Inexperienced Doctors Research Articles

Deep learning based detection and classification of fetal lip in ultrasound images.

Fetal cleft lip is a common congenital defect. Considering the delicacy and difficulty of observing fetal lips, we have utilized deep learning technology to develop a new model aimed at quickly and accurately assessing the development of fetal lips during prenatal examinations. This model can detect ultrasound images of the fetal lips and classify them, aiming to provide a more objective prediction for the development of fetal lips. This study included 632 pregnant women in their mid-pregnancy stage, who underwent ultrasound examinations of the fetal lips, collecting both normal and abnormal fetal lip ultrasound images. To improve the accuracy of the detection and classification of fetal lips, we proposed and validated the Yolov5-ECA model. The experimental results show that, compared with the currently popular 10 models, our model achieved the best results in the detection and classification of fetal lips. In terms of the detection of fetal lips, the mean average precision (mAP) at 0.5 and mAP at 0.5:0.95 were 0.920 and 0.630, respectively. In the classification of fetal lip ultrasound images, the accuracy reached 0.925. The deep learning algorithm has accuracy consistent with manual evaluation in the detection and classification process of fetal lips. This automated recognition technology can provide a powerful tool for inexperienced young doctors, helping them to accurately conduct examinations and diagnoses of fetal lips.

Potential of digital chest radiography-based deep learning in screening and diagnosing pneumoconiosis: An observational study.

The diagnosis of pneumoconiosis is complex and subjective, leading to inevitable variability in readings. This is especially true for inexperienced doctors. To improve accuracy, a computer-assisted diagnosis system is used for more effective pneumoconiosis diagnoses. Three models (Resnet50, Resnet101, and DenseNet) were used for pneumoconiosis classification based on 1250 chest X-ray images. Three experienced and highly qualified physicians read the collected digital radiography images and classified them from category 0 to category III in a double-blinded manner. The results of the 3 physicians in agreement were considered the relative gold standards. Subsequently, 3 models were used to train and test these images and their performance was evaluated using multi-class classification metrics. We used kappa values and accuracy to evaluate the consistency and reliability of the optimal model with clinical typing. The results showed that ResNet101 was the optimal model among the 3 convolutional neural networks. The AUC of ResNet101 was 1.0, 0.9, 0.89, and 0.94 for detecting pneumoconiosis categories 0, I, II, and III, respectively. The micro-average and macro-average mean AUC values were 0.93 and 0.94, respectively. The accuracy and Kappa values of ResNet101 were 0.72 and 0.7111 for quadruple classification and 0.98 and 0.955 for dichotomous classification, respectively, compared with the relative standard classification of the clinic. This study develops a deep learning based model for screening and staging of pneumoconiosis is using chest radiographs. The ResNet101 model performed relatively better in classifying pneumoconiosis than radiologists. The dichotomous classification displayed outstanding performance, thereby indicating the feasibility of deep learning techniques in pneumoconiosis screening.

Suggestion for Objective Evaluation of Comparative Pulse Diagnosis.

Pulse is a method of Korean medicine diagnosis and is an important clue to detect the organs, nature, and progress of the disease. Pulse examination is included in the basic examination of Korean medicine doctors, but there is no standardized method for diagnosing pulse although the types and methods of the pulse taking are briefly described in the literature, making it difficult to spread the examination method. In this regard, I would like to propose an objective evaluation method. Although the importance of pulse examination and the method of pulse examination are known in the literature, it is difficult for undergraduate students or inexperienced Korean medicine doctors to access it, so in this paper a method of marking the size of the pulse power in the blank space for objective evaluation was devised and presented. The size of the pulse power should be indicated using the 1-cell, 3-cell, or 5-cell method according to the left and right wrists and the cun, guan and chi on both sides. The method of pulse diagnosis is an important diagnostic method as a verification process for making a Korean medical diagnosis. The remaining Korean medicine diagnostic methods, including pulse diagnosis, also need to undergo objectification. It is believed that the objectification of these diagnostic methods will lead to an improvement in the treatment rate of Korean medicine.

Disastrous cerebral and ocular vascular complications after cosmetic facial filler injections: a retrospective case series study

Soft tissue filler injections are among the most popular facial rejuvenation methods. Cerebral infarction and ophthalmic artery occlusion are rare and catastrophic complications, especially when facial cosmetic fillers are injected by inexperienced doctors. Radiologists and plastic surgeons need to increase their awareness of the complications associated with fillers, which allows early diagnosis and intervention to improve patient prognosis. Regarding the mechanism by which vascular occlusion occurs after facial filler injections, a retrograde embolic mechanism is currently the predominant theory. Numerous case reports have been presented regarding complications associated with injections of facial aesthetics. However, the small sample sizes of these studies did not allow for an adequate assessment of the clinical and imaging manifestations based on the location of the occlusion and the type of filler, and detailed elaboration of multiple cerebral infarctions is also lacking. Therefore, this study aimed to investigate the clinical and radiological features of severe cerebral and ocular complications caused by cosmetic facial filler injections. In addition, we discuss the pathogenesis, treatment, and prognosis of these patients. The clinical, computed tomography (CT), magnetic resonance imaging (MRI), and digital subtraction angiography (DSA) findings were described and analysed. Radiological examinations are crucial for demonstrating severe complications, and brain MRI is especially strongly suggested for patients with cosmetic filler-induced vision loss to identify asymptomatic cerebral infarctions. Extreme caution and care should be taken during facial injections by plastic surgeons.

Establishment of machine learning-based tool for early detection of pulmonary embolism

Background and objectivesPulmonary embolism (PE) is a complex disease with high mortality and morbidity rate, leading to increasing society burden. However, current diagnosis is solely based on symptoms and laboratory data despite its complex pathology, which easily leads to misdiagnosis and missed diagnosis by inexperienced doctors. Especially, CT pulmonary angiography, the gold standard method, is not widely available. In this study, we aim to establish a rapid and accurate screening model for pulmonary embolism using machine learning technology. Importantly, data required for disease prediction are easily accessed, including routine laboratory data and medical record information of patients. MethodsWe extracted features from patients' routine laboratory results and medical records, including blood routine, biochemical group, blood coagulation routine and other test results, as well as symptoms and medical history information. Samples with a feature loss rate greater than 0.8 were deleted from the original database. Data from 4723 cases were retained, 231 of which were positive for pulmonary embolism. 50 features were retained through the positive and negative statistical hypothesis testing which was used to build the predictive model. In order to avoid identification as majority-class samples caused by the imbalance of sample proportion, we used the method of Synthetic Minority Oversampling Technique (SMOTE) to increase the amount of information on minority samples. Five typical machine learning algorithms were used to model the screening of pulmonary embolism, including Support Vector Machines, Logistic Regression, Random Forest, XGBoost, and Back Propagation Neural Networks. To evaluate model performance, sensitivity, specificity and AUC curve were analyzed as the main evaluation indicators. Furthermore, a baseline model was established using the characteristics of the pulmonary embolism guidelines as a comparison model. ResultsWe found that XGBoost showed better performance compared to other models, with the highest sensitivity and specificity (0.99 and 0.99, respectively). Moreover, it showed significant improvement in performance compared to the baseline model (sensitivity and specificity were 0.76 and 0.76 respectively). More important, our model showed low missed diagnosis rate (0.46) and high AUC value (0.992). Finally, the calculation time of our model is only about 0.05 s to obtain the possibility of pulmonary embolism. ConclusionsIn this study, five machine learning classification models were established to assess the likelihood of patients suffering from pulmonary embolism, and the XGBoost model most significantly improved the precision, sensitivity, and AUC for pulmonary embolism screening. Collectively, we have established an AI-based model to accurately predict pulmonary embolism at early stage.

Evaluating Chatgpt As an Adjunct for Analyzing Challenging Case

Introduction Despite the rapid development of medicine, clinicians still faced with many challenging cases in practice. For challenging cases, most inexperienced doctors are unable to make a clear diagnosis in the first place, resulting in misdiagnosis and delayed disease. Recently, OpenAI released a new Chatbot model, called ChatGPT (Generation Pre training Converter), which is an artificial intelligence (AI) system that uses reinforcement learning model based on human feedback for training. There have been many research reports on the application of ChatGPT in the medical field, and it has been proven to have a wide range of clinical applications. In our study, we aimed to assess the accuracy and reliability of ChatGPT in diagnosing and treating challenging case. Methods We provided 2 challenging cases. Case 1 provided a medical history and auxiliary test results for ChatGPT, and then ChatGPT answered the possible diagnosis, diagnostic basis, further auxiliary examination and treatment. Case 2 provides negative symptoms and auxiliary test results for ChatGPT, which answers all possible diagnoses. Conduct a comprehensive evaluation and scoring of the results of ChatGPT responses by 5 senior physicians. A score of 10 is the perfect score. Results The analysis results of 2 challenging cases by ChatGPT were comprehensively evaluated by senior experienced professional doctors and scored 9 points. We found that ChatGPT's analysis of challenging case is comprehensive. Especially for the diagnosis and treatment of diseases, ChatGPT analyzes all possible diagnoses and provides accurate treatment plans through the provided medical history, auxiliary examination results, and treatment processes. ChatGPT can directly interpret the auxiliary examination results, such as high level of procalcitonin (PCT), abnormal kidney function, and with negative results can rule out certain diseases, such as ANA, ENA, and ANCA negative, ruling out the possibility of autoimmune diseases. ChatGPT can summarize all the diagnoses including rare diseases, based on the results of an auxiliary test, and explain the cause of the disease and the pathogenesis in detail. Conclusions Our study shows that ChatGPT can provide important reference value as an adjunct to clinicians in challenging case analysis. In particular, it can assist inexperienced doctors in remote areas and underdeveloped medical conditions to make clear diagnosis and treatment as soon as possible, avoid unnecessary referral to higher-level hospitals, reduce medical burden, and reduce waste of medical resources. In the medical field, ChatGPT needs more research to evaluate in clinical applications and become a better adjunct tool in clinical work.

A Novel pedicle screw placement Surgery Based on Integration of Surgical guides and Augmented Reality.

Augmented reality is a new technology that, when applied to spinal surgery, offers the potential for efficient, safe, and accurate placement of pedicle screws. This study investigated whether augmented reality combined with a guide board improved the safety and accuracy of pedicle screw placement compared to traditional free-hand screw placement. Four trainers were divided into augmented reality navigation and free-hand groups. Each group consisted of a novice and an experienced spine surgeon. A total of 80 pedicle screws were implanted. First, the AR group reconstructed the 3D model and planned the screw insertion route according to the CT data of L2 lumbar vertebrae. Next, the Microsoft HoloLens™ 2 was used to identify the vertebral model, and the planned virtual path was superimposed on the real cone model. Next, the screw was placed according to the projected trajectory. Finally, Micron Tracker was used to measure the deviation of screws from the preoperatively planned trajectory, and pedicle screws were evaluated using the Gertzbein-Robbins scale. In the augmented reality group, the linear deviation of experienced doctors and novices was 1.59 ± 0.39 mm and 1.73 ± 0.52 mm, and the angle deviation was 2.72 ± 0.61° and 2.87 ± 0.63°, respectively. In the free-hand group, the linear deviation of experienced doctors and novices was 2.88 ± 0.58mm and 5.25 ± 0.62mm, and the angle deviation was 4.41 ± 1.18° and 7.15 ±1.45°, respectively. The screw accuracy rate was 97.5% in the augmented reality navigation group and 77.5% in the free-hand group. Augmented reality navigation improves the accuracy and safety of pedicle screw implantation compared with the traditional free-hand method and can assist inexperienced doctors in completing the surgery.

Convolutional-neural-network-based radiographs evaluation assisting in early diagnosis of the periodontal bone loss via periapical radiograph

Background/PurposeThe preciseness of detecting periodontal bone loss is examiners dependent, and this leads to low reliability. The need for automated assistance systems on dental radiographic images has been increased. To the best of our knowledge, no studies have quantitatively and automatically staged periodontitis using dental periapical radiographs. The purpose of this study was to evaluate periodontal bone loss and periodontitis stage on dental periapical radiographs using deep convolutional neural networks (CNNs). Materials and methods336 periapical radiographic images (teeth: 390) between January 2017 and December 2019 were collected and de-identified. All periapical radiographic image datasets were divided into training dataset (n = 82, teeth: 123) and test dataset (n = 336, teeth: 390). For creating an optimal deep CNN algorithm model, the training datasets were directly used for the segmentation and individual tooth detection. To evaluate the diagnostic power, we calculated the degree of alveolar bone loss deviation between our proposed method and ground truth, the Pearson correlation coefficients (PCC), and the diagnostic accuracy of the proposed method in the test datasets. ResultsThe periodontal bone loss degree deviation between our proposed method and the ground truth drawn by the three periodontists was 6.5 %. In addition, the overall PCC value of our proposed system and the periodontists’ diagnoses was 0.828 (P < 0.01). The total diagnostic accuracy of our proposed method was 72.8 %. The diagnostic accuracy was highest for stage III (97.0 %). ConclusionThis tool helps with diagnosis and prevents omission, and this may be especially helpful for inexperienced younger doctors and doctors in underdeveloped countries. It could also dramatically reduce the workload of clinicians and timely access to periodontist care for people requiring advanced periodontal treatment.

Respiratory sound-base disease classification and characterization with deep/machine learning techniques

Respiratory diseases (RDs) are a leading cause of death globally, with over 490,000 deaths in the EU and the US alone in 2017. Early detection of these diseases is crucial for improving treatment options and prolonging survival. However, diagnosing respiratory diseases can be challenging due to various factors such as physician shortages, financial constraints, and inexperienced doctors. To address these challenges, this paper presents a novel two-stage approach for the early detection and accurate classification of respiratory diseases using sound processing, machine learning, and deep learning techniques. The proposed approach utilizes a voice recorder and the internet to create a numerical feature vector using statistical properties of sound files from various datasets. The feature vector goes through a 2-stage feature elimination process for deep learning models and a 3-stage feature elimination process for machine learning models. The resulting feature vector is then fed into a supervised model to classify patients as healthy or suffering from a respiratory disease. A second classifier is then used to characterize the specific respiratory disease. Experiments with various machine/deep learning algorithms demonstrate that the proposed model outperforms existing methods in the literature in terms of accuracy and F1-score. The model achieves an accuracy score of 0.998 and an F1-score of 0.998 for distinguishing between healthy and respiratory disease-affected patients and an accuracy of 0.947 and an F1-score of 0.94 for categorizing ‘unhealthy’ records into COPD, asthma, and other respiratory diseases. The proposed approach has significant potential low-cost and efficient tool aiding early detection and classification. Simple SummaryEvery year, millions of people die as a result of respiratory diseases, particularly chronic obstructive pulmonary disease (COPD), asthma, and pneumonia. Early diagnosis is critical for reducing morbidity and mortality from respiratory diseases because the best treatment options are frequently discovered in the early stages. Even though there are numerous studies and screening programs that can aid in the early detection of these diseases, there are still significant gaps in screening tools. In this study, we investigated the prospects of a two-stage pipeline for an automated system for identifying respiratory diseases, as well as the potential applications of sound processing, machine learning, and deep learning techniques in respiratory diseases within the context of respiratory sounds. In the first stage, a feature vector was extracted directly from the respiratory sound, and it was determined whether the feature vector was related to a patient suffering from a respiratory disease using supervised machine learning techniques. In the second stage, the method can also distinguish between asthma, chronic obstructive pulmonary disease, and other respiratory diseases (Pneumonia, URTI, Bronchiectasis, Bronchiolitis, LRTI). Preliminary findings show that deep/machine learning-based methods have the potential to detect and classify respiratory illnesses in real time. The proposed model has great potential as a low-cost, easily accessible tool for screening and detecting respiratory diseases, thereby closing the gap in early diagnosis.

Modelling Differential Diagnosis of Febrile Diseases with Fuzzy Cognitive Map.

The report of the World Health Organization (WHO) about the poor accessibility of people living in low-to-middle-income countries to medical facilities and personnel has been a concern to both professionals and nonprofessionals in healthcare. This poor accessibility has led to high morbidity and mortality rates in tropical regions, especially when such a disease presents itself with confusable symptoms that are not easily differentiable by inexperienced doctors, such as those found in febrile diseases. This prompted the development of the fuzzy cognitive map (FCM) model to serve as a decision-support tool for medical health workers in the diagnosis of febrile diseases. With 2465 datasets gathered from four states in the febrile diseases-prone regions in Nigeria with the aid of 60 medical doctors, 10 of those doctors helped in weighting and fuzzifying the symptoms, which were used to generate the FCM model. Results obtained from computations to predict diagnosis results for the 2465 patients, and those diagnosed by the physicians on the field, showed an average of 87% accuracy for the 11 febrile diseases used in the study. The number of comorbidities of diseases with varying degrees of severity for most patients in the study also covary strongly with those found by the physicians in the field.

Differential diagnosis of secondary hypertension based on deep learning

Secondary hypertension is associated with higher risks of target organ damage and cardiovascular and cerebrovascular disease events. Early aetiology identification can eliminate aetiologies and control blood pressure. However, inexperienced doctors often fail to diagnose secondary hypertension, and comprehensively screening for all causes of high blood pressure increases health care costs. To date, deep learning has rarely been involved in the differential diagnosis of secondary hypertension. Relevant machine learning methods cannot combine textual information such as chief complaints with numerical information such as the laboratory examination results in electronic health records (EHRs), and the use of all features increases health care costs. To reduce redundant examinations and accurately identify secondary hypertension, we propose a two-stage framework that follows clinical procedures. The framework carries out an initial diagnosis process in the first stage, on which basis patients are recommended for disease-related examinations, followed by differential diagnoses of different diseases based on the different characteristics observed in the second stage. We convert the numerical examination results into descriptive sentences, thus blending textual and numerical characteristics. Medical guidelines are introduced through label embedding and attention mechanisms to obtain interactive features. Our model was trained and evaluated using a cross-sectional dataset containing 11,961 patients with hypertension from January 2013 to December 2019. The F1 scores of our model were 0.912, 0.921, 0.869 and 0.894 for primary aldosteronism, thyroid disease, nephritis and nephrotic syndrome and chronic kidney disease, respectively, which are four kinds of secondary hypertension with high incidence rates. The experimental results show that our model can powerfully use the textual and numerical data contained in EHRs to provide effective decision support for the differential diagnosis of secondary hypertension.

A novel cap-assisted endoscopic resection device for resection of a rectal lesion in an in vitro pig model.

In a previous study, we proved that modified cap-assisted endoscopic mucosal resection (mEMR-C) is non-inferior to endoscopic submucosal dissection (ESD) for rectal neuroendocrine tumors smaller than 10 mm with a similar complete resection rate. In addition, mEMR-C had a shorter procedure duration time and lower hospitalization costs [1]. However, a snare needs to be precisely placed into the inner groove of the cap during the mEMR-C procedure, which is time-consuming and difficult for an inexperienced doctor. To solve this problem, we proposed a novel cap-assisted endoscopic resection device for resecting of a rectal lesion that might make the procedure much simpler and easier. In the current study, we initially explored the feasibility of this device for endoscopic resection in an in vitro pig model.

Preparing for Reality: A Randomized Trial on Immersive Virtual Reality for Bronchoscopy Training

Background: Bronchoscopy is an essential procedure in the diagnosis and treatment of pulmonary diseases. However, the literature suggests that distractions affect the quality of bronchoscopy and affect inexperienced doctors more than experienced. Objectives: The objective of the study was as follows: does simulation-based bronchoscopy training with immersive virtual reality (iVR) improve the doctors’ ability to handle distractions and thereby increase the quality, measured in procedure time, structured progression score, diagnostic completeness (%), and hand motor movements of a diagnostic bronchoscopy in a simulated scenario. Exploratory outcomes were heart rate variability and a cognitive load questionnaire (Surg-TLX). Methods: Participants were randomized. The intervention group practiced in an iVR environment with a head-mounted display (HMD) while using the bronchoscopy simulator, while the control group trained without the HMD. Both groups were tested in the iVR environment using a scenario with distractions. Results: 34 participants completed the trial. The intervention group scored significantly higher in diagnostic completeness (100 i.q.r. 100–100 vs. 94 i.q.r. 89–100, p value = 0.03) and structured progress (16 i.q.r. 15–18 vs. 12 i.q.r. 11–15, p value 0.03) but not in procedure time (367 s standard deviation [SD] 149 vs. 445 s SD 219, p value = 0.06) or hand motor movements (−1.02 i.q.r. −1.03–[−1.02] versus −0.98 i.q.r. −1.02–[–0.98], p value = 0.27). The control group had a tendency toward a lower heart rate variability (5.76 i.q.r. 3.77–9.06 vs. 4.12 i.q.r. 2.68–6.27, p = 0.25). There was no significant difference in total Surg-TLX points between the two groups. Conclusion: iVR simulation training increases the quality of diagnostic bronchoscopy in a simulated scenario with distractions compared with conventional simulation-based training.

Introducing the "SIMline"-A Simulation Course in the Management of Severe Burns as a Tool in Undergraduate Medical Education.

Background: Management of burn injuries presents a complex and challenging situation for medical staff, especially for inexperienced young doctors. However, training on how to treat burn victims in the clinical setting is rarely taught in undergraduate medical classes. We have created the "SIMline", a simulation training program explicitly designed for coaching medical students in burn management. Methods: A total of 43 students participated in the "SIMline" course, which took place at the training facility at the Medical University of Graz, between 2018 and 2019. The course provided theoretical classes, practical exercises, and a full-scale care process simulation training. The learning progress of the students was monitored via a formative integrated test. Results: Students showed great progress throughout the course of the "SIMline" program, as their test scores improved by an average of 88%. The passing rate was 0% at the first exam (prior to course) as compared to 87% at the final exam, taken after the training. Conclusions: Comprehensive practical training programs in burn care are underrepresented in medical education. The "SIMline" course presents a novel and effective approach in training medical students in burn management. However, follow-up evaluation is necessary to confirm long-term educational benefits.

Requirement of image standardization for AI-based macroscopic diagnosis for surgical specimens of gastric cancer.

The pathological diagnosis of surgically resected gastric cancer involves both a macroscopic diagnosis by gross observation and a microscopic diagnosis by microscopy. Macroscopic diagnosis determines the location and stage of the disease and the involvement of other organs and surgical margin. Lesion recognition is, thus, an important diagnostic step that requires a skilled pathologist. Nonetheless, artificial intelligence (AI) technologies could allow even inexperienced doctors and laboratory technicians to examine surgically resected specimens without the need for pathologists. However, organ imaging conditions vary across hospitals, and an AI algorithm created in one setting may not work properly in another. Thus, we identified and standardized factors affecting the quality of pathological macroscopic images, which could further affect lesion identification using AI. We examined necessary image standardization for developing cancer detection AI for surgically resected gastric cancer by changing the following imaging conditions: focus, resolution, brightness, and contrast. Regarding focus, brightness, and contrast, the farther away the test data were from the training macro-image, the less likely the inference was to be correct. Little change was observed for resolution, even with differing conditions for the training and test data. Regarding focus, brightness, and contrast, there were conditions appropriate for AI. Contrast, in particular, was far from the conditions appropriate for humans. Standardizing focus, brightness, and contrast is important in the development of AI methodologies for lesion detection in surgically resected gastric cancer. This standardization is essential for AI to be implemented across hospitals.

A deep learning model for detection of leukocytes under various interference factors

The accurate detection of leukocytes is the basis for the diagnosis of blood system diseases. However, diagnosing leukocyte disorders by doctors is time-consuming and requires extensive experience. Automated detection methods with high accuracy can improve detection efficiency and provide recommendations to inexperienced doctors. Current methods and instruments either fail to automate the identification process fully or have low performance and need suitable leukocyte data sets for further study. To improve the current status, we need to develop more intelligent strategies. This paper investigates fulfilling high-performance automatic detection for leukocytes using a deep learning-based method. We established a new dataset more suitable for leukocyte detection, containing 6273 images (8595 leukocytes) and considering nine common clinical interference factors. Based on the dataset, the performance evaluation of six mainstream detection models is carried out, and a more robust ensemble model is proposed. The mean of average precision (mAP) @IoU = 0.50:0.95 and mean of average recall (mAR)@IoU = 0.50:0.95 of the ensemble model on the test set are 0.853 and 0.922, respectively. The detection performance of poor-quality images is robust. For the first time, it is found that the ensemble model yields an accuracy of 98.84% for detecting incomplete leukocytes. In addition, we also compared the test results of different models and found multiple identical false detections of the models, then provided correct suggestions for the clinic.

TCM Prescription Generation via Knowledge Source Guidance Network Combined with Herbal Candidate Mechanism.

Traditional Chinese medicine (TCM) prescriptions have made great contributions to the treatment of diseases and health preservation. To alleviate the shortage of TCM resources and improve the professionalism of automatically generated prescriptions, this paper deeply explores the connection between symptoms and herbs through deep learning technology, and realizes the automatic generation of TCM prescriptions. Particularly, this paper considers the significance of referring to similar underlying prescriptions as herbal candidates in the TCM prescribing process. Moreover, this paper incorporates the idea of referring to the potential guidance information of corresponding prescriptions when model extracts symptoms representations. To provide a reference for inexperienced young TCM doctors when they prescribe, this paper proposes a dual-branch guidance strategy combined with candidate attention model (DGSCAM) to automatically generate TCM prescriptions based on symptoms text. The format of the data used this paper is the "symptoms-prescription" data pair. The specific method is as follows. First, DGSCAM realizes the extraction of key information of prescription-guided symptoms through a dual-branch network. Then, herbal candidates in the form of prescriptions that can treat symptoms are proposed in view of the compatibility knowledge of TCM prescriptions. To our knowledge, this is the first attempt to use prescriptions as herbal candidates in the prescription generation process. We conduct extensive experiments on a mixed public and clinical dataset, achieving 37.39% precision, 25.04% recall, and 29.99% F1 score, with an average doctor score of 7.7 out of 10. The experimental results show that our proposed model is valid and can generate more specialized TCM prescriptions than the baseline models. The DGSCAM developed by us has broad application scenarios and greatly promotes the research on intelligent TCM prescribing.

The Application of Ultrasound Image-Based Radiomics in the Diagnosis of Mild Carpal Tunnel Syndrome.

The ultrasound diagnosis of mild carpal tunnel syndrome (CTS) is challenging. Radiomics can identify image information that the human eye cannot recognize. The purpose of our study was to explore the value of ultrasound image-based radiomics in the diagnosis of mild CTS. This retrospective study included 126 wrists in the CTS group and 88 wrists in the control group. The radiomics features were extracted from the cross-sectional ultrasound images at the entrance of median nerve carpal tunnel, and the modeling was based on robust features. Two radiologists with different experiences diagnosed CTS according to two guidelines. The area under receiver (AUC) operating characteristic curve, sensitivity, specificity, and accuracy were used to evaluate the diagnostic efficacy of the two radiologists and the radiomics model. According to guideline one, the AUC values of the two radiologists for CTS were 0.72 and 0.67, respectively; according to guideline two, the AUC were 0.73 and 0.68, respectively. The radiomics model achieved the best accuracy when 16 important robust features were selected. The AUC values of training set and test set were 0.92 and 0.90, respectively. The radiomics label based on ultrasound images had excellent diagnostic efficacy for mild CTS. It is expected to help radiologists to identify early CTS patients as soon as possible, especially for inexperienced doctors.

Demarcation Line Determination for Diagnosis of Gastric Cancer Disease Range Using Unsupervised Machine Learning in Magnifying Narrow-Band Imaging.

Background and Aims: It is important to determine an accurate demarcation line (DL) between the cancerous lesions and background mucosa in magnifying narrow-band imaging (M-NBI)-based diagnosis. However, it is difficult for novice endoscopists. We aimed to automatically determine the accurate DL using a machine learning method. Methods: We used an unsupervised machine learning approach to determine the DLs. Our method consists of the following four steps: (1) an M-NBI image is segmented into superpixels using simple linear iterative clustering; (2) the image features are extracted for each superpixel; (3) the superpixels are grouped into several clusters using the k-means method; and (4) the boundaries of the clusters are extracted as DL candidates. The 23 M-NBI images of 11 cases were used for performance evaluation. The evaluation investigated the similarity of the DLs identified by endoscopists and our method, and the Euclidean distance between the two DLs was calculated. For the single case of 11 cases, the histopathological examination was also conducted to evaluate the proposed system. Results: The average Euclidean distances for the 11 cases were 10.65, 11.97, 7.82, 8.46, 8.59, 9.72, 12.20, 9.06, 22.86, 8.45, and 25.36. The results indicated that the proposed method could identify similar DLs to those identified by experienced doctors. Additionally, it was confirmed that the proposed system could generate pathologically valid DLs by increasing the number of clusters. Conclusions: Our proposed system can support the training of inexperienced doctors as well as enrich the knowledge of experienced doctors in endoscopy.

A Residual Fusion Network for Osteosarcoma MRI Image Segmentation in Developing Countries.

Among primary bone cancers, osteosarcoma is the most common, peaking between the ages of a child's rapid bone growth and adolescence. The diagnosis of osteosarcoma requires observing the radiological appearance of the infected bones. A common approach is MRI, but the manual diagnosis of MRI images is prone to observer bias and inaccuracy and is rather time consuming. The MRI images of osteosarcoma contain semantic messages in several different resolutions, which are often ignored by current segmentation techniques, leading to low generalizability and accuracy. In the meantime, the boundaries between osteosarcoma and bones or other tissues are sometimes too ambiguous to separate, making it a challenging job for inexperienced doctors to draw a line between them. In this paper, we propose using a multiscale residual fusion network to handle the MRI images. We placed a novel subnetwork after the encoders to exchange information between the feature maps of different resolutions, to fuse the information they contain. The outputs are then directed to both the decoders and a shape flow block, used for improving the spatial accuracy of the segmentation map. We tested over 80,000 osteosarcoma MRI images from the PET-CT center of a well-known hospital in China. Our approach can significantly improve the effectiveness of the semantic segmentation of osteosarcoma images. Our method has higher F1, DSC, and IOU compared with other models while maintaining the number of parameters and FLOPS.