Comparing Micro-Ultrasound to mpMRI in Detecting Clinically Significant Prostate Cancer

Objectives To compare the performance of micro-ultrasound (mUS) with multi-parametric magnetic resonance imaging (mpMRI) in detecting clinically significant prostate cancer. Materials and Methods Retrospective data from consecutive patients with any indication for prostate biopsy in 2 academic institutions were included. The operator, blinded to mpMRI, would first scan the prostate and annotate any mUS lesions. All mUS lesions were biopsied. Any mpMRI lesions that did not correspond to mUS lesion upon unblinding were additionally biopsied. Grade group (GG) ≥ 2 was considered clinically significant cancer. The Jeffreys interval method was used to compare performance of mUS with mpMRI with the non-inferiority limit set at −5%. Results Imaging and biopsy were performed in 82 patients with 153 lesions. mUS had similar sensitivity to mpMRI (per-lesion analysis: 78.4% versus 72.5%), but lower specificity, positive predictive value, negative predictive value, and area under the curve. Micro-ultrasound found GG ≥ 2 in 13% of cases missed by mpMRI, while mpMRI found GG ≥ 2 in 11% of cases missed by mUS. The difference 0.020 (95% CI −0.070 to 0.110) was not statistically significant (P = 0.33). Conclusion The sensitivity of mUS in detecting GG ≥ 2 disease was similar to that of mpMRI, but the specificity was lower. Further evaluation with a larger sample size and experienced operators is warranted.

Multi-Task Model for Esophageal Lesion Analysis Using Endoscopic Images: Classification with Image Retrieval and Segmentation with Attention

The automatic analysis of endoscopic images to assist endoscopists in accurately identifying the types and locations of esophageal lesions remains a challenge. In this paper, we propose a novel multi-task deep learning model for automatic diagnosis, which does not simply replace the role of endoscopists in decision making, because endoscopists are expected to correct the false results predicted by the diagnosis system if more supporting information is provided. In order to help endoscopists improve the diagnosis accuracy in identifying the types of lesions, an image retrieval module is added in the classification task to provide an additional confidence level of the predicted types of esophageal lesions. In addition, a mutual attention module is added in the segmentation task to improve its performance in determining the locations of esophageal lesions. The proposed model is evaluated and compared with other deep learning models using a dataset of 1003 endoscopic images, including 290 esophageal cancer, 473 esophagitis, and 240 normal. The experimental results show the promising performance of our model with a high accuracy of 96.76% for the classification and a Dice coefficient of 82.47% for the segmentation. Consequently, the proposed multi-task deep learning model can be an effective tool to help endoscopists in judging esophageal lesions.

Weakly-supervised lesion analysis with a CNN-based framework for COVID-19

Lesions of COVID-19 can be visualized clearly by chest CT images, therefore, providing valuable evidence for clinicians when making a diagnosis. However, due to the variety of COVID-19 lesions and the complexity of the manual delineation procedure, automatic analysis of lesions with unknown and diverse types from a CT image remains a challenging task. In this paper we propose a weakly-supervised framework for this task, requiring only a series of normal and abnormal CT images without the need for annotations of the specific locations and types of lesions. Specifically, this framework employs a deep learning-based diagnosis branch for the classification of the CT image and then leverages a lesion identification branch to capture multiple types of lesions. We verify our framework on publicly available datasets and CT data collected from 13 patients of the First Affiliated Hospital of Shantou University Medical College, China. The results show that the proposed framework can achieve state-of-the-art diagnosis prediction, and the extracted lesion features are capable of distinguishing between lesions showing ground glass opacity and consolidation. Further exploration also demonstrates that this framework has the potential to discover lesion types that have not been reported and can potentially be generalized to lesion detection of other chest-based diseases.

819 Radiomic markers associated with clinical benefit in advanced uveal melanoma patients with radiographic progression on tebentafusp

BackgroundTebentafusp, a bispecific fusion protein consisting of affinity-enhanced T cell receptor targeting a gp100 derived peptide fused to anti-CD3 effector, has shown overall survival (OS) benefit in untreated metastatic uveal melanoma (mUM). The OS benefit derives from all RECIST response categories, even progressive disease (PD). In Ph2 trial of previously treated mUM (NCT02570308), one-third (35%) of 48 evaluable patients with best response of PD had ctDNA reduction (³0.5 log reduction) and longer OS (median 16.9 months) compared to the group without ctDNA reduction (median OS 8.5 months).Methods34 of 127 mUM patients from Ph2 trial1 were selected based on best response of PD and no ctDNA reduction (Group A, n=17) or 0.5 log ctDNA reduction (Group B, n=17). One patient per group were excluded due to poor image quality or limited CT/MRI sequences. Tumor lesions were manually segmented on CT and MRI. Radiomics features were extracted at baseline and Week-8 (first assessment). The objective was to use unsupervised machine-learning to develop two signatures using 16 features to classify the two groups. The per-patient analysis signature (n=32) combined 8 volumetric features on CT-scan at baseline and change by Week-8. The per-lesion analysis signature (n=148) combined 4 features (volume and 3 radiomics features previously associated with outcome to checkpoint immunotherapy in cutaneous melanoma) at two timepoints using CT and MRI. Performance was evaluated using area under the receiver operating characteristic curve (AUC).ResultsThe median OS for Groups A and B were 8.5 and 16.9 months, respectively. In the per-patient analysis, a volumetric signature classified patients into the groups with AUC 0.71 (95%CI: 0.53–0.90) with 63% specificity and 81% sensitivity at the optimal threshold (0.57). In the per-lesion analysis, a radiomic signature reached an AUC of 0.70 (95%CI: 0.58–0.81) with 66% specificity and 74% sensitivity at the optimal threshold (0.53). Group B had lower baseline tumor lesion volume (AUC=0.65), distinct baseline tumor heterogeneity (AUC=0.66), and distinct change in tumor heterogeneity by week 8 (AUC = 0.66/0.69 on CT/MRIConclusionsA radiomic analyses of a subset of PD patients was able to predict Group B, patients with ctDNA reduction and longer OS, at a patient and lesion level. The strongest radiomic predictor by CT/MRI was decrease on treatment in tumor heterogeneity. Confirmation in a larger dataset of these signatures is needed to identify which patients may be benefiting from tebentafusp despite radiographic progression.Trial RegistrationNCT02570308Reference1. Sacco JJ, Carvajal R, Butler MO, et al. A phase (ph) II, multi-center study of the safety and efficacy of tebentafusp (tebe) (IMCgp100) in patients (pts) with metastatic uveal melanoma (mUM). Ann Oncol 2020;31:S1442–S1143.Abstract 819 Figure 1Percent change in tumor measurement from baseline at week 8 per independent review committee by Group A and BAbstract 819 Figure 2Kaplan-Meier plot comparing overall survival rates in group A and group B patientsAbstract 819 Figure 3Blue color represents a high probability of the patient being in Group A while red color indicates high probability of being in Group B

Аpplication of artificial intelligence systems in neuroradiology of acute ischemic stroke

Introduction. Artificial intelligence is one of the fastest-growing areas of great importance to radiology. Purpose. In this article, we aimed to study the current state of the use of computer-aided imaging analysis in acute ischemic stroke. Results. There are many artificial intelligence softwares that automatic image processing can successfully identify neuroradiology image in stroke: early detection by diagnostic imaging methods, assessment of the time of disease onset, segmentation of the lesion, analysis of the presence and possibility of cerebral edema, and predicting complications and treatment outcomes. Conclusion. The first results of using artificial intelligence to evaluate neuroimaging data showed that machine-learning methods could be useful as decision-making tools when choosing a treatment for acute ischemic stroke.

Quantitative 7-Tesla Imaging of Cortical Myelin Changes in Early Multiple Sclerosis

Cortical demyelination occurs early in multiple sclerosis (MS) and relates to disease outcome. The brain cortex has endogenous propensity for remyelination as proven from histopathology study. In this study, we aimed at characterizing cortical microstructural abnormalities related to myelin content by applying a novel quantitative MRI technique in early MS. A combined myelin estimation (CME) cortical map was obtained from quantitative 7-Tesla (7T) T2* and T1 acquisitions in 25 patients with early MS and 19 healthy volunteers. Cortical lesions in MS patients were classified based on their myelin content by comparison with CME values in healthy controls as demyelinated, partially demyelinated, or non-demyelinated. At follow-up, we registered changes in cortical lesions as increased, decreased, or stable CME. Vertex-wise analysis compared cortical CME in the normal-appearing cortex in 25 MS patients vs. 19 healthy controls at baseline and investigated longitudinal changes at 1 year in 10 MS patients. Measurements from the neurite orientation dispersion and density imaging (NODDI) diffusion model were obtained to account for cortical neurite/dendrite loss at baseline and follow-up. Finally, CME maps were correlated with clinical metrics. CME was overall low in cortical lesions (p = 0.03) and several normal-appearing cortical areas (p < 0.05) in the absence of NODDI abnormalities. Individual cortical lesion analysis revealed, however, heterogeneous CME patterns from extensive to partial or absent demyelination. At follow-up, CME overall decreased in cortical lesions and non-lesioned cortex, with few areas showing an increase (p < 0.05). Cortical CME maps correlated with processing speed in several areas across the cortex. In conclusion, CME allows detection of cortical microstructural changes related to coexisting demyelination and remyelination since the early phases of MS, and shows to be more sensitive than NODDI and relates to cognitive performance.

The Additive Value of 3D Total Body Imaging for Sequential Monitoring of Skin Lesions: A Case Series

<b><i>Background:</i></b> Timely diagnosis is the cornerstone of melanoma morbidity and mortality reduction. 2D total body photography and dermoscopy are routinely used to assist with early detection of skin malignancies. Polarized 3D total body photography is a novel technique that enables fast image acquisition of almost the entire skin surface. We aimed to determine the added value of 3D total body photography alongside dermoscopy for monitoring cutaneous lesions. <b><i>Methods:</i></b> Lesion images from high-risk individuals were assessed for long-term substantial changes via dermoscopy and 3D total body photography. Three case studies are presented demonstrating how 3D total body photography may enhance lesion analysis alongside traditional dermoscopy. <b><i>Results:</i></b> 3D total body photography can assist clinicians by presenting cutaneous lesions in their skin ecosystem, thereby providing additional clinical context and enabling a more holistic assessment to aid dermoscopy interpretation. For lesion cases where previous dermoscopy is unavailable, corresponding 3D images can substitute for baseline dermoscopy. Additionally, 3D total body photography is not susceptible to artificial stretch artefacts. <b><i>Conclusion:</i></b> 3D total body photography is valuable alongside dermoscopy for monitoring cutaneous lesions. Furthermore, it is capable of surveilling almost the entire skin surface, including areas not traditionally monitored by sequential imaging.

68Ga-PSMA PET/CT and mpMRI for primary lymph node staging of intermediate to high-risk prostate cancer: a systematic review and meta-analysis of diagnostic test accuracy

Purpose To evaluate the diagnostic accuracy of Gallium-68 prostate-specific membrane antigen positron emission tomography-computed tomography (68Ga-PSMA PET/CT) compared with multiparametric magnetic resonance imaging (mpMRI) for detection of metastatic lymph nodes in intermediate to high-risk prostate cancer (PCa). Methods PRISMA-compliant systematic review updated to September 2020 was performed to identify studies that evaluated the diagnostic performance of 68Ga-PSMA PET/CT and mpMRI for detection of metastatic lymph nodes in the same cohort of PCa patients using histopathologic examination as a reference standard. The quality of each study was assessed using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) instrument. STATA version 16.0 was used to obtain the pooled estimates of diagnostic accuracy for per-patient and per-lesion analyses. Heterogeneity in the accuracy estimates was explored by reviewing the generated forest plots, summary receiver operator characteristic (SROC) curves, hierarchical SROC plots, chi-squared test, heterogeneity index, and Spearman’s correlation coefficients. Results Six studies, which included 476 patients, met the eligibility criteria for per-patient analysis and four of these studies, reporting data from 4859 dissected lymph nodes, were included in the per-lesion analysis. In the per-patient analysis (N = 6), the pooled sensitivity and specificity for 68Ga-PSMA PET/CT were 0.69 and 0.93, and for mpMRI the pooled sensitivity and specificity were 0.37 and 0.95. In the per-lesion analysis (N = 4), the pooled sensitivity and specificity for 68Ga-PSMA PET/CT were 0.58 and 0.99, and for mpMRI the pooled sensitivity and specificity were 0.44 and 0.99. There was high heterogeneity and a threshold effect in outcomes. A sensitivity analysis demonstrated that the pooled estimates were stable when excluding studies with patient selection concerns, whereas the variances of the pooled estimates became significant, and the characteristics of heterogeneity changed when excluding studies with concerns about index imaging tests. Conclusion Both imaging techniques have high specificity for the detection of nodal metastases of PCa. 68Ga-PSMA PET/CT has the advantage of being more sensitive and making it possible to detect distant metastases during the same examination. These modalities may play a complementary role in the diagnosis of PCa. Given the paucity of data and methodological limitations of the included studies, large scale trials are necessary to confirm their clinical values.