An official website of the United States government
ABSTRACT Necrotizing enterocolitis (NEC) is a devastating disease affecting premature infants. Broadband optical spectroscopy (BOS) is a method of noninvasive optical data collection from intra‐abdominal organs in premature infants, offering potential for disease detection. Herein, a novel machine learning approach, iterative principal component analysis (iPCA), is developed to select optimal wavelengths from BOS data collected in vivo from neonatal intensive care unit (NICU) patients for NEC classification. Neural network models were trained for classification, with a reduced‐feature model distinguishing NEC with an accuracy of 88%, a sensitivity of 89%, and a specificity of 88%. While whole‐spectrum models performed the best for accuracy and specificity, a reduced feature model excelled in sensitivity, with minimal cost to other metrics. This research supports the hypothesis that the analysis of human tissue via BOS may permit noninvasive disease detection. Furthermore, a medical device optimized with these models may potentially screen for NEC with as few as seven wavelengths.
more »
« less
Label-free flow cytometry of rare circulating tumor cell clusters in whole blood
Abstract Circulating tumor cell clusters (CTCCs) are rare cellular events found in the blood stream of metastatic tumor patients. Despite their scarcity, they represent an increased risk for metastasis. Label-free detection methods of these events remain primarily limited to in vitro microfluidic platforms. Here, we expand on the use of confocal backscatter and fluorescence flow cytometry (BSFC) for label-free detection of CTCCs in whole blood using machine learning for peak detection/classification. BSFC uses a custom-built flow cytometer with three excitation wavelengths (405 nm, 488 nm, and 633 nm) and five detectors to detect CTCCs in whole blood based on corresponding scattering and fluorescence signals. In this study, detection of CTCC-associated GFP fluorescence is used as the ground truth to assess the accuracy of endogenous back-scattered light-based CTCC detection in whole blood. Using a machine learning model for peak detection/classification, we demonstrated that the combined use of backscattered signals at the three wavelengths enable detection of ~ 93% of all CTCCs larger than two cells with a purity of > 82% and an overall accuracy of > 95%. The high level of performance established through BSFC and machine learning demonstrates the potential for label-free detection and monitoring of CTCCs in whole blood. Further developments of label-free BSFC to enhance throughput could lead to important applications in the isolation of CTCCs in whole blood with minimal disruption and ultimately their detection in vivo.
more »
« less
- Award ID(s):
- 2018149
- PAR ID:
- 10377361
- Date Published:
- Journal Name:
- Scientific Reports
- Volume:
- 12
- Issue:
- 1
- ISSN:
- 2045-2322
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Cancer diagnostics is an important field of cancer recovery and survival with many expensive procedures needed to administer the correct treatment. Machine Learning (ML) approaches can help with the diagnostic prediction from circulating tumor cells in liquid biopsy or from a primary tumor in solid biopsy. After predicting the metastatic potential from a deep learning model, doctors in a clinical setting can administer a safe and correct treatment for a specific patient. This paper investigates the use of deep convolutional neural networks for predicting a specific cancer cell line as a tool for label free identification. Specifically, deep learning strategies for weight initialization and performance metrics are described, with transfer learning and the accuracy metric utilized in this work. The equipment used for prediction involves brightfield microscopy without the use of chemical labels, advanced instruments, or time-consuming biological techniques, giving an advantage over current diagnostic methods. In the procedure, three different binary datasets of well-known cancer cell lines were collected, each having a difference in metastatic potential. Two different classification models were adopted (EfficientNetV2 and ResNet-50) with the analysis given for each stage in the ML architecture. The training results for each model and dataset are provided and systematically compared. We found that the test set accuracy showed favorable performance for both ML models with EfficientNetV2 accuracy reaching up to 99%. These test results allowed EfficientNetV2 to outperform ResNet-50 at an average percent increase of 3.5% for each dataset. The high accuracy obtained from the predictions demonstrates that the system can be retrained on a large-scale clinical dataset.more » « less
-
Abstract Phagocytosis is a critical component of innate immunity that helps the body defend itself against infection, foreign particles, and cellular debris. Investigating and quantifying phagocytosis can help understand how the immune system identifies foreign particles and how phagocytosis relates to other biomarkers, e.g., cytokines, cell surface receptors, or blood lactate levels. In particular, increased blood lactate levels can be a potential biomarker to study diseases, e.g., septic shock. Establishing a relationship between phagocytosis and lactate levels can serve as an effective tool to monitor the immune response and may help stratify patients. In this study, we use phagocytosis activity data to classify the patients into two groups of blood lactate levels (High and Low) with machine learning models. The neutrophils extracted from the whole blood samples of 19 patients were used to collect data on phagocytosis, where the neutrophils were allowed to internalize IgG coated fluorescent bioparticles. The data collection process involved collecting whole blood samples, neutrophil isolation, adding fluorescent beads, incubating, and imaging the sample using a fluorescence microscope. The phagocytosis assay images were used to generate a numerical dataset by manually counting the number of particles engulfed by each cell. The study first presents an improved understanding by employing hierarchical clustering and heatmaps to generate the graphical representation of phagocytosis data. By comparing the results of heat maps and clustering techniques, it can be observed that the phagocytosis activity data can be used to differentiate blood lactate levels in two groups (control and high-risk). Later, three machine learning models (Decision Tree, k-nearest Neighbor, and Naïve Bayes) were trained on the original and pruned datasets after the outliers were removed. The AI models classified the data into high-risk and low-risk groups of blood lactate levels. A maximum classification accuracy of 78% and an area under the curve of 0.78 was achieved using the trained models.more » « less
-
Abstract Circulating Tumor Cells (CTCs), which migrate from original sites in a body to distant organs through blood, are a key factor in cancer detection. Emerging Label-free techniques owing to their inherent advantage to preserve characteristics of sorted cells and low consumption of samples can be promising to the prediction of cancer progression and metastasis research. Deterministic Lateral Displacement (DLD) is one of the label-free separation techniques employing a specific arrangement of micro-posts for continuous separation of suspended cells in a buffer based on the size of cells. Separation based solely on size is challenging since the size distributions of CTCs might overlap with those of normal blood cells. To address this problem, DLD can be combined with dielectrophoresis (DEP) technique which is the phenomenon of particle movement in a non-uniform electric field owing to the polarization effect. Although, DLD devices employ the laminar flow in low Reynolds number (Re) fluid flow due to predictability of such flow regimes, they should be improved to work in higher Re flow regime so as to attain high throughput devices. In this paper, a particle tracing simulation is developed to study the effects of different post shapes, shift fraction of micropost arrays, and dielectrophoresis forces on separation of CTCs from peripheral blood cells. Our numerical model and results provide a groundwork for design and fabrication of high-throughput DLD-DEP devices for improvement of CTC separation.more » « less
-
Advances in algorithms and low-power computing hardware imply that machine learning is of potential use in off-grid medical data classification and diagnosis applications such as electrocardiogram interpretation. However, although support vector machine algorithms for electrocardiogram classification show high classification accuracy, hardware implementations for edge applications are impractical due to the complexity and substantial power consumption needed for kernel optimization when using conventional complementary metal–oxide–semiconductor circuits. Here we report reconfigurable mixed-kernel transistors based on dual-gated van der Waals heterojunctions that can generate fully tunable individual and mixed Gaussian and sigmoid functions for analogue support vector machine kernel applications. We show that the heterojunction-generated kernels can be used for arrhythmia detection from electrocardiogram signals with high classification accuracy compared with standard radial basis function kernels. The reconfigurable nature of mixed-kernel heterojunction transistors also allows for personalized detection using Bayesian optimization. A single mixed-kernel heterojunction device can generate the equivalent transfer function of a complementary metal–oxide–semiconductor circuit comprising dozens of transistors and thus provides a low-power approach for support vector machine classification applications.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1038/s41598-022-14003-5
