased tumor growth in distant tumor 7 days post treatment compared to controls (p < 0.001). For all models, Ki-67 and CD34 were elevated (p < 0.01, all comparisons). IL-6, HGF, and VEGF were also upregulated post incomplete tumor RFA (p < 0.01). These markers were suppressed to baseline levels with adjuvant c-MET kinase or VEGF receptor inhibition.

CONCLUSION: Incomplete RFA of a target tumor can sufficiently stimulate residual tumor cells to induce accelerated growth of distant tumors via the IL-6/c-Met/HGF pathway and VEGF production.

Artificial intelligence (AI) models for decision support have been developed for clinical settings such as radiology, but little work evaluates the potential impact of such systems. In this study, physicians received chest X-rays and diagnostic advice, some of which was inaccurate, and were asked to evaluate advice quality and make diagnoses. All advice was generated by human experts, but some was labeled as coming from an AI system. As a group, radiologists rated advice as lower quality when it appeared to come from an AI system; physicians with less task-expertise did not. Diagnostic accuracy was significantly worse when participants received inaccurate advice, regardless of the purported source. This work raises important considerations for how advice, AI and non-AI, should be deployed in clinical environments.