Generalizability and Diagnostic Performance of AI Models for Thyroid US.

Background Artificial intelligence (AI) models have improved US assessment of thyroid nodules; however, the lack of generalizability limits the application of these models. Purpose To develop AI models for segmentation and classification of thyroid nodules in US using diverse data sets from nationwide hospitals and multiple vendors, and to measure the impact of the AI models on diagnostic performance. Materials and Methods This retrospective study included consecutive patients with pathologically confirmed thyroid nodules who underwent US using equipment from 12 vendors at 208 hospitals across China from November 2017 to January 2019. The detection, segmentation, and classification models were developed based on the subset or complete set of images. Model performance was evaluated by precision and recall, Dice coefficient, and area under the receiver operating characteristic curve (AUC) analyses. Three scenarios (diagnosis without AI assistance, with freestyle AI assistance, and with rule-based AI assistance) were compared with three senior and three junior radiologists to optimize incorporation of AI into clinical practice. Results A total of 10 023 patients (median age, 46 years [IQR 37-55 years]; 7669 female) were included. The detection, segmentation, and classification models had an average precision, Dice coefficient, and AUC of 0.98 (95% CI: 0.96, 0.99), 0.86 (95% CI: 0.86, 0.87), and 0.90 (95% CI: 0.88, 0.92), respectively. The segmentation model trained on the nationwide data and classification model trained on the mixed vendor data exhibited the best performance, with a Dice coefficient of 0.91 (95% CI: 0.90, 0.91) and AUC of 0.98 (95% CI: 0.97, 1.00), respectively. The AI model outperformed all senior and junior radiologists (P < .05 for all comparisons), and the diagnostic accuracies of all radiologists were improved (P < .05 for all comparisons) with rule-based AI assistance. Conclusion Thyroid US AI models developed from diverse data sets had high diagnostic performance among the Chinese population. Rule-based AI assistance improved the performance of radiologists in thyroid cancer diagnosis. © RSNA, 2023 Supplemental material is available for this article.

Development of machine learning model to predict pulmonary function with low-dose CT-derived parameter response mapping in a community-based chest screening cohort.

PURPOSE: To construct and evaluate the performance of a machine learning-based low dose computed tomography (LDCT)-derived parametric response mapping (PRM) model for predicting pulmonary function test (PFT) results. MATERIALS AND METHODS: A total of 615 subjects from a community-based screening population (40-74 years old) with PFT parameters, including the ratio of the first second forced expiratory volume to forced vital capacity (FEV1/FVC), the percentage of forced expiratory volume in the one second predicted (FEV1%), and registered inspiration-to-expiration chest CT scanning were enrolled retrospectively. Subjects were classified into a normal, high risk, and COPD group based on PFT. Data of 72 PRM-derived quantitative parameters were collected, including volume and volume percentage of emphysema, functional-small airways disease, and normal lung tissue. A machine-learning with random forest regression model and a multilayer perceptron (MLP) model were constructed and tested on PFT prediction, which was followed by evaluation of classification performance based on the PFT predictions. RESULTS: The machine-learning model based on PRM parameters showed better performance for predicting PFT than MLP, with a coefficient of determination (R2 ) of 0.749 and 0.792 for FEV1/FVC and FEV1%, respectively. The Mean Squared Errors (MSE) for FEV1/FVC and FEV1% are 0.0030 and 0.0097 for the random forest model, respectively. The Root Mean Squared Errors (RMSE) for FEV1/FVC and FEV1% are 0.055 and 0.098, respectively. The sensitivity, specificity, and accuracy for differentiating between the normal group and high-risk group were 34/40 (85%), 65/72 (90%), and 99/112 (88%), respectively. For differentiating between the non-COPD group and COPD group, the sensitivity, specificity, and accuracy were 8/9 (89%), 112/112 (100%), 120/121 (99%), respectively. CONCLUSIONS: The machine learning-based random forest model predicts PFT results in a community screening population based on PRM, and it identifies high risk COPD from normal populations with high sensitivity and reliably predicts of high-risk COPD.

Deep convolutional neural network-based classification of cancer cells on cytological pleural effusion images.

Lung cancer is one of the leading causes of cancer-related death worldwide. Cytology plays an important role in the initial evaluation and diagnosis of patients with lung cancer. However, due to the subjectivity of cytopathologists and the region-dependent diagnostic levels, the low consistency of liquid-based cytological diagnosis results in certain proportions of misdiagnoses and missed diagnoses. In this study, we performed a weakly supervised deep learning method for the classification of benign and malignant cells in lung cytological images through a deep convolutional neural network (DCNN). A total of 404 cases of lung cancer cells in effusion cytology specimens from Shanghai Pulmonary Hospital were investigated, in which 266, 78, and 60 cases were used as the training, validation and test sets, respectively. The proposed method was evaluated on 60 whole-slide images (WSIs) of lung cancer pleural effusion specimens. This study showed that the method had an accuracy, sensitivity, and specificity respectively of 91.67%, 87.50% and 94.44% in classifying malignant and benign lesions (or normal). The area under the receiver operating characteristic (ROC) curve (AUC) was 0.9526 (95% confidence interval (CI): 0.9019-9.9909). In contrast, the average accuracies of senior and junior cytopathologists were 98.34% and 83.34%, respectively. The proposed deep learning method will be useful and may assist pathologists with different levels of experience in the diagnosis of cancer cells on cytological pleural effusion images in the future.

Transcriptomes of lineage-specific Drosophila neuroblasts profiled by genetic targeting and robotic sorting.

A brain consists of numerous distinct neurons arising from a limited number of progenitors, called neuroblasts in Drosophila. Each neuroblast produces a specific neuronal lineage. To unravel the transcriptional networks that underlie the development of distinct neuroblast lineages, we marked and isolated lineage-specific neuroblasts for RNA sequencing. We labeled particular neuroblasts throughout neurogenesis by activating a conditional neuroblast driver in specific lineages using various intersection strategies. The targeted neuroblasts were efficiently recovered using a custom-built device for robotic single-cell picking. Transcriptome analysis of mushroom body, antennal lobe and type II neuroblasts compared with non-selective neuroblasts, neurons and glia revealed a rich repertoire of transcription factors expressed among neuroblasts in diverse patterns. Besides transcription factors that are likely to be pan-neuroblast, many transcription factors exist that are selectively enriched or repressed in certain neuroblasts. The unique combinations of transcription factors present in different neuroblasts may govern the diverse lineage-specific neuron fates.

Graphene nanopore with a self-integrated optical antenna.

We report graphene nanopores with integrated optical antennae. We demonstrate that a nanometer-sized heated spot created by photon-to-heat conversion of a gold nanorod resting on a graphene membrane forms a nanoscale pore with a self-integrated optical antenna in a single step. The distinct plasmonic traits of metal nanoparticles, which have a unique capability to concentrate light into nanoscale regions, yield the significant advantage of parallel nanopore fabrication compared to the conventional sequential process using an electron beam. Tunability of both the nanopore dimensions and the optical characteristics of plasmonic nanoantennae are further achieved. Finally, the key optical function of our self-integrated optical antenna on the vicinity of graphene nanopore is manifested by multifold fluorescent signal enhancement during DNA translocation.

A new AI-assisted scoring system for PD-L1 expression in NSCLC.

BACKGROUND: Artificial intelligence (AI) analysis may serve as a scoring tool for programmed cell death ligand-1 (PD-L1) expression. In this study, a new AI-assisted scoring system for pathologists was tested for PD-L1 expression assessment in non-small cell lung cancer (NSCLC). METHODS: PD-L1 expression was evaluated using the tumor proportion score (TPS) categorized into three levels: negative (TPS < 1%), low expression (TPS 1-49%), and high expression (TPS ≥ 50%). In order to train, validate, and test the Aitrox AI segmentation model at the whole slide image (WSI) level, 54, 53, and 115 cases were used as training, validation, and test datasets, respectively. TPS reading results from five experienced pathologists, six inexperienced and the Aitrox AI model were analyzed on 115 PD-L1 stained WSIs. The Gold Standard for TPS was derived from the review of three expert pathologists. Spearman's correlation coefficient was calculated and compared between the results. RESULTS: Aitrox AI Model correlated strongly with the TPS Gold Standard and was comparable with the results of three of the five experienced pathologists. In contrast, the results of four of the six inexperienced pathologists correlated only moderately with the TPS Gold Standard. Aitrox AI Model performed better than the inexperienced pathologists and was comparable to experienced pathologists in both negative and low TPS groups. Despite the fact that the low TPS group showed 5.09% of cases with large fluctuations, the Aitrox AI Model still showed a higher correlation than the inexperienced pathologists. However, the AI model showed unsatisfactory performance in the high TPS groups, especially lower values than the Gold Standard in images with large regions of false-positive cells. CONCLUSION: The Aitrox AI Model demonstrates potential in assisting routine diagnosis of NSCLC by pathologists through scoring of PD-L1 expression.

Re-Defining High Risk COPD with Parameter Response Mapping Based on Machine Learning Models.

Purpose: To explore optimal threshold of FEV1% predicted value (FEV1%pre) for high-risk chronic obstructive pulmonary disease (COPD) using the parameter response mapping (PRM) based on machine learning classification model. Patients and Methods: A total of 561 consecutive non-COPD subjects who were screened for chest diseases in our hospital between August and October 2018 and who had complete questionnaire surveys, pulmonary function tests (PFT), and paired respiratory chest CT scans were enrolled retrospectively. The CT quantitative parameter for small airway remodeling was PRM, and 72 parameters were obtained at the levels of whole lung, left and right lung, and five lobes. To identify a more reasonable thresholds of FEV1% predicted value for distinguishing high-risk COPD patients from the normal, 80 thresholds from 50% to 129% were taken with a partition of 1% to establish a random forest classification model under each threshold, such that novel PFT-parameter-based high-risk criteria would be more consistent with the PRM-based machine learning classification model. Results: Machine learning-based PRM showed that consistency between PRM parameters and PFT was better able to distinguish high-risk COPD from the normal, with an AUC of 0.84 when the threshold was 72%. When the threshold was 80%, the AUC was 0.72 and when the threshold was 95%, the AUC was 0.64. Conclusion: Machine learning-based PRM is feasible for redefining high-risk COPD, and setting the optimal FEV1% predicted value lays the foundation for redefining high-risk COPD diagnosis.

An imaging-based artificial intelligence model for non-invasive grading of hepatic venous pressure gradient in cirrhotic portal hypertension.

The hepatic venous pressure gradient (HVPG) is the gold standard for cirrhotic portal hypertension (PHT), but it is invasive and specialized. Alternative non-invasive techniques are needed to assess the hepatic venous pressure gradient (HVPG). Here, we develop an auto-machine-learning CT radiomics HVPG quantitative model (aHVPG), and then we validate the model in internal and external test datasets by the area under the receiver operating characteristic curves (AUCs) for HVPG stages (≥10, ≥12, ≥16, and ≥20 mm Hg) and compare the model with imaging- and serum-based tools. The final aHVPG model achieves AUCs over 0.80 and outperforms other non-invasive tools for assessing HVPG. The model shows performance improvement in identifying the severity of PHT, which may help non-invasive HVPG primary prophylaxis when transjugular HVPG measurements are not available.

Histological Subtypes Classification of Lung Cancers on CT Images Using 3D Deep Learning and Radiomics.

RATIONALE AND OBJECTIVES: Histological subtypes of lung cancers are critical for clinical treatment decision. In this study, we attempt to use 3D deep learning and radiomics methods to automatically distinguish lung adenocarcinomas (ADC), squamous cell carcinomas (SCC), and small cell lung cancers (SCLC) respectively on Computed Tomography images, and then compare their performance. MATERIALS AND METHODS: 920 patients (mean age 61.2, range, 17-87; 340 Female and 580 Male) with lung cancer, including 554 patients with ADC, 175 patients with lung SCC and 191 patients with SCLC, were included in this retrospective study from January 2013 to August 2018. Histopathologic analysis was available for every patient. The classification models based on 3D deep learning (named the ProNet) and radiomics (named com_radNet) were designed to classify lung cancers into the three types mentioned above according to histopathologic results. The training, validation and testing cohorts counted 0.70, 0.15, and 0.15 of the whole datasets respectively. RESULTS: The ProNet model used to classify the three types of lung cancers achieved the F1-scores of 90.0%, 72.4%, 83.7% in ADC, SCC, and SCLC respectively, and the weighted average F1-score of 73.2%. For com_radNet, the F1-scores achieved 83.1%, 75.4%, 85.1% in ADC, SCC, and SCLC, and the weighted average F1-score was 72.2%. The area under the receiver operating characteristic curve of the ProNet model and com_radNet were 0.840 and 0.789, and the accuracy were 71.6% and 74.7% respectively. CONCLUSION: The ProNet and com_radNet models we developed can achieve high performance in distinguishing ADC, SCC, and SCLC and may be promising approaches for non-invasive predicting histological subtypes of lung cancers.

Static conformation and dynamics of single DNA molecules confined in nanoslits.

Nearly thirty years ago, Daoud and de Gennes derived the scaling predictions for the linear polymer chains trapped in a slit with dimension close to the Kuhn length; however, these predictions have yet to be compared with experiments. We have fabricated nanoslits with vertical dimension similar to the Kuhn length of ds-DNA (110nm) using standard photolithography techniques. Fluorescently labeled single DNA molecules with contour lengths L ranging from 4 to 75 microm were successfully injected into the slits and the chain molecules undergoing Brownian motions were imaged by fluorescence microscopy. The distributions of the chain radius of gyration and the two-dimensional asphericity were measured. It is found that the DNA molecules exhibit highly anisotropic shape and the mean asphericity is chain length independence. The shape anisotropy of DNA in our measurements is between two and three dimensions (2D and 3D). The static scaling law of the chain extension and the radius of gyration , approximately L(nu) were observed with nuR(parallel)=0.65+/-0.02 and nu(Rg)=0.68+/-0.05. These results are close to the average value between two (nuR parallel,Rg=0.75) and three (nuR parallel,Rg=0.6) -dimensional theoretical value. The scaling of the extensional and rotational relaxation time are between Rouse model in nanoslits and Zimm model in the bulk solution, respectively. We show that the conformation and chain relaxation of DNA confined in a slit close to Kuhn length exhibit the quasi-two-dimensional behavior.

Self-powered integrated microfluidic point-of-care low-cost enabling (SIMPLE) chip.

Portable, low-cost, and quantitative nucleic acid detection is desirable for point-of-care diagnostics; however, current polymerase chain reaction testing often requires time-consuming multiple steps and costly equipment. We report an integrated microfluidic diagnostic device capable of on-site quantitative nucleic acid detection directly from the blood without separate sample preparation steps. First, we prepatterned the amplification initiator [magnesium acetate (MgOAc)] on the chip to enable digital nucleic acid amplification. Second, a simplified sample preparation step is demonstrated, where the plasma is separated autonomously into 224 microwells (100 nl per well) without any hemolysis. Furthermore, self-powered microfluidic pumping without any external pumps, controllers, or power sources is accomplished by an integrated vacuum battery on the chip. This simple chip allows rapid quantitative digital nucleic acid detection directly from human blood samples (10 to 105 copies of methicillin-resistant Staphylococcus aureus DNA per microliter, ~30 min, via isothermal recombinase polymerase amplification). These autonomous, portable, lab-on-chip technologies provide promising foundations for future low-cost molecular diagnostic assays.

Opposing intrinsic temporal gradients guide neural stem cell production of varied neuronal fates.

Neural stem cells show age-dependent developmental potentials, as evidenced by their production of distinct neuron types at different developmental times. Drosophila neuroblasts produce long, stereotyped lineages of neurons. We searched for factors that could regulate neural temporal fate by RNA-sequencing lineage-specific neuroblasts at various developmental times. We found that two RNA-binding proteins, IGF-II mRNA-binding protein (Imp) and Syncrip (Syp), display opposing high-to-low and low-to-high temporal gradients with lineage-specific temporal dynamics. Imp and Syp promote early and late fates, respectively, in both a slowly progressing and a rapidly changing lineage. Imp and Syp control neuronal fates in the mushroom body lineages by regulating the temporal transcription factor Chinmo translation. Together, the opposing Imp/Syp gradients encode stem cell age, specifying multiple cell fates within a lineage.

Calcium transients closely reflect prolonged action potentials in iPSC models of inherited cardiac arrhythmia.

Long-QT syndrome mutations can cause syncope and sudden death by prolonging the cardiac action potential (AP). Ion channels affected by mutations are various, and the influences of cellular calcium cycling on LQTS cardiac events are unknown. To better understand LQTS arrhythmias, we performed current-clamp and intracellular calcium ([Ca(2+)]i) measurements on cardiomyocytes differentiated from patient-derived induced pluripotent stem cells (iPS-CM). In myocytes carrying an LQT2 mutation (HERG-A422T), APs and [Ca(2+)]i transients were prolonged in parallel. APs were abbreviated by nifedipine exposure and further lengthened upon releasing intracellularly stored Ca(2+). Validating this model, control iPS-CM treated with HERG-blocking drugs recapitulated the LQT2 phenotype. In LQT3 iPS-CM, expressing NaV1.5-N406K, APs and [Ca(2+)]i transients were markedly prolonged. AP prolongation was sensitive to tetrodotoxin and to inhibiting Na(+)-Ca(2+) exchange. These results suggest that LQTS mutations act partly on cytosolic Ca(2+) cycling, potentially providing a basis for functionally targeted interventions regardless of the specific mutation site.

Fluorescence enhancement and lifetime modification of single nanodiamonds near a nanocrystalline silver surface.

Fluorescent nanodiamond (FND) contains nitrogen-vacancy defect centers as fluorophores. The intensity of its fluorescence can be significantly enhanced after deposition of the particle (35 or 140 nm in size) on a nanocrystalline Ag film without a buffer layer. The excellent photostability (i.e. neither photobleaching nor photoblinking) of the material is preserved even on the Ag film. Concurrent decrease of excited state lifetimes and increase of fluorescence intensities indicate that the enhancement results from surface plasmon resonance. Such a fluorescence enhancement effect is diminished when the individual FND particle is wrapped around by DNA molecules, as a result of an increase in the distance between the color-center emitters inside the FND and the nearby Ag nanoparticles. A fluorescence intensity enhancement up to 10-fold is observed for 35 nm FNDs, confirmed by fluorescence lifetime imaging microscopy.

Mass production and dynamic imaging of fluorescent nanodiamonds.

Fluorescent nanodiamond is a new nanomaterial that possesses several useful properties, including good biocompatibility, excellent photostability and facile surface functionalizability. Moreover, when excited by a laser, defect centres within the nanodiamond emit photons that are capable of penetrating tissue, making them well suited for biological imaging applications. Here, we show that bright fluorescent nanodiamonds can be produced in large quantities by irradiating synthetic diamond nanocrystallites with helium ions. The fluorescence is sufficiently bright and stable to allow three-dimensional tracking of a single particle within the cell by means of either one- or two-photon-excited fluorescence microscopy. The excellent photophysical characteristics are maintained for particles as small as 25 nm, suggesting that fluorescent nanodiamond is an ideal probe for long-term tracking and imaging in vivo, with good temporal and spatial resolution.

Characterization and application of single fluorescent nanodiamonds as cellular biomarkers.

Type Ib diamonds emit bright fluorescence at 550-800 nm from nitrogen-vacancy point defects, (N-V)(0) and (N-V)(-), produced by high-energy ion beam irradiation and subsequent thermal annealing. The emission, together with noncytotoxicity and easiness of surface functionalization, makes nano-sized diamonds a promising fluorescent probe for single-particle tracking in heterogeneous environments. We present the result of our characterization and application of single fluorescent nanodiamonds as cellular biomarkers. We found that, under the same excitation conditions, the fluorescence of a single 35-nm diamond is significantly brighter than that of a single dye molecule such as Alexa Fluor 546. The latter photobleached in the range of 10 s at a laser power density of 10(4) W/cm(2), whereas the nanodiamond particle showed no sign of photobleaching even after 5 min of continuous excitation. Furthermore, no fluorescence blinking was detected within a time resolution of 1 ms. The photophysical properties of the particles do not deteriorate even after surface functionalization with carboxyl groups, which form covalent bonding with polyL-lysines that interact with DNA molecules through electrostatic forces. The feasibility of using surface-functionalized fluorescent nanodiamonds as single-particle biomarkers is demonstrated with both fixed and live HeLa cells.