Comparing accuracy of machine learning approaches to identifying parathyroid adenomas: Lessons and new directions.

PURPOSE: The purpose of this investigation is to understand the accuracy of machine learning techniques to detect biopsy-proven adenomas from similar appearing lymph nodes and factors that influence accuracy by comparing support vector machine (SVM) and bidirectional Long short-term memory (Bi-LSTM) analyses. This will provide greater insight into how these tools could integrate multidimensional data and aid the detection of parathyroid adenomas consistently and accurately. METHODS: Ninety-nine patients were identified; 93 4D-CTs of patients with pathology-proven parathyroid adenomas were reviewed; 94 parathyroid adenomas and 112 lymph nodes were analyzed. A 2D slice through the lesions in each phase was used to perform sequence classification with ResNet50 as the pre-trained network to construct the Bi-LSTM model, and the mean enhancement curves were used to form an SVM model. The model characteristics and accuracy were calculated for the training and validation data sets. RESULTS: On the training data, the area under the curve (AUC) of the Bi-LSTM was 0.99, while the SVM was 0.95 and statistically significant on the DeLong test. The overall accuracy of the Bi-LSTM on the validation data set was 92 %, while the SVM was 88 %. The accuracy for parathyroid adenomas specifically was 93 % for the Bi-LSTM and 83 % for the SVM model. CONCLUSION: Enhancement characteristics are a distinguishing feature that accurately identifies parathyroid adenomas alone. The Bi-LSTM performs statistically better in identifying parathyroid adenomas than the SVM analysis when using both morphologic and enhancement information to distinguish between parathyroid adenomas and lymph nodes. SUMMARY STATEMENT: The Bi-LSTM more accurately identifies parathyroid adenomas than the SVM analysis, which uses both morphologic and enhancement information to distinguish between parathyroid adenomas and lymph nodes, performs statistically better.

Pretreatment CT perfusion collateral parameters correlate with penumbra salvage in middle cerebral artery occlusion.

BACKGROUND AND PURPOSE: Acute ischemic stroke due to large vessel occlusion (AIS-LVO) is a major cause of functional dependence. Collateral status (CS) is an important determinant of functional outcomes. Pretreatment CT perfusion (CTP) parameters serve as reliable surrogates of CS. Penumbra Salvage Index (PSI) is another parameter predictive of functional outcomes in AIS-LVO. The aim of this study is to assess the relationship of pretreatment CTP parameters with PSI. METHODS: In this prospectively collected, retrospectively reviewed multicenter analysis, inclusion criteria were as follows: (1) CT angiography confirmed middle cerebral artery (MCA) M1-segment and proximal M2-segment occlusion from 9/1/2017 to 9/22/2022; (2) diagnostic CTP; and (3) available diagnostic Magnetic resonance Imaging (MRI) diffusion-weighted images. Pearson correlation analysis was performed to assess the association between cerebral blood volume (CBV) index and hypoperfusion intensity ratio (HIR) with PSI. p value ≤.05 was considered statistically significant. RESULTS: In total, 131 patients (n = 86, M1 and n = 45, proximal M2 occlusion) met our inclusion criteria. CBV index showed a modest positive correlation with PSI (r = 0.34, p<.001) in patients with proximal MCA occlusion. Similar trends were noted in subgroup analysis of patients with M1 occlusion, and proximal M2 occlusion. Whereas, HIR did not have a strong trend or correlation with PSI. CONCLUSION: CBV index correlates with PSI, whereas HIR does not. Future studies are needed to expand our understanding of the adjunct role of CBV index with other similar pretreatment CTP-based markers in clinical evaluation and decision-making in patients with MCA occlusion.

Redefining CT perfusion-based ischemic core estimates for the ghost core in early time window stroke.

BACKGROUND AND PURPOSE: In large vessel occlusion (LVO) stroke patients, relative cerebral blood flow (rCBF)<30% volume thresholds are commonly used in treatment decisions. In the early time window, nearly infarcted but salvageable tissue volumes may lead to pretreatment overestimates of infarct volume, and thus potentially exclude patients who may otherwise benefit from intervention. Our multisite analysis aims to explore the strength of relationships between widely used pretreatment CT parameters and clinical outcomes for early window stroke patients. METHODS: Patients from two sites in a prospective registry were analyzed. Patients with LVOs, presenting within 3 hours of last known well, and who were successfully reperfused were included. Primary short-term neurological outcome was percent National Institutes of Health Stroke Scale (NIHSS) change from admission to discharge. Secondary long-term outcome was 90-day modified Rankin score. Spearman's correlations were performed. Significance was attributed to p-value ≤.05. RESULTS: Among 73 patients, median age was 66 (interquartile range 54-76) years. Among all pretreatment imaging parameters, rCBF<30%, rCBF<34%, and rCBF<38% volumes were significantly, inversely correlated with percentage NIHSS change (p<.048). No other parameters significantly correlated with outcomes. CONCLUSIONS: Our multisite analysis shows that favorable short-term neurological recovery was significantly correlated with rCBF volumes in the early time window. However, modest strength of correlations provides supportive evidence that the applicability of general ischemic core estimate thresholds in this subpopulation is limited. Our results support future larger-scale efforts to liberalize or reevaluate current rCBF parameter thresholds guiding treatment decisions for early time window stroke patients.

Prolonged venous transit on perfusion imaging is associated with higher odds of mortality in successfully reperfused patients with large vessel occlusion stroke.

BACKGROUND: Poor venous outflow (VO) profiles are associated with unfavorable outcomes in patients with acute ischemic stroke caused by large vessel occlusion (AIS-LVO), despite achieving successful reperfusion. The objective of this study is to assess the association between mortality and prolonged venous transit (PVT), a novel visual qualitative VO marker on CT perfusion (CTP) time to maximum (Tmax) maps. METHODS: We performed a retrospective analysis of prospectively collected data from consecutive adult patients with AIS-LVO with successful reperfusion (modified Thrombolysis in Cerebral Infarction 2b/2c/3). PVT+ was defined as Tmax ≥10 s timing on CTP Tmax maps in at least one of the following: superior sagittal sinus (proximal venous drainage) and/or torcula (deep venous drainage). PVT- was defined as lacking this in both regions. The primary outcome was mortality at 90 days. In a 1:1 propensity score-matched cohort, regressions were performed to determine the effect of PVT on 90-day mortality. RESULTS: In 127 patients of median (IQR) age 71 (64-81) years, mortality occurred in a significantly greater proportion of PVT+ patients than PVT- patients (32.5% vs 12.6%, P=0.01). This significant difference persisted after matching (P=0.03). PVT+ was associated with a significantly increased likelihood of 90-day mortality (OR 1.22 (95% CI 1.02 to 1.46), P=0.03) in the matched cohort. CONCLUSIONS: PVT+ was significantly associated with 90-day mortality despite successful reperfusion therapy in patients with AIS-LVO. PVT is a simple VO profile marker with potential as an adjunctive metric during acute evaluation of AIS-LVO patients. Future studies will expand our understanding of using PVT in the evaluation of patients with AIS-LVO.

The Compensation Index Is Better Associated with DSA ASITN Collateral Score Compared to the Cerebral Blood Volume Index and Hypoperfusion Intensity Ratio.

BACKGROUND: Pretreatment CT Perfusion (CTP) parameters serve as reliable surrogates of collateral status (CS). In this study, we aim to assess the relationship between the novel compensation index (CI, Tmax > 4 s/Tmax > 6 s) and already established CTP collateral markers, namely cerebral blood volume (CBV) index and Hypoperfusion Intensity Ratio (HIR), with the reference standard American Society of Interventional and Therapeutic Neuroradiology (ASITN) collateral score (CS) on DSA. METHODS: In this retrospective study, inclusion criteria were the following: (a) CT angiography confirmed anterior circulation large vessel occlusion from 9 January 2017 to 10 January 2023; (b) diagnostic CT perfusion; and (c) underwent mechanical thrombectomy with documented DSA-CS. Student t-test, Mann-Whitney-U-test and Chi-square test were used to assess differences. Spearman's rank correlation and logistic regression analysis were used to assess associations. p ≤ 0.05 was considered significant. RESULTS: In total, 223 patients (mean age: 67.8 ± 15.8, 56% female) met our inclusion criteria. The CI (ρ = 0.37, p < 0.001) and HIR (ρ = -0.29, p < 0.001) significantly correlated with DSA-CS. Whereas the CBV Index (ρ = 0.1, p > 0.05) did not correlate with DSA-CS. On multivariate logistic regression analysis taking into account age, sex, ASPECTS, tPA, premorbid mRS, NIH stroke scale, prior history of TIA, stroke, atrial fibrillation, diabetes mellitus, hyperlipidemia, heart disease and hypertension, only CI was not found to be independently associated with DSA-CS (adjusted OR = 1.387, 95% CI: 1.09-1.77, p < 0.01). CONCLUSION: CI demonstrates a stronger correlation with DSA-CS compared to both the HIR and CBV Index where it may show promise as an additional quantitative pretreatment CS biomarker.

Cultivation of HEK 293 cell line and production of a member of the superfamily of G-protein coupled receptors for drug discovery applications using a highly efficient novel bioreactor.

A process of producing a receptor in HEK-293 cells used for the drug discovery program at Pfizer Inc. has been successfully developed with a novel BelloCell bioreactor to replace the conventional 2-D cell culturing devices including Cell Factories and roller bottles. A single BelloCell-500 has produced >1.4 x 10(9) HEK-293 cells, which are equivalent to those produced by 12 roller bottles, with substantially easier operation, single inoculation, less inoculum, less medium consumption and better space utilization. The receptor expression levels are better than those obtained by the traditional process. 3.7 pmoles of radioligandY mg(-1) protein were attained in the bioreactor compared to 2.3 pmoles of radioligandY mg(-1) protein in roller bottles. This may be attributed to the three dimensional attachment during cell growth. A 92% cell recovery from the bioreactor has been attained using Acutase or Trypsin treatment followed by four washes. It has been proven to be a viable and efficient device to produce adherent cells and express target components of interest for drug discovery applications.