3D Breast Cancer Segmentation in DCE-MRI Using Deep Learning With Weak Annotation.

BACKGROUND: Deep learning models require large-scale training to perform confidently, but obtaining annotated datasets in medical imaging is challenging. Weak annotation has emerged as a way to save time and effort. PURPOSE: To develop a deep learning model for 3D breast cancer segmentation in dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) using weak annotation with reliable performance. STUDY TYPE: Retrospective. POPULATION: Seven hundred and thirty-six women with breast cancer from a single institution, divided into the development (N = 544) and test dataset (N = 192). FIELD STRENGTH/SEQUENCE: 3.0-T, 3D fat-saturated gradient-echo axial T1-weighted flash 3D volumetric interpolated brain examination (VIBE) sequences. ASSESSMENT: Two radiologists performed a weak annotation of the ground truth using bounding boxes. Based on this, the ground truth annotation was completed through autonomic and manual correction. The deep learning model using 3D U-Net transformer (UNETR) was trained with this annotated dataset. The segmentation results of the test set were analyzed by quantitative and qualitative methods, and the regions were divided into whole breast and region of interest (ROI) within the bounding box. STATISTICAL TESTS: As a quantitative method, we used the Dice similarity coefficient to evaluate the segmentation result. The volume correlation with the ground truth was evaluated with the Spearman correlation coefficient. Qualitatively, three readers independently evaluated the visual score in four scales. A P-value <0.05 was considered statistically significant. RESULTS: The deep learning model we developed achieved a median Dice similarity score of 0.75 and 0.89 for the whole breast and ROI, respectively. The volume correlation coefficient with respect to the ground truth volume was 0.82 and 0.86 for the whole breast and ROI, respectively. The mean visual score, as evaluated by three readers, was 3.4. DATA CONCLUSION: The proposed deep learning model with weak annotation may show good performance for 3D segmentations of breast cancer using DCE-MRI. LEVEL OF EVIDENCE: 3 TECHNICAL EFFICACY: Stage 2.

χ-Separation Imaging for Diagnosis of Multiple Sclerosis versus Neuromyelitis Optica Spectrum Disorder.

Background Use of χ-separation imaging can provide surrogates for iron and myelin that relate closely to abnormal changes in multiple sclerosis (MS) lesions. Purpose To evaluate the appearances of MS and neuromyelitis optica spectrum disorder (NMOSD) brain lesions on χ-separation maps and explore their diagnostic value in differentiating the two diseases in comparison with previously reported diagnostic criteria. Materials and Methods This prospective study included individuals with MS or NMOSD who underwent χ-separation imaging from October 2017 to October 2020. Positive (χpos) and negative (χneg) susceptibility were estimated separately by using local frequency shifts and calculating R2' (R2' = R2* - R2). R2 mapping was performed with a machine learning approach. For each lesion, presence of the central vein sign (CVS) and paramagnetic rim sign (PRS) and signal characteristics on χneg and χpos maps were assessed and compared. For each participant, the proportion of lesions with CVS, PRS, and hypodiamagnetism was calculated. Diagnostic performances were assessed using receiver operating characteristic (ROC) curve analysis. Results A total of 32 participants with MS (mean age, 34 years ± 10 [SD]; 25 women, seven men) and 15 with NMOSD (mean age, 52 years ± 17; 14 women, one man) were evaluated, with a total of 611 MS and 225 NMOSD brain lesions. On the χneg maps, 80.2% (490 of 611) of MS lesions were categorized as hypodiamagnetic versus 13.8% (31 of 225) of NMOSD lesions (P < .001). Lesion appearances on the χpos maps showed no evidence of a difference between the two diseases. In per-participant analysis, participants with MS showed a higher proportion of hypodiamagnetic lesions (83%; IQR, 72-93) than those with NMOSD (6%; IQR, 0-14; P < .001). The proportion of hypodiamagnetic lesions achieved excellent diagnostic performance (area under the ROC curve, 0.96; 95% CI: 0.91, 1.00). Conclusion On χ-separation maps, multiple sclerosis (MS) lesions tend to be hypodiamagnetic, which can serve as an important hallmark to differentiate MS from neuromyelitis optica spectrum disorder. © RSNA, 2022 Supplemental material is available for this article.

Diagnosis of nasal bone fractures on plain radiographs via convolutional neural networks.

This study aimed to assess the performance of deep learning (DL) algorithms in the diagnosis of nasal bone fractures on radiographs and compare it with that of experienced radiologists. In this retrospective study, 6713 patients whose nasal radiographs were examined for suspected nasal bone fractures between January 2009 and October 2020 were assessed. Our dataset was randomly split into training (n = 4325), validation (n = 481), and internal test (n = 1250) sets; a separate external dataset (n = 102) was used. The area under the receiver operating characteristic curve (AUC), sensitivity, and specificity of the DL algorithm and the two radiologists were compared. The AUCs of the DL algorithm for the internal and external test sets were 0.85 (95% CI, 0.83-0.86) and 0.86 (95% CI, 0.78-0.93), respectively, and those of the two radiologists for the external test set were 0.80 (95% CI, 0.73-0.87) and 0.75 (95% CI, 0.68-0.82). The DL algorithm therefore significantly exceeded radiologist 2 (P = 0.021) but did not significantly differ from radiologist 1 (P = 0.142). The sensitivity and specificity of the DL algorithm were 83.1% (95% CI, 71.2-93.2%) and 83.7% (95% CI, 69.8-93.0%), respectively. Our DL algorithm performs comparably to experienced radiologists in diagnosing nasal bone fractures on radiographs.

[Applications of Artificial Intelligence in MR Image Acquisition and Reconstruction]. / MRI ì‹ í˜¸íšë“ê³¼ 영상재구성에서의 인공지능 ì ìš©.

Recently, artificial intelligence (AI) technology has shown potential clinical utility in a wide range of MRI fields. In particular, AI models for improving the efficiency of the image acquisition process and the quality of reconstructed images are being actively developed by the MR research community. AI is expected to further reduce acquisition times in various MRI protocols used in clinical practice when compared to current parallel imaging techniques. Additionally, AI can help with tasks such as planning, parameter optimization, artifact reduction, and quality assessment. Furthermore, AI is being actively applied to automate MR image analysis such as image registration, segmentation, and object detection. For this reason, it is important to consider the effects of protocols or devices in MR image analysis. In this review article, we briefly introduced issues related to AI application of MR image acquisition and reconstruction.

[Brain Iron Imaging in Aging and Cognitive Disorders: MRI Approaches]. / ë ¸í™” 및 ì¸ì§€ê¸°ëŠ¥ìž¥ì• ì—ì„œ 뇌 ì² ì˜ìƒ 기법: ìžê¸°ê³µëª ì˜ìƒì„ 이용한 ì ‘ê·¼.

Iron has a vital role in the human body, including the central nervous system. Increased deposition of iron in the brain has been reported in aging and important neurodegenerative diseases. Owing to the unique magnetic resonance properties of iron, MRI has great potential for in vivo assessment of iron deposition, distribution, and non-invasive quantification. In this paper, we will review the MRI methods for iron assessment and their changes in aging and neurodegenerative diseases, focusing on Alzheimer's disease. In addition, we will summarize the limitations of current approaches and introduce new areas and MRI methods for iron imaging that are expected in the future.

Clinical Implications of Focal Mineral Deposition in the Globus Pallidus on CT and Quantitative Susceptibility Mapping of MRI.

OBJECTIVE: To assess focal mineral deposition in the globus pallidus (GP) by CT and quantitative susceptibility mapping (QSM) of MRI scans and evaluate its clinical significance, particularly cerebrovascular degeneration. MATERIALS AND METHODS: This study included 105 patients (66.1 ± 13.7 years; 40 male and 65 female) who underwent both CT and MRI with available QSM data between January 2017 and December 2019. The presence of focal mineral deposition in the GP on QSM (GPQSM) and CT (GPCT) was assessed visually using a three-point scale. Cerebrovascular risk factors and small vessel disease (SVD) imaging markers were also assessed. The clinical and radiological findings were compared between the different grades of GPQSM and GPCT. The relationship between GP grades and cerebrovascular risk factors and SVD imaging markers was assessed using univariable and multivariable linear regression analyses. RESULTS: GPCT and GPQSM were significantly associated (p < 0.001) but were not identical. Higher GPCT and GPQSM grades showed smaller gray matter (p = 0.030 and p = 0.025, respectively) and white matter (p = 0.013 and p = 0.019, respectively) volumes, as well as larger GP volumes (p < 0.001 for both). Among SVD markers, white matter hyperintensity was significantly associated with GPCT (p = 0.006) and brain atrophy was significantly associated with GPQSM (p = 0.032) in at univariable analysis. In multivariable analysis, the normalized volume of the GP was independently positively associated with GPCT (p < 0.001) and GPQSM (p = 0.002), while the normalized volume of the GM was independently negatively associated with GPCT (p = 0.040) and GPQSM (p = 0.035). CONCLUSION: Focal mineral deposition in the GP on CT and QSM might be a potential imaging marker of cerebral vascular degeneration. Both were associated with increased GP volume.

Multi-Domain Neumann Network with Sensitivity Maps for Parallel MRI Reconstruction.

MRI is an imaging technology that non-invasively obtains high-quality medical images for diagnosis. However, MRI has the major disadvantage of long scan times which cause patient discomfort and image artifacts. As one of the methods for reducing the long scan time of MRI, the parallel MRI method for reconstructing a high-fidelity MR image from under-sampled multi-coil k-space data is widely used. In this study, we propose a method to reconstruct a high-fidelity MR image from under-sampled multi-coil k-space data using deep-learning. The proposed multi-domain Neumann network with sensitivity maps (MDNNSM) is based on the Neumann network and uses a forward model including coil sensitivity maps for parallel MRI reconstruction. The MDNNSM consists of three main structures: the CNN-based sensitivity reconstruction block estimates coil sensitivity maps from multi-coil under-sampled k-space data; the recursive MR image reconstruction block reconstructs the MR image; and the skip connection accumulates each output and produces the final result. Experiments using the fastMRI T1-weighted brain image dataset were conducted at acceleration factors of 2, 4, and 8. Qualitative and quantitative experimental results show that the proposed MDNNSM method reconstructs MR images more accurately than other methods, including the generalized autocalibrating partially parallel acquisitions (GRAPPA) method and the original Neumann network.

Fully Automatic Assessment of Background Parenchymal Enhancement on Breast MRI Using Machine-Learning Models.

BACKGROUND: Automated measurement and classification models with objectivity and reproducibility are required for accurate evaluation of the breast cancer risk of fibroglandular tissue (FGT) and background parenchymal enhancement (BPE). PURPOSE: To develop and evaluate a machine-learning algorithm for breast FGT segmentation and BPE classification. STUDY TYPE: Retrospective. POPULATION: A total of 794 patients with breast cancer, 594 patients assigned to the development set, and 200 patients to the test set. FIELD STRENGTH/SEQUENCE: 3T and 1.5T; T2 -weighted, fat-saturated T1 -weighted (T1 W) with dynamic contrast enhancement (DCE). ASSESSMENT: Manual segmentation was performed for the whole breast and FGT regions in the contralateral breast. The BPE region was determined by thresholding using the subtraction of the pre- and postcontrast T1 W images and the segmented FGT mask. Two radiologists independently assessed the categories of FGT and BPE. A deep-learning-based algorithm was designed to segment and measure the volume of whole breast and FGT and classify the grade of BPE. STATISTICAL TESTS: Dice similarity coefficients (DSC) and Spearman correlation analysis were used to compare the volumes from the manual and deep-learning-based segmentations. Kappa statistics were used for agreement analysis. Comparison of area under the receiver operating characteristic (ROC) curves (AUC) and F1 scores were calculated to evaluate the performance of BPE classification. RESULTS: The mean (±SD) DSC for manual and deep-learning segmentations was 0.85 ± 0.11. The correlation coefficient for FGT volume from manual- and deep-learning-based segmentations was 0.93. Overall accuracy of manual segmentation and deep-learning segmentation in BPE classification task was 66% and 67%, respectively. For binary categorization of BPE grade (minimal/mild vs. moderate/marked), overall accuracy increased to 91.5% in manual segmentation and 90.5% in deep-learning segmentation; the AUC was 0.93 in both methods. DATA CONCLUSION: This deep-learning-based algorithm can provide reliable segmentation and classification results for BPE. LEVEL OF EVIDENCE: 3 TECHNICAL EFFICACY STAGE: 2.

Regulatory T Cells Conditioned Media Stimulates Migration in HaCaT Keratinocytes: Involvement of Wound Healing.

PURPOSE: Regulatory T (Treg) cells, a type of immune cell, play a very important role in the immune response as a subpopulation of T cells. In this study, we investigated the effects of Treg cells conditioned media (CM) on cell migration. Various cytokines and growth factors of Treg cells CM can effect on re-epithelialization stage during the wound healing. METHODS: Isolated CD4+CD25+ Treg cells from Peripheral Blood Mononuclear Cells (PBMCs) were cultured and CM obtained. HaCaT keratinocytes were treated with various concentration of Treg cells CM. Cell migration, proliferation and expression of proteins that are related to the Epithelial-Mesenchymal Transition (EMT) process, matrix metalloproteinase-1 (MMP-1) were analyzed. RESULTS: Above 90% CD4+CD25+ Treg cells were obtained from CD8+ depleted PBMCs and the CM have various cytokines and growth factors.One percent and 5% concentration of Treg cells CM increased HaCaT keratinocytes migration. The Treg cells CM stimulated EMT, which led to the down-regulation of E-cadherin in the HaCaT keratinocytes at the wound edge. The Treg cells CM increased MMP-1, which is involved in tissue remodeling. CONCLUSION: Our results suggest that Treg cells CM which has various cytokines and growth factors promote wound healing by stimulating HaCaT keratinocytes migration.

The complete chloroplast genome of Codonopsis lanceolata (Campanulaceae).

The complete chloroplast genome sequence of Codonopsis lanceolata was determined by next generation sequencing. The total length of chloroplast genome of C. lanceolata was 169,447 bp long, including a large single-copy (LSC) region of 85,253 bp, a small single-copy (SSC) region of 8060 bp, and a pair of identical inverted repeat regions (IRs) of 38,067 bp. A total of 110 genes was annotated, resulting in 79 protein-coding genes, 27 tRNA genes, and 4 rRNA genes. The phylogenetic analysis of C. lanceolata with related chloroplast genome sequences in this study provided the taxonomical relationship of C. lanceolata in the genus Campanula.

The complete chloroplast genome of Caltha Palustris (Ranunculaceae).

The complete chloroplast genome sequence of Caltha palustris, a species of the Ranunculaceae family, was characterized from the de novo assembly of HiSeq (Illumina Co.) paired-end sequencing data. The chloroplast genome of C. palustris was 155,292 bp in length, with a large single-copy (LSC) region of 84,120 bp, a small single-copy (SSC) region of 18,342 bp, and a pair of identical inverted repeat regions (IRs) of 26,415 bp. The genome contained a total of 114 genes, including 80 protein-coding genes, 30 transfer RNA (tRNA) genes, and 4 ribosomal RNA (rRNA) genes. The phylogenetic analysis of C. palustris with 14 related species revealed the closest taxonomical relationship with Hydrastis canadensis in the Ranunculaceae family.