Search

Objective. Unrolled algorithms are a promising approach for reconstruction of CT images in challenging scenarios, such as low-dose, sparse-view and limited-angle imaging. In an unrolled algorithm, a fixed number of iterations of a reconstruction method are unrolled into multiple layers of a neural network, and interspersed with trainable layers. The entire network is then trained end-to-end in a supervised fashion, to learn an appropriate regularizer from training data. In this paper we propose a novel unrolled algorithm, and compare its performance with several other approaches on sparse-view and limited-angle CT. Approach. The proposed algorithm is inspired by the superiorization methodology, an optimization heuristic in which iterates of a feasibility-seeking method are perturbed between iterations, typically using descent directions of a model-based penalty function. Our algorithm instead uses a modified U-net architecture to introduce the perturbations, allowing a network to learn beneficial perturbations to the image at various stages of the reconstruction, based on the training data. Main Results. In several numerical experiments modeling sparse-view and limited angle CT scenarios, the algorithm provides excellent results. In particular, it outperforms several competing unrolled methods in limited-angle scenarios, while providing comparable or better performance on sparse-view scenarios. Significance. This work represents a first step towards exploiting the power of deep learning within the superiorization methodology. Additionally, it studies the effect of network architecture on the performance of unrolled methods, as well as the effectiveness of the unrolled approach on both limited-angle CT, where previous studies have primarily focused on the sparse-view and low-dose cases., (© 2022 Institute of Physics and Engineering in Medicine.)

Purpose: We investigated whether the additional use of grey-scale inversion technique improves the interpretation of eight chest abnormalities, in terms of diagnostic performance and interobserver variability., Material and Methods: A total of 507 patients who underwent a chest computed tomography (CT) examination and a chest radiography (CXR) within 24 h were enrolled. CT was the standard of reference. Images were retrospectively reviewed for the presence of atelectasis, consolidation, interstitial abnormality, nodule, mass, pleural effusion, pneumothorax and rib fractures. Four CXR reading settings, involving 3 readers were organized: only standard; only inverted; standard followed by inverted; and inverted followed by standard. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy, assessed with the area under the curve (AUC), and their 95% confidence interval were calculated for each reader and setting. Interobserver agreement was tested by Cohen's K test with quadratic weights (K w ) and its 95%CI., Results: CXR sensitivity % for any finding was 35.1 (95% CI: 33 to 37) for setting 1, 35.9 (95% CI: 33 to 37), for setting 2, 32.59 (95% CI: 30 to 34) for setting 3, and 35.56 (95% CI: 33 to 37) for setting 4; specificity % 93.78 (95% CI: 91 to 95), 93.92 (95% CI: 91 to 95), 94.43 (95% CI: 92 to 96), 93.86 (95% CI: 91 to 95); PPV % 56.22 (95% CI: 54.2 to 58.2), 56.49 (95% CI: 54.5 to 58.5), 57.15 (95% CI: 55 to 59), 56.75 (95% CI: 54 to 58); NPV % 85.66 (95% CI: 83 to 87), 85.74 (95% CI: 83 to 87), 85.29 (95% CI: 83 to 87), 85.73 (95% CI: 83 to 87); AUC values 0.64 (95% CI: 0.62 to 0.66), 0.65 (95% CI: 0.63 to 0.67), 0.64 (95% CI: 0.62 to 0.66), 0.65 (95% CI: 0.63 to 0.67); K w values 0.42 (95% CI: 0.4 to 0.44), 0.40 (95% CI: 0.38 to 0.42), 0.42 (95% CI: 0.4 to 0.44), 0.41 (95% CI: 0.39 to 0.43) for settings 1, 2, 3 and 4, respectively., Conclusions: No significant advantages were observed in the use of grey-scale inversion technique neither over standard display mode nor in combination at the detection of eight chest abnormalities., (© 2022. The Author(s).)

Decreased expression of Numb, resulting in activation of the proto-oncogene Notch1 and reduction in the tumor suppressor p53, has been demonstrated in mammary carcinomas. The aim of this study was to investigate the relationship between Numb protein expression and clinicopathological characteristics, tumor biological subtypes and putative cancer stem cell markers in a well-characterized cohort of primary human breast cancers. Immunohistochemistry was performed on tissue microarrays of primary invasive breast tumors using a polyclonal anti-Numb primary antibody. Of the 241 tumors evaluated, 50 (21%) displayed deficient or reduced Numb immunoreactivity. Retained Numb expression was significantly correlated to estrogen (ER) and progesterone receptor (PR) positivity (P < 0.001 and P = 0.004, respectively). Interestingly, we found that a higher percentage of the tumors with deficient or reduced Numb expression belonged to the triple-negative (ER-/PR-/HER2-) subgroup compared to tumors with retained Numb expression (P = 0.004). Transcriptional profiling of a subset of these tumors linked NOTCH1 and BIRC5, both downstream targets of Numb, to the triple-negative subgroup in an inverse manner. Typically, subgroups characterized by the low expression of Numb expressed higher levels of NOTCH1 and BIRC5 (encoding survivin). We also found deficient expression of Numb in a significantly higher proportion of BRCA1 dependent tumors, which are usually triple-negative, compared to sporadic tumors. The expression of Numb in 14 breast cancer cell lines correlated similarly to their respective molecular subtypes. We further established an inverse correlation between the Numb expression levels and the CD44+/CD24- cancer stem cell phenotype (P = 0.05) in primary tumors. Finally, decreased Numb expression was associated with poorer distant disease-free survival (P = 0.01). Taken together, our results indicate that loss of Numb expression is a marker of tumor aggressiveness, potentially linked to BRCA1 status and a cancer stem cell phenotype in primary breast cancer.