Machine Learning Assisted MRI Characterization for Diagnosis of Neonatal Acute Bilirubin Encephalopathy

Abnormal neonatal brain microstructure in gestational diabetes mellitus revealed by MRI texture analysis

To investigate the value of MRI texture analysis in evaluating the effect of gestational diabetes mellitus (GDM) on neonatal brain microstructure development, we retrospectively collected images of neonates undergoing head MRI scans, including a GDM group (N1 = 37) and a healthy control group (N2 = 34). MaZda texture analysis software was used to extract the texture features from different sequence images and perform dimensionality reduction, and then the texture features selected by the lowest misjudgement rate method were imported into SPSS software for statistical analysis. In our study, we found that GDM affects the development of the microstructure of the neonatal brain, and different combinations of texture features have different recognition performances, such as different sequences and different brain regions. As a consequence, texture analysis combining multiple conventional MRI sequences has a high recognition performance in revealing the abnormal development of the brain microstructure of neonates born of mothers with GDM.

The differential value of radiomics based on traditional T1-weighted sequences in newborns with hyperbilirubinemia

BACKGROUND

On the basis of visual-dependent reading method, radiological recognition and assessment of neonatal hyperbilirubinemia (NH) or acute bilirubin encephalopathy (ABE) on conventional magnetic resonance imaging (MRI) sequences are challenging. Prior studies had shown that radiomics was possible to characterize ABE-induced intensity and morphological changes on MRI sequences, and it has emerged as a desirable and promising future in quantitative and objective MRI data extraction. To investigate the utility of radiomics based on T1-weighted sequences for identifying neonatal ABE in patients with hyperbilirubinemia and differentiating between those with NH and the normal controls.

METHODS

A total of 88 patients with NH were enrolled, including 50 patients with ABE and 38 ABE-negative individuals, and 70 age-matched normal neonates were included as controls. All participants were divided into training and validation cohorts in a 7:3 ratio. Radiomics features extracted from the basal ganglia of T1-weighted sequences on magnetic resonance imaging were evaluated and selected to set up the prediction model using the K-nearest neighbour-based bagging algorithm. A receiver operating characteristic curve was plotted to assess the differentiating performance of the radiomics-based model.

RESULTS

Four of 744 radiomics features were selected for the diagnostic model of ABE. The radiomics model yielded an area under the curve (AUC) of 0.81 and 0.82 in the training and test cohorts, with accuracy, precision, sensitivity, and specificity of 0.82, 0.80, 0.91, and 0.69 and 0.78, 0.8, 0.8, and 0.75, respectively. Six radiomics features were selected in this model to distinguish those with NH from the normal controls. The AUC for the training cohort was 0.97, with an accuracy of 0.92, a precision of 0.92, a sensitivity of 0.93, and a specificity of 0.90. The performance of the radiomics model was confirmed by testing the test cohort, and the AUC, accuracy, precision, sensitivity, and specificity were 0.97, 0.92, 0.96, 0.89, and 0.95, respectively.

CONCLUSIONS

The proposed radiomics model based on traditional TI-weighted sequences may be used effectively for identifying ABE and even differentiating patients with NH from the normal controls, which can provide microcosmic information beyond experience-dependent vision and potentially assist in clinical diagnosis and treatment.

Unsupervised abnormality detection in neonatal MRI brain scans using deep learning

Analysis of 3D medical imaging data has been a large topic of focus in the area of Machine Learning/Artificial Intelligence, though little work has been done in algorithmic (particularly unsupervised) analysis of neonatal brain MRI's. A myriad of conditions can manifest at an early age, including neonatal encephalopathy (NE), which can result in lifelong physical consequences. As such, there is a dire need for better biomarkers of NE and other conditions. The objective of the study is to improve identification of anomalies and prognostication of neonatal MRI brain scans. We introduce a framework designed to support the analysis and assessment of neonatal MRI brain scans, the results of which can be used as an aid to neuroradiologists. We explored the efficacy of the framework through iterations of several deep convolutional Autoencoder (AE) unsupervised modeling architectures designed to learn normalcy of the neonatal brain structure. We tested this framework on the developing human connectome project (dHCP) dataset with 97 patients that were previously categorized by severity. Our framework demonstrated the model's ability to identify and distinguish subtle morphological signatures present in brain structures. Normal and abnormal neonatal brain scans can be distinguished with reasonable accuracy, correctly categorizing them in up to 83% of cases. Most critically, new brain anomalies originally missed during the radiological reading were identified and corroborated by a neuroradiologist. This framework and our modeling approach demonstrate an ability to improve prognostication of neonatal brain conditions and are able to localize new anomalies.

Deep Learning Network with Spatial Attention Module for Detecting Acute Bilirubin Encephalopathy in Newborns Based on Multimodal MRI

BACKGROUND

Acute bilirubin encephalopathy (ABE) is a significant cause of neonatal mortality and disability. Early detection and treatment of ABE can prevent the further development of ABE and its long-term complications. Due to the limited classification ability of single-modal magnetic resonance imaging (MRI), this study aimed to validate the classification performance of a new deep learning model based on multimodal MRI images. Additionally, the study evaluated the effect of a spatial attention module (SAM) on improving the model's diagnostic performance in distinguishing ABE.

METHODS

This study enrolled a total of 97 neonates diagnosed with ABE and 80 neonates diagnosed with hyperbilirubinemia (HB, non-ABE). Each patient underwent three types of multimodal imaging, which included T1-weighted imaging (T1WI), T2-weighted imaging (T2WI), and an apparent diffusion coefficient (ADC) map. A multimodal MRI classification model based on the ResNet18 network with spatial attention modules was built to distinguish ABE from non-ABE. All combinations of the three types of images were used as inputs to test the model's classification performance, and we also analyzed the prediction performance of models with SAMs through comparative experiments.

RESULTS

The results indicated that the diagnostic performance of the multimodal image combination was better than any single-modal image, and the combination of T1WI and T2WI achieved the best classification performance (accuracy = 0.808 ± 0.069, area under the curve = 0.808 ± 0.057). The ADC images performed the worst among the three modalities' images. Adding spatial attention modules significantly improved the model's classification performance.

CONCLUSION

Our experiment showed that a multimodal image classification network with spatial attention modules significantly improved the accuracy of ABE classification.

Development and evaluation clinical-radiomics analysis based on T1-weighted imaging for diagnosing neonatal acute bilirubin encephalopathy

Purpose

To investigate the value of clinical-radiomics analysis based on T1-weighted imaging (T1WI) for predicting acute bilirubin encephalopathy (ABE) in neonates.

Methods

In this retrospective study, sixty-one neonates with clinically confirmed ABE and 50 healthy control neonates were recruited between October 2014 and March 2019. Two radiologists' visual diagnoses for all subjects were independently based on T1WI. Eleven clinical and 216 radiomics features were obtained and analyzed. Seventy percent of samples were randomly selected as the training group and were used to establish a clinical-radiomics model to predict ABE; the remaining samples were used to validate the performance of the models. The discrimination performance was assessed by receiver operating characteristic (ROC) curve analysis.

Results

Seventy-eight neonates were selected for training (median age, 9 days; interquartile range, 7-20 days; 49 males) and 33 neonates for validation (median age, 10 days; interquartile range, 6-13 days; 24 males). Two clinical features and ten radiomics features were finally selected to construct the clinical-radiomics model. In the training group, the area under the ROC curve (AUC) was 0.90 (sensitivity: 0.814; specificity: 0.914); in the validation group, the AUC was 0.93 (sensitivity: 0.944; specificity: 0.800). The AUCs of two radiologists' and the radiologists' final visual diagnosis results based on T1WI were 0.57, 0.63, and 0.66, respectively. The discriminative performance of the clinical-radiomics model in the training and validation groups was increased compared to the radiologists' visual diagnosis (P < 0.001).

Conclusions

A combined clinical-radiomics model based on T1WI has the potential to predict ABE. The application of the nomogram could potentially provide a visualized and precise clinical support tool.

Machine Learning Approaches in Traditional Chinese Medicine: A Systematic Review

Machine learning (ML), as a branch of artificial intelligence, acquires the potential and meaningful rules from the mass of data via diverse algorithms. Owing to all research of traditional Chinese medicine (TCM) belonging to the digitalization of clinical records or experimental works, a massive and complex amount of data has become an inextricable part of the related studies. It is thus not surprising that ML approaches, as novel and efficient tools to mine the useful knowledge from data, have created inroads in a diversity of scopes of TCM over the past decade of years. However, by browsing lots of literature, we find that not all of the ML approaches perform well in the same field. Upon further consideration, we infer that the specificity may inhere between the ML approaches and their applied fields. This systematic review focuses its attention on the four categories of ML approaches and their eight application scopes in TCM. According to the function, ML approaches are classified into four categories, including classification, regression, clustering, and dimensionality reduction, and into 14 models as follows in more detail: support vector machine, least square-support vector machine, logistic regression, partial least squares regression, k-means clustering, hierarchical cluster analysis, artificial neural network, back propagation neural network, convolutional neural network, decision tree, random forest, principal component analysis, partial least squares-discriminant analysis, and orthogonal partial least squares-discriminant analysis. The eight common applied fields are divided into two parts: one for TCM, such as the diagnosis of diseases, the determination of syndromes, and the analysis of prescription, and the other for the related researches of Chinese herbal medicine, such as the quality control, the identification of geographic origins, the pharmacodynamic material basis, the medicinal properties, and the pharmacokinetics and pharmacodynamics. Additionally, this paper discusses the function and feature difference among ML approaches when they are applied to the corresponding fields via comparing their principles. The specificity of each approach to its applied fields has also been affirmed, whereby laying a foundation for subsequent studies applying ML approaches to TCM.

The role of artificial intelligence in paediatric neuroradiology

Imaging plays a fundamental role in the managing childhood neurologic, neurosurgical and neuro-oncological disease. Employing multi-parametric MRI techniques, such as spectroscopy and diffusion- and perfusion-weighted imaging, to the radiophenotyping of neuroradiologic conditions is becoming increasingly prevalent, particularly with radiogenomic analyses correlating imaging characteristics with molecular biomarkers of disease. However, integration into routine clinical practice remains elusive. With modern multi-parametric MRI now providing additional data beyond anatomy, informing on histology, biology and physiology, such metric-rich information can present as information overload to the treating radiologist and, as such, information relevant to an individual case can become lost. Artificial intelligence techniques are capable of modelling the vast radiologic, biological and clinical datasets that accompany childhood neurologic disease, such that this information can become incorporated in upfront prognostic modelling systems, with artificial intelligence techniques providing a plausible approach to this solution. This review examines machine learning approaches than can be used to underpin such artificial intelligence applications, with exemplars for each machine learning approach from the world literature. Then, within the specific use case of paediatric neuro-oncology, we examine the potential future contribution for such artificial intelligence machine learning techniques to offer solutions for patient care in the form of decision support systems, potentially enabling personalised medicine within this domain of paediatric radiologic practice.

Development and validation of a machine learning model to predict the risk of lymph node metastasis in renal carcinoma

SIMPLE SUMMARY

Studies have shown that about 30% of kidney cancer patients will have metastasis, and lymph node metastasis (LNM) may be related to a poor prognosis. Our retrospective study aims to provide a reliable machine learning-based model to predict the occurrence of LNM in kidney cancer. We screened the pathological grade, liver metastasis, M staging, primary site, T staging, and tumor size from the training group (n=39016) formed by the SEER database and the validation group (n=771) formed by the medical center. Independent predictors of LNM in cancer patients. Using six different algorithms to build a prediction model, it is found that the prediction performance of the XGB model in the training group and the validation group is significantly better than any other machine learning model. The results show that prediction tools based on machine learning can accurately predict the probability of LNM in patients with kidney cancer and have satisfactory clinical application prospects.

BACKGROUND

Lymph node metastasis (LNM) is associated with the prognosis of patients with kidney cancer. This study aimed to provide reliable machine learning-based (ML-based) models to predict the probability of LNM in kidney cancer.

METHODS

Data on patients diagnosed with kidney cancer were extracted from the Surveillance, Epidemiology and Outcomes (SEER) database from 2010 to 2017, and variables were filtered by least absolute shrinkage and selection operator (LASSO), univariate and multivariate logistic regression analyses. Statistically significant risk factors were used to build predictive models. We used 10-fold cross-validation in the validation of the model. The area under the receiver operating characteristic curve (AUC) was used to assess the performance of the model. Correlation heat maps were used to investigate the correlation of features using permutation analysis to assess the importance of predictors. Probability density functions (PDFs) and clinical utility curves (CUCs) were used to determine clinical utility thresholds.

RESULTS

The training cohort of this study included 39,016 patients, and the validation cohort included 771 patients. In the two cohorts, 2544 (6.5%) and 66 (8.1%) patients had LNM, respectively. Pathological grade, liver metastasis, M stage, primary site, T stage, and tumor size were independent predictive factors of LNM. In both model validation, the XGB model significantly outperformed any of the machine learning models with an AUC value of 0.916.A web calculator (https://share.streamlit.io/liuwencai4/renal_lnm/main/renal_lnm.py) were built based on the XGB model. Based on the PDF and CUC, we suggested 54.6% as a threshold probability for guiding the diagnosis of LNM, which could distinguish about 89% of LNM patients.

CONCLUSIONS

The predictive tool based on machine learning can precisely indicate the probability of LNM in kidney cancer patients and has a satisfying application prospect in clinical practice.

Amplitude-integrated electroencephalography improves the predictive ability of acute bilirubin encephalopathy

BACKGROUND

To establish a clinical prediction model of acute bilirubin encephalopathy (ABE) using amplitude-integrated electroencephalography (aEEG).

METHODS

A total of 114 neonatal hyperbilirubinemia patients in the Beijing Chaoyang Hospital from August 2015 to October 2018 were enrolled in this study. There were 62 (54.38%) males, and the age of patients undergoing aEEG examination was 2-23 days, with an average of 7.61±4.08 days. Participant clinical information, peak bilirubin value, albumin value, hyperbilirubinemia, and the graphic indicators of aEEG were extracted from medical records, and ABE was diagnosed according to a bilirubin-induced neurological dysfunction (BIND) score >0. Multivariable logistic regression was used to establish a clinical prediction model of ABE. Furthermore, decision curve analysis (DCA) was performed to evaluate the model's predictive value.

RESULTS

According to the BIND score, there were a total of 23 (20.18%) ABE cases. The multivariable logistic regression analysis showed that the value of bilirubin/albumin (B/A), presence of hyperbilirubinemia risk factors, number of sleep-wake cycling (SWC) within 3 hours, widest bandwidth, duration of SWC, and type of SWC were significantly associated with ABE. A clinical prediction model was developed as: p=ex/ (1+ex), X=0.278+0.713*B/A+2.602*with risk factors (with risk factors equals 1) - 1.500*SWC number within 3 hours + 0.219*the widest bandwidth-0.065*the duration of one SWC + 1.491* SWC (mature SWC equals 0, immature SWC equals 1). The area under the curve (AUC) was 0.85 [95% confidence interval (CI): 0.75-0.94], which was significantly higher than the AUC only based on conventional clinical information of B/A (AUC: 0.58, 95% CI: 0.45-0.72). The DCA also showed good predictive ability compared to B/A.

CONCLUSIONS

A clinical prediction model can be established based on the patients' B/A, presence of risk factors for hyperbilirubinemia, number of SWC within 3 hours, widest bandwidth, duration of 1 SWC, and the type of SWC. It has good predictive ability and may improve the diagnostic accuracy of ABE.

CT Angiography-Based Radiomics for Classification of Intracranial Aneurysm Rupture

Background: Intracranial aneurysm rupture is a devastating medical event with a high morbidity and mortality rate. Thus, timely detection and management are critical. The present study aimed to identify the aneurysm radiomics features associated with rupture and to build and evaluate a radiomics classification model of aneurysm rupture. Methods: Radiomics analysis was applied to CT angiography (CTA) images of 393 patients [152 (38.7%) with ruptured aneurysms]. Patients were divided at a ratio of 7:3 into retrospective training (n = 274) and prospective test (n = 119) cohorts. A total of 1,229 radiomics features were automatically calculated from each aneurysm. The feature number was systematically reduced, and the most important classifying features were selected. A logistic regression model was constructed using the selected features and evaluated on training and test cohorts. Radiomics score (Rad-score) was calculated for each patient and compared between ruptured and unruptured aneurysms. Results: Nine radiomics features were selected from the CTA images and used to build the logistic regression model. The radiomics model has shown good performance in the classification of the aneurysm rupture on training and test cohorts [area under the receiver operating characteristic curve: 0.92 [95% confidence interval CI: 0.89-0.95] and 0.86 [95% CI: 0.80-0.93], respectively, p < 0.001]. Rad-score showed statistically significant differences between ruptured and unruptured aneurysms (median, 2.50 vs. -1.60 and 2.35 vs. -1.01 on training and test cohorts, respectively, p < 0.001). Conclusion: The results indicated the potential of aneurysm radiomics features for automatic classification of aneurysm rupture on CTA images.

Current Status of Tissue Clearing and the Path Forward in Neuroscience

Due to the complexity and limited availability of human brain tissues, for decades, pathologists have sought to maximize information gained from individual samples, based on which (patho)physiological processes could be inferred. Recently, new understandings of chemical and physical properties of biological tissues and multiple chemical profiling have given rise to the development of scalable tissue clearing methods allowing superior optical clearing of across-the-scale samples. In the past decade, tissue clearing techniques, molecular labeling methods, advanced laser scanning microscopes, and data visualization and analysis have become commonplace. Combined, they have made 3D visualization of brain tissues with unprecedented resolution and depth widely accessible. To facilitate further advancements and applications, here we provide a critical appraisal of these techniques. We propose a classification system of current tissue clearing and expansion methods that allows users to judge the applicability of individual ones to their questions, followed by a review of the current progress in molecular labeling, optical imaging, and data processing to demonstrate the whole 3D imaging pipeline based on tissue clearing and downstream techniques for visualizing the brain. We also raise the path forward of tissue-clearing-based imaging technology, that is, integrating with state-of-the-art techniques, such as multiplexing protein imaging, in situ signal amplification, RNA detection and sequencing, super-resolution imaging techniques, multiomics studies, and deep learning, for drawing the complete atlas of the human brain and building a 3D pathology platform for central nervous system disorders.

Analysis on the MRI and BAEP Results of Neonatal Brain With Different Levels of Bilirubin

BACKGROUND

To explore whether there is abnormality of neonatal brains' MRI and BAEP with different bilirubin levels, and to provide an objective basis for early diagnosis on the bilirubin induced subclinical damage on brains.

METHODS

To retrospectively analyze the clinical data of 103 neonatal patients, to conduct routine brain MRI examination and BAEP testing, and to analyze BAEP and MRI image results of the neonatal patients, who were divided into three groups based on the levels of total serum bilirubin concentration (TSB): 16 cases in mild group (TSB: 0.0-229.0 ěmol/L), 49 cases in moderate group (TSB: 229.0-342.0 ěmol/L), and 38 cases in severe group (TSB ≥ 342.0 ěmol/L).

RESULTS

We found the following: A. Comparison of the bilirubin value of the different group: The bilirubin value of the mild group is 171.99 ± 33.50 ěmol/L, the moderate group is 293.98 ± 32.09 ěmol/L, and the severe group is 375.59 ± 34.25 ěmol/L. The comparison of bilirubin values of the three groups of neonates (p < 0.01) indicates the difference is statistically significant (p < 0.01). B. The weight value of the <2,500 g group is 2.04 ± 0.21 and the ≥2,500 g group is 3.39 ± 0.46; the weight comparison of the two groups indicates that the difference is statistically significant (p < 0.01). C. Comparison of the abnormal MRI of the different groups: The brain MRI result's abnormal ratio of the mild group is 31.25%, the moderate group is 16.33%, and the severe group is 21.05%, but the comparison of brain MRI results of the three neonates groups indicates that the difference is not statistically significant (p > 0.05). D. Comparison of abnormal MRI signal values of globus pallidus on T1WI in different groups: 1. The comparison of normal group signal values with that of mild group (p < 0.05), with that of moderate group, and with that of severe group (p < 0.01) indicates that the difference is statistically significant.

CONCLUSION

At low level of bilirubin, central nervous system damage may also occur and can be detected as abnormality by MRI and BAEP. Meanwhile, MRI and BAEP can also provide early abnormal information for the judgment of central nervous system damage of the children with NHB who have no acute bilirubin encephalopathy (ABE) clinical features, and provide clues for early treatment and early intervention.

Nomogram for Acute Bilirubin Encephalopathy Risk in Newborns With Extreme Hyperbilirubinemia

Background and Objectives: This work aimed to develop a predictive model of neonatal acute bilirubin encephalopathy.

Methods: We retrospectively analyzed the data on extreme hyperbilirubinemia (EHB) newborns hospitalized in the First Hospital of Jilin University from January 1, 2012 to December 31, 2019. The demographic characteristics, pathological information, and admission examination results of newborns were collected to analyze the factors affecting acute bilirubin encephalopathy and to establish a predictive model.

Results: A total of 517 newborns were included in this study, of which 102 (19.7%) had acute bilirubin encephalopathy. T1WI hyperintensity [18.819 (8.838–40.069)], mother's age > 35 years [2.618 (1.096–6.2530)], abnormal white blood cell (WBC) [6.503 (0.226–18.994)], TSB level [1.340 (1.242–1.445)], and albumin level [0.812 (0.726–0.907)] were independently associated with neonatal acute bilirubin encephalopathy (ABE). All independently associated risk factors were used to form an ABE risk estimation nomogram. The bootstrap validation method was used to internally validate the resulting model. The nomogram demonstrated good accuracy in predicting the risk of ABE, with an unadjusted C index of 0.943 (95% CI, 0.919–0.962) and a bootstrap-corrected C index of 0.900.

Conclusion: A nomogram was constructed using five risk factors of ABE. This model can help clinicians determine the best treatment for neonatal hyperbilirubinemia.

Increased Serum Expression of Inflammatory Cytokines may Serve as Potential Diagnostic Biomarker for Bilirubin Encephalopathy

OBJECTIVES

The present study was designed to explore the roles of inflammatory cytokines interleukin-1β (IL-1β) and Tumor growth factor-β (TGF-β) in the diagnosis and treatment of neonate bilirubin encephalopathy (BE).

METHODS

A total of 128 BE neonates and 128 normal neonates were included. The serum samples of the BE children and controls were collected, and the levels of IL-1β and TGF-β were examined. Moreover, the correlation between the level of bilirubin and serum expression of IL-1β or TGF-β in BE patients was analyzed. Finally, receiver operating characteristic (ROC) curves were generated to determine the diagnostic value of the cytokines.

RESULTS

IL-1β and TGF-β levels were higher in the serum of BE patients than those in non-BE patients, and the expression of either IL-1β or TGF-β showed a strong positive correlation with the serum expression of bilirubin in BE patients. Moreover, the results of ROC analysis showed that either IL-1β or TGF-β could distinguish BE patients from healthy controls.

CONCLUSION

IL-1β and TGF-β levels were upregulated in BE and might function as potential biomarkers or therapeutic targets for BE patients.