Urine Sediment Recognition Method Based on Multi-View Deep Residual Learning in Microscopic Image

Deep learning radiomics of multimodal ultrasound for classifying metastatic cervical lymphadenopathy into primary cancer sites: a feasibility study

PURPOSE

To investigate the feasibility of deep learning radiomics (DLR) based on multimodal ultrasound to differentiate the primary cancer sites of metastatic cervical lymphadenopathy (CLA).

MATERIALS AND METHODS

This study analyzed 280 biopsy-confirmed metastatic CLAs from 280 cancer patients, including 54 from head and neck squamous cell carcinoma (HNSCC), 58 from thyroid cancer (TC), 92 from lung cancer (LC), and 76 from gastrointestinal cancer (GIC). Before biopsy, patients underwent conventional ultrasound (CUS), ultrasound elastography (UE), and contrast-enhanced ultrasound (CEUS). Based on CUS, DLR models using CUS, CUS+UE, CUS+CEUS, and CUS+UE+CEUS data were developed and compared. The best model was integrated with key clinical indicators selected by univariate analysis to achieve the best classification performance.

RESULTS

All DLR models achieved similar performance with respect to classifying four primary tumor sites of metastatic CLA (AUC:0.708~0.755). After integrating key clinical indicators (age, sex, and neck level), the US+UE+CEUS+clinical model yielded the best performance with an overall AUC of 0.822 in the validation cohort, but there was no significance compared with the basal CUS+clinical model (P>0.05), both of which identified metastasis from HNSCC, TC, LC, and GIC with 0.869 and 0.911, 0.838 and 0.916, 0.750 and 0.610, and 0.829 and 0.769, respectively.

CONCLUSION

The ultrasound-based DLR model can be used to classify the primary cancer sites of metastatic CLA, and the CUS combined with clinical indicators is adequate to provide a high discriminatory performance. The addition of the combination of UE and CEUS data is expected to further improve performance.

Artificial intelligence in the clinical laboratory

Laboratory medicine has become a highly automated medical discipline. Nowadays, artificial intelligence (AI) applied to laboratory medicine is also gaining more and more attention, which can optimize the entire laboratory workflow and even revolutionize laboratory medicine in the future. However, only a few commercially available AI models are currently approved for use in clinical laboratories and have drawbacks such as high cost, lack of accuracy, and the need for manual review of model results. Furthermore, there are a limited number of literature reviews that comprehensively address the research status, challenges, and future opportunities of AI applications in laboratory medicine. Our article begins with a brief introduction to AI and some of its subsets, then reviews some AI models that are currently being used in clinical laboratories or that have been described in emerging studies, and explains the existing challenges associated with their application and possible solutions, finally provides insights into the future opportunities of the field. We highlight the current status of implementation and potential applications of AI models in different stages of the clinical testing process.

Deep learning radiomics-based prediction model of metachronous distant metastasis following curative resection for retroperitoneal leiomyosarcoma: a bicentric study

BACKGROUND

Combining conventional radiomics models with deep learning features can result in superior performance in predicting the prognosis of patients with tumors; however, this approach has never been evaluated for the prediction of metachronous distant metastasis (MDM) among patients with retroperitoneal leiomyosarcoma (RLS). Thus, the purpose of this study was to develop and validate a preoperative contrast-enhanced computed tomography (CECT)-based deep learning radiomics model for predicting the occurrence of MDM in patients with RLS undergoing complete surgical resection.

METHODS

A total of 179 patients who had undergone surgery for the treatment of histologically confirmed RLS were retrospectively recruited from two tertiary sarcoma centers. Semantic segmentation features derived from a convolutional neural network deep learning model as well as conventional hand-crafted radiomics features were extracted from preoperative three-phase CECT images to quantify the sarcoma phenotypes. A conventional radiomics signature (RS) and a deep learning radiomics signature (DLRS) that incorporated hand-crafted radiomics and deep learning features were developed to predict the risk of MDM. Additionally, a deep learning radiomics nomogram (DLRN) was established to evaluate the incremental prognostic significance of the DLRS in combination with clinico-radiological predictors.

RESULTS

The comparison of the area under the curve (AUC) values in the external validation set, as determined by the DeLong test, demonstrated that the integrated DLRN, DLRS, and RS models all exhibited superior predictive performance compared with that of the clinical model (AUC 0.786 [95% confidence interval 0.649-0.923] vs. 0.822 [0.692-0.952] vs. 0.733 [0.573-0.892] vs. 0.511 [0.359-0.662]; both P < 0.05). The decision curve analyses graphically indicated that utilizing the DLRN for risk stratification provided greater net benefits than those achieved using the DLRS, RS and clinical models. Good alignment with the calibration curve indicated that the DLRN also exhibited good performance.

CONCLUSIONS

The novel CECT-based DLRN developed in this study demonstrated promising performance in the preoperative prediction of the risk of MDM following curative resection in patients with RLS. The DLRN, which outperformed the other three models, could provide valuable information for predicting surgical efficacy and tailoring individualized treatment plans in this patient population.

TRIAL REGISTRATION

Not applicable.

Deep Learning Radiomics Nomogram Based on Enhanced CT to Predict the Response of Metastatic Lymph Nodes to Neoadjuvant Chemotherapy in Locally Advanced Gastric Cancer

BACKGROUND

We aimed to construct and validate a deep learning (DL) radiomics nomogram using baseline and restage enhanced computed tomography (CT) images and clinical characteristics to predict the response of metastatic lymph nodes to neoadjuvant chemotherapy (NACT) in locally advanced gastric cancer (LAGC).

METHODS

We prospectively enrolled 112 patients with LAGC who received NACT from January 2021 to August 2022. After applying the inclusion and exclusion criteria, 98 patients were randomized 7:3 to the training cohort (n = 68) and validation cohort (n = 30). We established and compared three radiomics signatures based on three phases of CT images before and after NACT, namely radiomics-baseline, radiomics-delta, and radiomics-restage. Then, we developed a clinical model, DL model, and a nomogram to predict the response of LAGC after NACT. We evaluated the predictive accuracy and clinical validity of each model using the receiver operating characteristic curve and decision curve analysis, respectively.

RESULTS

The radiomics-delta signature was the best predictor among the three radiomics signatures. So, we developed and validated a DL delta radiomics nomogram (DLDRN). In the validation cohort, the DLDRN produced an area under the receiver operating curve of 0.94 (95% confidence interval, 0.82-0.96) and demonstrated adequate differentiation of good response to NACT. Furthermore, the DLDRN significantly outperformed the clinical model and DL model (p < 0.001). The clinical utility of the DLDRN was confirmed through decision curve analysis.

CONCLUSIONS

In patients with LAGC, the DLDRN effectively predicted a therapeutic response in metastatic lymph nodes, which could provide valuable information for individualized treatment.

Fully automated segmentation and radiomics feature extraction of hypopharyngeal cancer on MRI using deep learning

OBJECTIVES

To use convolutional neural network for fully automated segmentation and radiomics features extraction of hypopharyngeal cancer (HPC) tumor in MRI.

METHODS

MR images were collected from 222 HPC patients, among them 178 patients were used for training, and another 44 patients were recruited for testing. U-Net and DeepLab V3 + architectures were used for training the models. The model performance was evaluated using the dice similarity coefficient (DSC), Jaccard index, and average surface distance. The reliability of radiomics parameters of the tumor extracted by the models was assessed using intraclass correlation coefficient (ICC).

RESULTS

The predicted tumor volumes by DeepLab V3 + model and U-Net model were highly correlated with those delineated manually (p < 0.001). The DSC of DeepLab V3 + model was significantly higher than that of U-Net model (0.77 vs 0.75, p < 0.05), particularly in those small tumor volumes of < 10 cm3 (0.74 vs 0.70, p < 0.001). For radiomics extraction of the first-order features, both models exhibited high agreement (ICC: 0.71-0.91) with manual delineation. The radiomics extracted by DeepLab V3 + model had significantly higher ICCs than those extracted by U-Net model for 7 of 19 first-order features and for 8 of 17 shape-based features (p < 0.05).

CONCLUSION

Both DeepLab V3 + and U-Net models produced reasonable results in automated segmentation and radiomic features extraction of HPC on MR images, whereas DeepLab V3 + had a better performance than U-Net.

CLINICAL RELEVANCE STATEMENT

The deep learning model, DeepLab V3 + , exhibited promising performance in automated tumor segmentation and radiomics extraction for hypopharyngeal cancer on MRI. This approach holds great potential for enhancing the radiotherapy workflow and facilitating prediction of treatment outcomes.

KEY POINTS

• DeepLab V3 + and U-Net models produced reasonable results in automated segmentation and radiomic features extraction of HPC on MR images. • DeepLab V3 + model was more accurate than U-Net in automated segmentation, especially on small tumors. • DeepLab V3 + exhibited higher agreement for about half of the first-order and shape-based radiomics features than U-Net.

Automated Urine Cell Image Classification Model Using Chaotic Mixer Deep Feature Extraction

Microscopic examination of urinary sediments is a common laboratory procedure. Automated image-based classification of urinary sediments can reduce analysis time and costs. Inspired by cryptographic mixing protocols and computer vision, we developed an image classification model that combines a novel Arnold Cat Map (ACM)- and fixed-size patch-based mixer algorithm with transfer learning for deep feature extraction. Our study dataset comprised 6,687 urinary sediment images belonging to seven classes: Cast, Crystal, Epithelia, Epithelial nuclei, Erythrocyte, Leukocyte, and Mycete. The developed model consists of four layers: (1) an ACM-based mixer to generate mixed images from resized 224 × 224 input images using fixed-size 16 × 16 patches; (2) DenseNet201 pre-trained on ImageNet1K to extract 1,920 features from each raw input image, and its six corresponding mixed images were concatenated to form a final feature vector of length 13,440; (3) iterative neighborhood component analysis to select the most discriminative feature vector of optimal length 342, determined using a k-nearest neighbor (kNN)-based loss function calculator; and (4) shallow kNN-based classification with ten-fold cross-validation. Our model achieved 98.52% overall accuracy for seven-class classification, outperforming published models for urinary cell and sediment analysis. We demonstrated the feasibility and accuracy of deep feature engineering using an ACM-based mixer algorithm for image preprocessing combined with pre-trained DenseNet201 for feature extraction. The classification model was both demonstrably accurate and computationally lightweight, making it ready for implementation in real-world image-based urine sediment analysis applications.

Automatic Classification of Particles in the Urine Sediment Test with the Developed Artificial Intelligence-Based Hybrid Model

Urine sediment examination is one of the main tests used in the diagnosis of many diseases. Thanks to this test, many diseases can be detected in advance. Examining the results of this test is an intensive and time-consuming process. Therefore, it is very important to automatically interpret the urine sediment test results using computer-aided systems. In this study, a data set consisting of eight classes was used. The data set used in the study consists of 8509 particle images obtained by examining the particles in the urine sediment. A hybrid model based on textural and Convolutional Neural Networks (CNN) was developed to classify the images in the related data set. The features obtained using textural-based methods and the features obtained from CNN-based architectures were combined after optimizing using the Minimum Redundancy Maximum Relevance (mRMR) method. In this way, we aimed to extract different features of the same image. This increased the performance of the proposed model. The CNN-based ResNet50 architecture and textural-based Local Binary Pattern (LBP) method were used for feature extraction. Finally, the optimized and combined feature map was classified at different machine learning classifiers. In order to compare the performance of the model proposed in the study, results were also obtained from different CNN architectures. A high accuracy value of 96.0% was obtained in the proposed model.

Automatic tumor segmentation and metachronous single-organ metastasis prediction of nasopharyngeal carcinoma patients based on multi-sequence magnetic resonance imaging

Background

Distant metastases is the main failure mode of nasopharyngeal carcinoma. However, early prediction of distant metastases in NPC is extremely challenging. Deep learning has made great progress in recent years. Relying on the rich data features of radiomics and the advantages of deep learning in image representation and intelligent learning, this study intends to explore and construct the metachronous single-organ metastases (MSOM) based on multimodal magnetic resonance imaging.

Patients and methods

The magnetic resonance imaging data of 186 patients with nasopharyngeal carcinoma before treatment were collected, and the gross tumor volume (GTV) and metastatic lymph nodes (GTVln) prior to treatment were defined on T1WI, T2WI, and CE-T1WI. After image normalization, the deep learning platform Python (version 3.9.12) was used in Ubuntu 20.04.1 LTS to construct automatic tumor detection and the MSOM prediction model.

Results

There were 85 of 186 patients who had MSOM (including 32 liver metastases, 25 lung metastases, and 28 bone metastases). The median time to MSOM was 13 months after treatment (7–36 months). The patients were randomly assigned to the training set (N = 140) and validation set (N = 46). By comparison, we found that the overall performance of the automatic tumor detection model based on CE-T1WI was the best (6). The performance of automatic detection for primary tumor (GTV) and lymph node gross tumor volume (GTVln) based on the CE-T1WI model was better than that of models based on T1WI and T2WI (AP@0.5 is 59.6 and 55.6). The prediction model based on CE-T1WI for MSOM prediction achieved the best overall performance, and it obtained the largest AUC value (AUC = 0.733) in the validation set. The precision, recall, precision, and AUC of the prediction model based on CE-T1WI are 0.727, 0.533, 0.730, and 0.733 (95% CI 0.557–0.909), respectively. When clinical data were added to the deep learning prediction model, a better performance of the model could be obtained; the AUC of the integrated model based on T2WI, T1WI, and CE-T1WI were 0.719, 0.738, and 0.775, respectively. By comparing the 3-year survival of high-risk and low-risk patients based on the fusion model, we found that the 3-year DMFS of low and high MSOM risk patients were 95% and 11.4%, respectively (p &lt; 0.001).

Conclusion

The intelligent prediction model based on magnetic resonance imaging alone or combined with clinical data achieves excellent performance in automatic tumor detection and MSOM prediction for NPC patients and is worthy of clinical application.

Deep Learning Radiomics Nomogram to Predict Lung Metastasis in Soft-Tissue Sarcoma: A Multi-Center Study

Objectives

To build and evaluate a deep learning radiomics nomogram (DLRN) for preoperative prediction of lung metastasis (LM) status in patients with soft tissue sarcoma (STS).

Methods

In total, 242 patients with STS (training set, n=116; external validation set, n=126) who underwent magnetic resonance imaging were retrospectively enrolled in this study. We identified independent predictors for LM-status and evaluated their performance. The minimum redundancy maximum relevance (mRMR) method and least absolute shrinkage and selection operator (LASSO) algorithm were adopted to screen radiomics features. Logistic regression, decision tree, random forest, support vector machine (SVM), and adaptive boosting classifiers were compared for their ability to predict LM. To overcome the imbalanced distribution of the LM data, we retrained each machine-learning classifier using the synthetic minority over-sampling technique (SMOTE). A DLRN combining the independent clinical predictors with the best performing radiomics prediction signature (mRMR+LASSO+SVM+SMOTE) was established. Area under the receiver operating characteristics curve (AUC), calibration curves, and decision curve analysis (DCA) were used to assess the performance and clinical applicability of the models.

Result

Comparisons of the AUC values applied to the external validation set revealed that the DLRN model (AUC=0.833) showed better prediction performance than the clinical model (AUC=0.664) and radiomics model (AUC=0.799). The calibration curves indicated good calibration efficiency and the DCA showed the DLRN model to have greater clinical applicability than the other two models.

Conclusion

The DLRN was shown to be an accurate and efficient tool for LM-status prediction in STS.

Diagnostic Accuracy of Wireless Capsule Endoscopy in Polyp Recognition Using Deep Learning: A Meta-Analysis

AIM

As the completed studies have small sample sizes and different algorithms, a meta-analysis was conducted to assess the accuracy of WCE in identifying polyps using deep learning.

METHOD

Two independent reviewers searched PubMed, Embase, the Web of Science, and the Cochrane Library for potentially eligible studies published up to December 8, 2021, which were analysed on a per-image basis. STATA RevMan and Meta-DiSc were used to conduct this meta-analysis. A random effects model was used, and a subgroup and regression analysis was performed to explore sources of heterogeneity.

RESULTS

Eight studies published between 2017 and 2021 included 819 patients, and 18,414 frames were eventually included in the meta-analysis. The summary estimates for the WCE in identifying polyps by deep learning were sensitivity 0.97 (95% confidence interval (CI), 0.95-0.98); specificity 0.97 (95% CI, 0.94-0.98); positive likelihood ratio 27.19 (95% CI, 15.32-50.42); negative likelihood ratio 0.03 (95% CI 0.02-0.05); diagnostic odds ratio 873.69 (95% CI, 387.34-1970.74); and the area under the sROC curve 0.99.

CONCLUSION

WCE uses deep learning to identify polyps with high accuracy, but multicentre prospective randomized controlled studies are needed in the future.

CNN-GRU-AM for Shared Bicycles Demand Forecasting

The demand forecast of shared bicycles directly determines the utilization rate of vehicles and projects operation benefits. Accurate prediction based on the existing operating data can reduce unnecessary delivery. Since the use of shared bicycles is susceptible to time dependence and external factors, most of the existing works only consider some of the attributes of shared bicycles, resulting in insufficient modeling and unsatisfactory prediction performance. In order to address the aforementioned limitations, this paper establishes a novelty prediction model based on convolutional recurrent neural network with the attention mechanism named as CNN-GRU-AM. There are four parts in the proposed CNN-GRU-AM model. First, a convolutional neural network (CNN) with two layers is used to extract local features from the multiple sources data. Second, the gated recurrent unit (GRU) is employed to capture the time-series relationships of the output data of CNN. Third, the attention mechanism (AM) is introduced to mining the potential relationships of the series features, in which different weights will be assigned to the corresponding features according to their importance. At last, a fully connected layer with three layers is added to learn features and output the prediction results. To evaluate the performance of the proposed method, we conducted massive experiments on two datasets including a real mobile bicycle data and a public shared bicycle data. The experimental results show that the prediction performance of the proposed model is better than other prediction models, indicating the significance of the social benefits.

Calcium Spark Detection and Event-Based Classification of Single Cardiomyocyte Using Deep Learning

Ca²⁺ sparks are the elementary Ca²⁺ release events in cardiomyocytes, altered properties of which lead to impaired Ca²⁺ handling and finally contribute to cardiac pathology under various diseases. Despite increasing use of machine-learning algorithms in deciphering the content of biological and medical data, Ca²⁺ spark images and data are yet to be deeply learnt and analyzed. In the present study, we developed a deep residual convolutional neural network method to detect Ca²⁺ sparks. Compared to traditional detection methods with arbitrarily defined thresholds to distinguish signals from noises, our new method detected more Ca²⁺ sparks with lower amplitudes but similar spatiotemporal distributions, thereby indicating that our new algorithm detected many very weak events that are usually omitted when using traditional detection methods. Furthermore, we proposed an event-based logistic regression and binary classification model to classify single cardiomyocytes using Ca²⁺ spark characteristics, which to date have generally been used only for simple statistical analyses and comparison between normal and diseased groups. Using this new detection algorithm and classification model, we succeeded in distinguishing wild type (WT) vs RyR2-R2474S^± cardiomyocytes with 100% accuracy, and vehicle vs isoprenaline-insulted WT cardiomyocytes with 95.6% accuracy. The model can be extended to judge whether a small number of cardiomyocytes (and so the whole heart) are under a specific cardiac disease. Thus, this study provides a novel and powerful approach for the research and application of calcium signaling in cardiac diseases.

Medical Image Recognition Technology in the Effect of Substituting Soybean Meal for Fish Meal on the Diversity of Intestinal Microflora in Channa argus

Purpose

To study the application of medical image recognition technology based on backpropagation neural network (BPNN) in the effect of soybean meal replacing fish meal on intestinal microbial diversity of Channa argus and to evaluate the application value of this intelligent algorithm, Channa argus was fed with different contents of soybean meal instead of fish meal.

Methods

After intestinal samples were collected and bacteria were isolated, microscopic imaging was performed, and the images were classified and identified. BPNN was constructed to perform denoising, smoothing, and segmentation.

Results

After BPNN processing, the bacteria were completely separated from the original image background, and the bacteria was in the closed state, which was beneficial to feature extraction and species recognition. If there were 2 hidden layer nodes, the segmentation accuracy of bacterial microscopic images was the highest, up to 97.3%. With the replacement ratio of fish meal increased, the species of intestinal microbiome gradually enriched, and the relative abundance of intestinal microbiome was higher after fish meal was completely replaced by soybean meal (replacement). The intestinal microbial enzyme activities were affected by different fish meal and soybean meal contents in the diet. The glutamate transaminase and adenosine deaminase activities were increased after the replacement and were higher than those before the replacement, with statistically significant differences (P < 0.05).

Conclusion

Replacement of fish meal with soybean meal has a significant effect on the intestinal flora diversity of Channa argus, and there is a close relationship between them. The image recognition technology based on BPNN has high recognition rate and segmentation accuracy for microbiological microscopic images.

Deep learning radiomic nomogram to predict recurrence in soft tissue sarcoma: a multi-institutional study

OBJECTIVES

To evaluate the performance of a deep learning radiomic nomogram (DLRN) model at predicting tumor relapse in patients with soft tissue sarcomas (STS) who underwent surgical resection.

METHODS

In total, 282 patients who underwent MRI and resection for STS at three independent centers were retrospectively enrolled. In addition, 113 of the 282 patients received additional contrast-enhanced MRI scans. We separated the participants into a development cohort and an external test cohort. The development cohort consisted of patients from one center and the external test cohort consisted of patients from two other centers. Two MRI-based DLRNs for prediction of tumor relapse after resection of STS were established. We universally tested the DLRNs and compared them with other prediction models constructed by using widespread adopted predictors (i.e., staging systems and Ki67) instead of radiomics features.

RESULTS

The DLRN1 model incorporated plain MRI-based radiomics signature into the clinical data, and the DLRN2 model integrated radiomics signature extracted from plain and contrast-enhanced MRI with the clinical predictors. Across both study sets, the two MRI-based DLRNs had relatively better prognostic capability (C index ≥ 0.721 and median AUC ≥ 0.746; p < 0.05 compared with most other models and predictors) and less opportunity for prediction error (integrated Brier score ≤ 0.159). The decision curve analysis indicates that the DLRNs have greater benefits than staging systems, Ki67, and other models. We selected appropriate cutoff values for the DLRNs to divide STS recurrence into three risk strata (low, medium, and high) and calculated those groups' cumulative risk rates.

CONCLUSION

The DLRNs were shown to be a reliable and externally validated tool for predicting STS recurrence by comparing with other prediction models.

KEY POINTS

• The prediction of a high recurrence rate of STS before emergence of local recurrence can help to determine whether more active treatment should be implemented. • Two MRI-based DLRNs for prediction of tumor relapse were shown to be a reliable and externally validated tool for predicting STS recurrence. • We used the DLRNs to divide STS recurrence into three risk strata (low, medium, and high) to facilitate more targeted postoperative management in the clinic.

AI based colorectal disease detection using real-time screening colonoscopy

Colonoscopy is an effective tool for early screening of colorectal diseases. However, the application of colonoscopy in distinguishing different intestinal diseases still faces great challenges of efficiency and accuracy. Here we constructed and evaluated a deep convolution neural network (CNN) model based on 117 055 images from 16 004 individuals, which achieved a high accuracy of 0.933 in the validation dataset in identifying patients with polyp, colitis, colorectal cancer (CRC) from normal. The proposed approach was further validated on multi-center real-time colonoscopy videos and images, which achieved accurate diagnostic performance on detecting colorectal diseases with high accuracy and precision to generalize across external validation datasets. The diagnostic performance of the model was further compared to the skilled endoscopists and the novices. In addition, our model has potential in diagnosis of adenomatous polyp and hyperplastic polyp with an area under the receiver operating characteristic curve of 0.975. Our proposed CNN models have potential in assisting clinicians in making clinical decisions with efficiency during application.

Improved efficiency of urine cell image segmentation using droplet microfluidics technology

Recent advances in the recognition of biological samples using machine vision have made this technology increasingly important in research and detection. Image segmentation is an important step in this process. This study focuses on how to reduce the interference factors such as the overlap between different types (or within the same type) of urine cells according to microfluidics and improve the machine vision segmentation accuracy for cell images. In this study, we demonstrate that the platform can realize this hypothesis using urine cell image segmentation as an example application. We first discuss the reported urine cell droplet microfluidic chip system, which can realize the test conditions in which urine cells are encapsulated in the droplet and isolated from salt crystallization and/or bacteria and other urine-formed elements. Then, based on the analysis conditions set in the aforementioned experiment, the proportions of red blood cells, white blood cells, and squamous epithelial cells covered by various formed elements in the total urine cells in the same urine sample are measured. We simultaneously analyze the percentage of urine cells covered by salt crystallization and the incidence of overlapping between urine cells. Finally, the Otsu algorithm is used to segment the urine cell images encapsulated by the droplet and the urine cell images not encapsulated by the droplet, and the Dice, Jaccard, precision, and recall values are calculated. The results suggest that the method of encapsulating single cells based on droplets can improve the image segmentation effect without optimizing the algorithm.

Visual Interpretability in Computer-Assisted Diagnosis of Thyroid Nodules Using Ultrasound Images

BACKGROUND The number of studies on deep learning in artificial intelligence (AI)-assisted diagnosis of thyroid nodules is increasing. However, it is difficult to explain what the models actually learn in artificial intelligence-assisted medical research. Our aim is to investigate the visual interpretability of the computer-assisted diagnosis of malignant and benign thyroid nodules using ultrasound images. MATERIAL AND METHODS We designed and implemented 2 experiments to test whether our proposed model learned to interpret the ultrasound features used by ultrasound experts to diagnose thyroid nodules. First, in an anteroposterior/transverse (A/T) ratio experiment, multiple models were trained by changing the A/T ratio of the original nodules, and their classification, accuracy, sensitivity, and specificity were tested. Second, in a visualization experiment, class activation mapping used global average pooling and a fully connected layer to visualize the neural network to show the most important features. We also examined the importance of data preprocessing. RESULTS The A/T ratio experiment showed that after changing the A/T ratio of the nodules, the accuracy of the neural network model was reduced by 9.24-30.45%, indicating that our neural network model learned the A/T ratio information of the nodules. The visual experiment results showed that the nodule margins had a strong influence on the prediction of the neural network. CONCLUSIONS This study was an active exploration of interpretability in the deep learning classification of thyroid nodules. It demonstrated the neural network-visualized model focused on irregular nodule margins and the A/T ratio to classify thyroid nodules.

A Microfluidic Detection System for Bladder Cancer Tumor Cells

The clinical characteristics of excreted tumor cells can be found in the urine of bladder cancer patients, meaning the identification of tumor cells in urine can assist in bladder cancer diagnosis. The presence of white blood cells and epithelial cells in the urine interferes with the recognition of tumor cells. In this paper, a technique for detecting cancer cells in urine based on microfluidics provides a novel approach to bladder cancer diagnosis. The bladder cancer cell line (T24) and MeT-5A were used as positive bladder tumor cells and non-tumor cells, respectively. The practicality of the tumor cell detection system based on microfluidic cell chip detection technology is discussed. The tumor cell (T24) concentration was around 1 × 10⁴ to 300 × 10⁴ cells/mL. When phosphate buffer saline (PBS) was the diluted solution, the tumor cell detected rate was 63–71% and the detection of tumor cell number stability (coefficient of variation, CV%) was 6.7–4.1%, while when urine was the diluted solution, the tumor cell detected rate was 64–72% and the detection of tumor cell number stability (CV%) was 6.3–3.9%. In addition, both PBS and urine are tumor cell dilution fluid solutions. The sample was analyzed at a speed of 750 microns per hour. Based on the above experiments, a system for detecting bladder cancer cells in urine by microfluidic analysis chip technology was reported. The rate of recognizing bladder cancer cells reached 68.4%, and the speed reached 2 mL/h.