Faculty Bibliography

Head computed tomography (CT) imaging is a widely used imaging modality with multitudes of medical indications, particularly in assessing pathology of the brain, skull and cerebrovascular system. It is commonly used as the first-line imaging in neurologic emergencies given its rapidity of image acquisition, safety, cost and ubiquity. Deep learning models may facilitate detection of a wide range of diseases. However, the scarcity of high-quality labels and annotations, particularly among less common conditions, substantially hinders the development of powerful models. To address this challenge, we introduce FM-HCT, a Foundation Model for Head CT for generalizable disease detection, trained using self-supervised learning. Our approach pretrains a deep learning model on a large, diverse dataset of 361,663 non-contrast 3D head CT scans without the need for manual annotations, enabling the model to learn robust, generalizable features. Our results demonstrate that the self-supervised foundation model substantially improves performance on downstream diagnostic tasks compared to models trained from scratch and previous 3D CT foundation models trained on scarce annotated datasets.

BACKGROUND AND PURPOSE/OBJECTIVE:Photon-counting CT (PCCT) offers several advantages over conventional CT for cochlear implant (CI) imaging, including improved spatial resolution, and both signal-to-noise and contrast-to-noise ratios. However, the optimal PCCT reconstruction parameters for CI imaging has not been established. This study compared six PCCT reconstruction approaches for temporal bone CI imaging in a multi-reader design. MATERIALS AND METHODS/METHODS:20 patients with CIs (24 implants) underwent temporal bone PCCT on a NAEOTOM Alpha scanner. Raw data was reconstructed using six different algorithms, as follows: Hr84 0.2mm T3D (head-regular, sharpness level 84, polyenergetic), Hr84 0.4mm T3D, Qr56 0.4mm iMAR T3D (quantitative-regular, sharpness level 56, iterative metal artifact reduction), Qr76 0.4mm M_140 (virtual monoenergetic 140 keV), Qr76 0.4mm M_70 (virtual monoenergetic 70 keV), and Qr76 0.4mm T3D.Two fellowship-trained neuroradiologists independently rated electrode visibility and overall image quality for all implants, and wire visibility for implants with visible wires, on 0-2 Likert scales. Inter-reader agreement was assessed with quadratic weighted Cohen's kappa. A mixed effects model was used to evaluate reconstruction differences for each metric. RESULTS:Mean patient age was 50.9 ± 26 years; 8 were women. Inter-reader agreement was substantial for electrode visibility (κ = 0.66) and overall image quality (κ = 0.79), and moderate for wire visibility (κ = 0.52). Reconstruction type significantly affected all three metrics. The sharp kernel reconstructions (Hr84 0.2mm T3D and Hr84 0.4mm T3D) consistently ranked highest, with significantly greater scores than most other reconstructions. Qr56 0.4mm iMAR T3D was the lowest rated in every category, significantly worse than all other reconstructions. CONCLUSIONS:PCCT reconstruction parameters substantially influence postoperative CI image quality. Ultra-high-resolution reconstructions provided the best combination of artifact suppression and fine structural detail, while iterative MAR and high-keV monoenergetic reconstructions performed the worst. These findings can guide reconstruction selection to optimize PCCT protocols for CI evaluation.

Dual-energy computed tomography (DECT) has emerged as a transformative tool in the past decade. Initially employed in gout within the field of rheumatology to distinguish and quantify monosodium urate crystals through its dual-material discrimination capability, DECT has since broadened its clinical applications. It now encompasses various rheumatic diseases, employing advanced techniques such as bone marrow edema assessment, iodine mapping, and collagen-specific imaging. This review article aims to examine the unique characteristics of DECT, discuss its strengths and limitations, illustrate its applications for accurately evaluating various rheumatic diseases in clinical practice, and propose future directions for DECT in rheumatology.

OBJECTIVE:To determine the optimal virtual monoenergy for viewing the bowel at photon-counting CT enterography using quantitative assessment of mural attenuation, contrast-to-noise ratio, signal-to-noise ratio and noise. METHODS:This study was institutional review board approved and Health Insurance Portability and Accountability Act compliant. Consecutive adults (≥ 18 years) who underwent photon-counting CT enterography from 5/1/2022-5/31/2022 with available Spectral Postprocessing (SPP) images for retrospective virtual monoenergy creation were identified. Nine virtual monoenergetic series (40-120 keV, 10 keV increments) were created. Two region-of-interest measurements were placed in the stomach wall, jejunum wall, ileum wall, and each psoas muscle by two radiologists on 0.6 mm images in PACS. Region-of-interests were copied to other virtual monoenergies to ensure identical placement and size. Attenuation (HU) and noise (HU standard deviation) were recorded from each region-of-interest. Signal-to-noise ratio and contrast-to-noise ratio were computed for stomach, jejunum, ileum, and all bowel combined. Pairwise comparisons for attenuation, noise, signal-to-noise ratio and contrast-to-noise ratio for each virtual monoenergy were performed with ANOVA. A p <.05 indicated statistical significance. RESULTS:50 patients (32 female; mean[SD] age: 57 years) were included. Attenuation and noise for all bowel regions were highest at 40 keV with statistically significant pairwise comparisons from 40 to 70 keV (all p <.05), but similar for 70-120 keV (all p >.05). Signal-to-noise ratio was similar from 40 to 70 keV (all p >.05) for all bowel regions. Contrast-to-noise ratio decreased with increasing keV. Contrast-to-noise ratio was similar for all bowel at 40 keV and 50 keV (p =.06), for stomach from 40 to 70 keV (all p >.05), for jejunum from 40 to 50 keV (p =.21), and for ileum from 40 to 60 keV (all p >.05). CONCLUSION/CONCLUSIONS:50 keV virtual monoenergetic images from photon-counting CT enterography optimizes contrast-to-noise ratio while mitigating noise and should routinely be utilized for bowel assessment at photon-counting CT enterography. As most photon-counting CT users primarily interpret virtual monoenergetic images in clinical practice, knowledge of the optimal virtual monoenergy can inform protocol development.

PURPOSE/OBJECTIVE:As the pancreas is a low contrast visibility organ, pancreatic ductal adenocarcinoma detection is challenging due to subtle attenuation differences between tumor and pancreatic parenchyma. Photon counting CT (PCCT) has superior iodine contrast-to-noise ratio than conventional CT and also affords the creation of low keV virtual monoenergetic images, both of which increase adenocarcinoma conspicuity. The purpose therefore was to identify the optimal virtual monoenergy for visualizing PDAC during the pancreatic parenchymal phase of enhancement at PCCT using both quantitative and qualitative analyses. METHODS:Consecutive patients with pancreatic parenchymal phase PCCT source data were retrospectively identified by PACS search. For the quantitative analysis, region of interest (ROI) measurements were drawn in the pancreatic head, body, tail, pancreatic adenocarcinoma (if present), and psoas muscles on 40-120 keV virtual monoenergetic images in 10 keV increments. Based on the quantitative analysis results and vendor recommendations, four virtual monoenergies(40 keV, 55 keV, 70 keV, and 85 keV) were selected for additional qualitative analysis. Three radiologists blinded to four virtual monoenergies assessed overall image quality, image noise, pancreatic enhancement, and pancreatic mass conspicuity on 5-point Likert scales. RESULTS:54 patients (28/54 male, mean[SD] age: 62 [13] years) were included. Quantitatively, 40 keV had the highest pancreatic parenchymal CNR and attenuation difference between the adenocarcinoma and parenchyma, but also the highest noise (HUsd). Qualitatively, 70 keV had the best overall image quality (Mean [SE]: 3.7[0.1]) and lower noise than 40 and 55 keV (3.6[0.08] vs. 1.8[0.07] and 2.7[0.05], respectively, p < .001). 40 keV had the greatest pancreatic enhancement (mean[SE] 4.6[0.11]). Adenocarcinoma conspicuity ratings were greatest at 40 keV and 55 keV, and not significantly different from each other (mean[SE] 4.4[0.13] and 4.3[0.14], respectively, Tukey adj-p =.20). 55 keV had greater overall image quality and lower noise than 40 keV (mean[SE] 3.4[0.08] vs. 2.5[0.08], Tukey adj-p < .001 and 2.7[0.05] vs. 1.8[0.07], Tukey adj-p < .001 respectively). CONCLUSION/CONCLUSIONS:55 keV pancreatic parenchymal phase virtual monoenergetic images afford optimal pancreatic assessment at PCCT for the visualization of pancreatic adenocarcinoma. Routinely viewing 55 keV virtual monoenergetic images at PCCT may improve PDAC detection.

PURPOSE/OBJECTIVE:Retrospectively compare image quality, radiologist diagnostic confidence, and time for images to reach PACS for contrast enhanced abdominopelvic CT examinations created on the scanner console by technologists versus those generated automatically by thin-client artificial intelligence (AI) mechanisms. METHODS:A retrospective PACS search identified adults who underwent an emergency department contrast-enhanced abdominopelvic CT in 07/2022 (Console Cohort) and 07/2023 (Server Cohort). Coronal and sagittal multiplanar reformatted images (MPR) were created by AI software in the Server cohort. Time to completion of MPR images was compared using 2-sample t-tests for all patients in both cohorts. Two radiologists qualitatively assessed image quality and diagnostic confidence on 5-point Likert scales for 50 consecutive examinations from each cohort. Additionally, they assessed for acute abdominopelvic findings. Continuous variables and qualitative scores were compared with the Mann-Whitney U test. A p < .05 indicated statistical significance. RESULTS:Mean[SD] time to exam completion in PACS was 8.7[11.1] minutes in the Console cohort (n = 728) and 4.6[6.6] minutes in the Server cohort (n = 892), p < .001. 50 examinations in the Console Cohort (28 women 22 men, 51[19] years) and Server cohort (27 women 23 men, 57[19] years) were included for radiologist review. Age, sex, CTDlvol, and DLP were not statistically different between the cohorts (all p > .05). There was no significant difference in image quality or diagnostic confidence for either reader when comparing the Console and Server cohorts (all p > .05). CONCLUSION/CONCLUSIONS:Examinations utilizing AI generated MPRs on a thin-client architecture were completed approximately 50% faster than those utilizing reconstructions generated at the console with no statistical difference in diagnostic confidence or image quality.

PURPOSE/OBJECTIVE:To determine the performance of volumetric dual energy low kV and iodine radiomic features for the differentiation of intrathoracic lymph node histopathology, and influence of contrast protocol. MATERIALS AND METHODS/METHODS:Intrathoracic lymph nodes with histopathologic correlation (neoplastic, granulomatous sarcoid, benign) within 90 days of DECT chest imaging were volumetrically segmented. 1691 volumetric radiomic features were extracted from iodine maps and low-kV images, totaling 3382 features. Univariate analysis was performed using 2-sample t-test and filtered for false discoveries. Multivariable analysis was used to compute AUCs for lymph node classification tasks. RESULTS:129 lymph nodes from 72 individuals (mean age 61 ± 15 years) were included, 52 neoplastic, 51 benign, and 26 granulomatous-sarcoid. Among all contrast enhanced DECT protocol exams (routine, PE and CTA), univariable analysis demonstrated no significant differences in iodine and low kV features between neoplastic and non-neoplastic lymph nodes; in the subset of neoplastic versus benign lymph nodes with routine DECT protocol, 199 features differed (p = .01- < 0.05). Multivariable analysis using both iodine and low kV features yielded AUCs >0.8 for differentiating neoplastic from non-neoplastic lymph nodes (AUC 0.86), including subsets of neoplastic from granulomatous (AUC 0.86) and neoplastic from benign (AUC 0.9) lymph nodes, among all contrast protocols. CONCLUSIONS:Volumetric DECT radiomic features demonstrate strong collective performance in differentiation of neoplastic from non-neoplastic intrathoracic lymph nodes, and are influenced by contrast protocol.

Dual-energy computed tomography (DECT) has emerged as a transformative tool in the past decade. Initially employed in gout within the field of rheumatology to distinguish and quantify monosodium urate crystals through its dual-material discrimination capability, DECT has since broadened its clinical applications. It now encompasses various rheumatic diseases, employing advanced techniques such as bone marrow edema assessment, iodine mapping, and collagen-specific imaging. This review article aims to examine the unique characteristics of DECT, discuss its strengths and limitations, illustrate its applications for accurately evaluating various rheumatic diseases in clinical practice, and propose future directions for DECT in rheumatology.