Dave, Jaydev

Purpose: To measure the computed tomography (CT) attenuation of acute pulmonary embolism (PE) and establish an upper Hounsfield unit (HU) threshold above which it is unlikely to occur. Materials and Methods: This retrospective study analyzed CT pulmonary angiography (CTPA) exams in patients with acute PE (2018–2023) who had a prior negative CTPA in the preceding 2 weeks. Clot and artifact attenuation were measured on the positive and prior negative CTPA respectively. Analysis of variance with post hoc tests and linear regression analysis were performed to determine the effects of technical and other parameters on clot and artifact attenuation. Results: A total of 60 patients were included, each with 1 record or a pair of CTPA exams, yielding 105 clots and 41 artifacts. Mean attenuation for acute PE was 50±40 HU, and for artifact was 110±146 HU. A threshold of 125 HU achieved 97% (102/105) sensitivity for acute PE below it and the same specificity for artifact above it. Specificity for acute PE below this threshold was 37% (15/41). Subsegmental clots had a lower attenuation than all other clots as a group. Kilovoltage peak (kVp) and proximal vessel attenuation were poor predictors of clot attenuation . Conclusions: Most acute PE demonstrated a CT attenuation ≤125 HU on standard CTPA, an upper threshold higher than previously reported, while artifacts demonstrated both lower and higher attenuation. An attenuation >125 HU on standard CTPA, favoring artifact rather than acute PE, may be a useful adjunct to morphologic differentiation between these entities.

Elevated intratumoral interstitial fluid pressure (IFP) can negatively impact the efficacy of neoadjuvant chemotherapy (NAC) of breast cancer. This study aims to predict breast cancer pathological complete response (pCR) to NAC based on IFP measures by 3-D subharmonic-aided pressure estimation (SHAPE). SHAPE uses the inverse linear relationship between hydrostatic pressure and subharmonic signal magnitude from intravascular ultrasound contrast agents. Thirty-eight breast cancer patients undergoing NAC from three clinical sites were enrolled for this study. SHAPE data were collected for each subject at 0%, 10%, and 30% of NAC completion using a Logiq E9 system (GE HeathCare) equipped with an RSP6-16D probe. The contrast agent used was Definity (Lantheus Medical Imaging). The subharmonic signal gradients between tumor and adjacent normal tissue were calculated and compared with patient pathological response outcome using t-test and receiver operating characteristic (ROC) curve analysis. Pressure gradients between pCR (15 patients) and non-pCR (23 patients) groups were significantly different at 0 (p = 0.02), 10 (p = 0.03), and 30% (p = 0.0003) completion of NAC. Area under the curve values of 0.69, 0.77, and 0.86 were achieved at 0%, 10%, and 30% of NAC completion, respectively. Overall, SHAPE pressure gradients may aid prediction of NAC pathological response in breast cancer patients.

Objectives Uniformity artifacts caused by defective transducer elements or scanner malfunctions degrade diagnostic image quality. Traditional quality control (QC) methods, such as phantom testing and visual or quantitative image analysis, can be labor‐intensive and limited in test frequency. This study aims to develop a deep learning framework to detect uniformity artifacts and complement traditional QC. Methods Clinical median images were generated by aggregating co‐registered grayscale ultrasound images acquired by each transducer and computing median intensity values across the image stack. A pretrained ResNet‐18 model was fine‐tuned on a dataset consisting of clinical median images from linear and curvilinear transducers. The dataset was divided into training, validation, and testing subsets, ensuring no overlap between training and test transducers. To assess generalizability, the model was also evaluated on an independent test set of 396 phantom‐validated images from linear and curvilinear transducers. Results The model achieved 100% accuracy on the first dataset's test set. On the independent test set, it attained 87.4% accuracy with high sensitivity (0.84) and specificity (0.88), demonstrating robust generalization. Occlusion sensitivity maps confirmed the model's attention to uniformity artifact regions. Conclusion The deep learning framework using clinical median images demonstrated robust performance across several linear and curvilinear transducer models. It could be integrated into the clinical QC workflow by automating artifact detection in an effective and timely manner. In our practice, it can flag median images classified as artifact‐present, reducing human review time by approximately 80% while preserving detection accuracy.