An automatic and accurate deep learning-based neuroimaging pipeline for the neonatal brain.

BACKGROUND: Accurate segmentation of neonatal brain tissues and structures is crucial for studying normal development and diagnosing early neurodevelopmental disorders. However, there is a lack of an end-to-end pipeline for automated segmentation and imaging analysis of the normal and abnormal neonatal brain. OBJECTIVE: To develop and validate a deep learning-based pipeline for neonatal brain segmentation and analysis of structural magnetic resonance images (MRI). MATERIALS AND METHODS: Two cohorts were enrolled in the study, including cohort 1 (582 neonates from the developing Human Connectome Project) and cohort 2 (37 neonates imaged using a 3.0-tesla MRI scanner in our hospital).We developed a deep leaning-based architecture capable of brain segmentation into 9 tissues and 87 structures. Then, extensive validations were performed for accuracy, effectiveness, robustness and generality of the pipeline. Furthermore, regional volume and cortical surface estimation were measured through in-house bash script implemented in FSL (Oxford Centre for Functional MRI of the Brain Software Library) to ensure reliability of the pipeline. Dice similarity score (DSC), the 95th percentile Hausdorff distance (H95) and intraclass correlation coefficient (ICC) were calculated to assess the quality of our pipeline. Finally, we finetuned and validated our pipeline on 2-dimensional thick-slice MRI in cohorts 1 and 2. RESULTS: The deep learning-based model showed excellent performance for neonatal brain tissue and structural segmentation, with the best DSC and the 95th percentile Hausdorff distance (H95) of 0.96 and 0.99 mm, respectively. In terms of regional volume and cortical surface analysis, our model showed good agreement with ground truth. The ICC values for the regional volume were all above 0.80. Considering the thick-slice image pipeline, the same trend was observed for brain segmentation and analysis. The best DSC and H95 were 0.92 and 3.00 mm, respectively. The regional volumes and surface curvature had ICC values just below 0.80. CONCLUSIONS: We propose an automatic, accurate, stable and reliable pipeline for neonatal brain segmentation and analysis from thin and thick structural MRI. The external validation showed very good reproducibility of the pipeline.

Application of Imaging Indicators Based on 18F-Fluorodeoxyglucose Positron Emission Tomography/Computed Tomography in Colorectal Peritoneal Carcinomatosis.

Objective: The imaging features of peritoneal carcinomatosis (PC) with different locations and pathological types of colorectal cancer (CRC) on 18F-fluorodeoxyglucose positron emission tomography/computed tomography (18F-FDG PET/CT) were analyzed and discussed. Methods: The PET/CT data of 132 patients with colorectal peritoneal carcinomatosis (CRPC) who met the inclusion and exclusion criteria between May 30, 2016, and December 31, 2019, were collected and analyzed. Observations included the location and pathological type of CRC, the peritoneal cancer index (PCI), standardized uptake maximum value (SUVmax), and retention index (RI) of the CRPC. Statistical analysis was performed using SPSS 20.0 software, and P < 0.05 was considered statistically significant. Results: (1) The range of the PCI in the 132 patients studied was 2-30, with a mean value of 7.40 ± 8.14. The maximum long diameter of the CRPC lesions ranged from 0.6 to 12.1 cm, with an average of 3.23 ± 1.94 cm. The SUVmax ranged from 1.2 to 31.0, with a mean value of 9.65 ± 6.01. The SUVmax and size correlation coefficient for maximal CRPC lesions was r = 0.47 (P < 0.001). The RI range of the 72 patients who underwent time-lapse scanning was -10.0-112.2%, with RI quartiles of 13.5-48.9%; RI was ≥5% in 65 cases and <5% in seven cases. (2) The patients were grouped by the location of their CRC: the right-sided colon cancer (RCC, n = 37), left-sided colon cancer (LCC, n = 44), and rectal cancer groups (RC, n = 51). There were significant differences in the CRC pathological types (P = 0.009) and PCI scores (P = 0.02) between the RCC and RC groups and the RI between the RCC group and the other two groups (P < 0.001). (3) There were 88 patients organized into three groups by the pathology of their CRC: the moderately well-differentiated adenocarcinoma (group A, n = 57), poorly differentiated adenocarcinoma (group B, n = 16), and mucinous adenocarcinoma groups (group C, n = 15 cases, including one case of signet-ring cell carcinoma). There were significant differences in the CRC position (P = 0.003) and SUVmax (P = 0.03) between groups A and C. Conclusion: The PCI, SUVmax, and RI of peritoneal metastatic carcinoma caused by CRC in different locations and pathological types vary. Mucinous adenocarcinoma and poorly differentiated adenocarcinoma are relatively common in the right colon, and the PCI of peritoneal metastatic carcinoma is fairly high, but the SUVmax and RI are somewhat low.

Combining quantitative susceptibility mapping to radiomics in diagnosing Parkinson's disease and assessing cognitive impairment.

OBJECTIVE: To explore whether magnetic susceptibility value (MSV) and radiomics features of the nigrostriatal system could be used as imaging markers for diagnosing Parkinson's disease (PD) and its related cognitive impairment (CI). METHODS: A total of 104 PD patients and 45 age-sex-matched healthy controls (HCs) underwent quantitative susceptibility mapping (QSM). The former completed Hoehn-Yahr Stage and Montreal Cognitive Assessment (MoCA). The patients were divided into several subgroups according to disease stages, courses, and MoCA scores. The ROI was subdivided into the substantia nigra (SN), head of caudate nucleus (HCN), and putamen. The MSVs and radiomics features were obtained from QSM. The multivariable logistic regression (MLR) and support vector machine (SVM) models were constructed to diagnose PD. The correlations between MSVs, radiomics features, and MoCA scores were evaluated. RESULTS: The MSVs in bilateral SN pars compacta (SNc) of PD patients were higher than those of the HCs (p < 0.001). There were differences in some radiomics features between the two groups (p < 0.05). The MSVs of the right SNc and the radiomics features of the right SN had the highest area under the curve (AUC), respectively. The comprehensive MLR model (0.90) and SVM model (0.95) revealed better classification performance than MSVs (p < 0.05) in diagnosing PD. The MSVs from the HCN were negatively correlated with MoCA scores in PD subgroups. There were correlations between radiomics features and MoCA scores in PD patients. CONCLUSIONS: Radiomics features and MSVs of the nigrostriatal system from QSM could have crucial role in diagnosing PD and assessing CI. KEY POINTS: • The MLR and the SVM models have excellent diagnostic performance in the diagnosis of PD. • A PD diagnostic nomogram, created based on MSV and the radiomics scores of SVM model, is very convenient for clinical use. • The radiomics features of the nigrostriatal system based on QSM help to evaluate the cognitive impairment in PD patients.