AI Diagnosis Research Articles

Diagnosing Necrotizing Enterocolitis via Fine-Grained Visual Classification.

Necrotizing Enterocolitis (NEC) is a devastating condition affecting prematurely born neonates. Reviewing Abdominal X-rays (AXRs) is a key step in NEC diagnosis, staging and treatment decision-making, but poses significant challenges due to the subtle, difficult-to-identify radiological signs of the disease. In this paper, we propose AIDNEC - AI Diagnosis of NECrotizing enterocolitis, a deep learning method to automatically detect and stratify the severity (surgical or medical) of NEC from no pathology in AXRs. The model is trainable end-to-end and integrates a Detection Transformer and Graph Convolution modules for localizing discriminative areas in AXRs, used to formulate subtle local embeddings. These are then combined with global image features to perform Fine-Grained Visual Classification (FGVC). We evaluate AIDNEC on our GOSH NEC dataset of 1153 images from 334 patients, achieving 79.7% accuracy in classifying NEC against No Pathology. AIDNEC outperforms the backbone by 2.6%, FGVC models by 2.5% and CheXNet by 4.2%, with statistically significant (two-tailed p 0.05) improvements, while providing meaningful discriminative regions to support the classification decision. Additional validation in the publicly available Chest X-ray14 dataset yields comparable performance to state-of-the-art methods, illustrating AIDNEC's robustness in a different X-ray classification task.

6937 A Rare Case of Pembrolizumab-induced Multiple Endocrinopathies Presenting with Acute Adrenal Crisis Secondary to Isolated ACTH Deficiency - A Case Report

Abstract Disclosure: R. Abdelmasih: None. W. Pan: None. Introduction: Pembrolizumab is an immune checkpoint inhibitor (ICI) that blocks programmed cell death receptor 1 (PD-1) which took a massive leap in the oncology world. However, we are still understanding more about its immune-related adverse events (irAEs). Here we report a case of adrenal crisis due to pembrolizumab-induced hypophysitis and adrenalitis. Case Description: A 74-year-old male with a history of left renal cell carcinoma post left nephrectomy 5 years ago, complicated with lung metastasis for which Pembrolizumab was started. 5 months after, patient developed hypothyroidism and was started on Levothyroxine. 3 months later, Patient presented with lethargy and hypotension. Physical exam revealed hypotension (MAP of 55), fever, and acute encephalopathy. AM cortisol 4.2 ug/dL, Patient was started on IV pressors and a stress dose of Hydrocortisone. Hemodynamics improved, and Hydrocortisone was tapered down. ACTH stimulation test revealed a subnormal response of cortisol 0.2, 1.7, and 2.7 ug/dL at 0, 30, and 60 minutes respectively in the setting of very low ACTH of < 1.5 pg/ml, other pituitary hormones were checked, revealing normal insulin-like growth factor was 116 ng/mL, LH 15.29 mIU/mL, and FSH 27.45 mIU/mL. TSH was high 15 mIU/mL and free T4 was 1.4 ng/dl, levothyroxine dose was increased, and hydrocortisone was maintained at 10 in am and 5 in pm with excellent response. Discussion:ICIs inhibit the immunological pathways that control T-cell activation, leading to multiple immune-related endocrinopathies. Thyroiditis is the most reported one (30%), though anti-PD1-induced hypophysitis is exceptionally rare (< 1%). To date, the pathophysiology of pembrolizumab-induced central AI is not clearly understood, it’s presumed that it causes lymphocytic hypophysitis often presenting with isolated ACTH deficiency. Also, Pembrolizumab-induced primary adrenal insufficiency has seldom been reported and not yet understood, interestingly, anti-steroidogenic enzyme antibodies were absent in those few most reported cases which suggests a different mechanism compared to traditional primary AI. We suggest that our patient had combined primary and secondary adrenal insufficiency based on ACTH stim test result with very low cortisol responses in the setting of very low ACTH, which we believe is what led to such severe presentation of adrenal crisis. Diagnosis of PD1-induced AI can be challenging due to the overlapping of advanced malignancy and AI presentation. A higher level of suspicion is prompted especially in patients who experienced other endocrine irAEs (as our patient who had PDL-1-induced hypothyroidism). We present this case to shed light on this rare yet potentially fatal endocrine adverse event that can be reversed by prompt diagnosis and timely management due to the increasing use of pembrolizumab. More cases are needed for a better understanding of such a rare adverse event. Presentation: 6/2/2024

A Comprehensive Review of AI Diagnosis Strategies for Age-Related Macular Degeneration (AMD).

The rapid advancement of computational infrastructure has led to unprecedented growth in machine learning, deep learning, and computer vision, fundamentally transforming the analysis of retinal images. By utilizing a wide array of visual cues extracted from retinal fundus images, sophisticated artificial intelligence models have been developed to diagnose various retinal disorders. This paper concentrates on the detection of Age-Related Macular Degeneration (AMD), a significant retinal condition, by offering an exhaustive examination of recent machine learning and deep learning methodologies. Additionally, it discusses potential obstacles and constraints associated with implementing this technology in the field of ophthalmology. Through a systematic review, this research aims to assess the efficacy of machine learning and deep learning techniques in discerning AMD from different modalities as they have shown promise in the field of AMD and retinal disorders diagnosis. Organized around prevalent datasets and imaging techniques, the paper initially outlines assessment criteria, image preprocessing methodologies, and learning frameworks before conducting a thorough investigation of diverse approaches for AMD detection. Drawing insights from the analysis of more than 30 selected studies, the conclusion underscores current research trajectories, major challenges, and future prospects in AMD diagnosis, providing a valuable resource for both scholars and practitioners in the domain.

P-047 Sperm moving pattern analysis with artificial intelligence solution predicts aneuploidy ratios of zygotes

Abstract Study question Can an analysis of sperm movements with artificial intelligence predict aneuploidy ratios of zygotes? Summary answer And sperm movement patterns analyzed by AI diagnosis predicted the aneuploidy ratios of zygotes. What is known already Sperm motilities present a specific pattern like strength, strength with curving, curving only and none progressive et al. Sperm movement patterns are determined by the sperm neck protein which is the centromere protein. Also, this centromere protein of the sperms was involved in the chromosome segregation in the zygote. Study design, size, duration A total of 8,000 clinical participants were recruited by CHA Infertility Center Seoul Station and 5 videos were taken from the semen sample of each participant. Machine learning with sperm moving clip for the diagnosis algorithm predicted aneuploidy ratios of fertilized embryos. Participants/materials, setting, methods We used YOLO_v8 and EfficientNet algorithm on the extracted sperm movement trajectory image to classify sperm into classes according to their motion pattern. This information was combined with the age of the patient to predict the possibility of infertility by analyzing the association with aneuploidy ratios of zygotes using deep learning model. Main results and the role of chance Deep learning AI algorithm utilized diagnosis for unexplained male infertility. No strength moving sperm pattern has shown association with male origin aneuploidy after fertilization. Especially high ratios of cycle moving sperm significantly increased the aneuploidy ratios in the zygote. But paternal aging has no correlation with aneuploidy ratios in the embryo. AI diagnosis have a 0.9 accuracy score. It is expected that the AI solution generated through this system can help overcome the problem of finding the cause of male infertility by predicting aneuploidy ratios of zygotes. Limitations, reasons for caution This study is designed for and performed with a minimum number of sperm move data. For clinical application, further mass data was set-up to increase accuracy of diagnosis. Wider implications of the findings Our AI solution was constructed to predict the possibility of aneuploidy ratios in the zygote by analyzing and classifying the movement patterns of sperm. Therefore, sperm movement-based AI diagnosis solution can be applied to define unexplained male infertility. Trial registration number non-clinical trials

Small RNA signatures of acute ischemic stroke in L1CAM positive extracellular vesicles

L1CAM-positive extracellular vesicles (L1EV) are an emerging biomarker that may better reflect ongoing neuronal damage than other blood-based biomarkers. The physiological roles and regulation of L1EVs and their small RNA cargoes following stroke is unknown. We sought to characterize L1EV small RNAs following stroke and assess L1EV RNA signatures for diagnosing stroke using weighted gene co-expression network analysis and random forest (RF) machine learning algorithms. Interestingly, small RNA sequencing of plasma L1EVs from patients with stroke and control patients (n = 28) identified micro(mi)RNAs known to be enriched in the brain. Weighted gene co-expression network analysis (WGCNA) revealed small RNA transcript modules correlated to diagnosis, initial NIH stroke scale, and age. L1EV RNA signatures associated with the diagnosis of AIS were derived from WGCNA and RF classification. These small RNA signatures demonstrated a high degree of accuracy in the diagnosis of AIS with an area under the curve (AUC) of the signatures ranging from 0.833 to 0.932. Further work is necessary to understand the role of small RNA L1EV cargoes in the response to brain injury, however, this study supports the utility of L1EV small RNA signatures as a biomarker of stroke.

Investigating Factors Shaping Future Doctors' Willingness to Adopt AI Diagnosis Support Systems

Investigating Factors Shaping Future Doctors' Willingness to Adopt AI Diagnosis Support Systems

Combined detection of serum IL-6 and CEA contributes to the diagnosis of lung adenocarcinoma in situ.

Effective discrimination of lung adenocarcinoma (LUAD) in situ (AIS) from benign pulmonary nodules (BPN) is critical for the early diagnosis of AIS. Our pilot study in a small cohort of 90 serum samples has shown that serum interleukin 6 (IL-6) detection can distinguish AIS from BPN and health controls (HC). In this study, we intend to comprehensively define the diagnostic value of individual and combined detection of serum IL-6 related to the traditional tumor markers carcinoembryonic antigen (CEA) and cytokeratin 19 fragment (CYFRA21-1) for AIS. The diagnostic performance of serum IL-6 along with CEA and CYFRA21-1 were evaluated in a large cohort of 300 serum samples by a chemiluminescence immunoassay and an electrochemiluminescence immunoassay. A training set comprised of 65 AIS, 65 BPN, and 65 HC samples was used to develop the predictive model for AIS. Data obtained from an independent validation set was applied to evaluate and validate the predictive model. In the training set, the levels of serum IL-6 and CEA in the AIS group were significantly higher than those in the BPN/HC group (P < 0.05). There was no significant difference in serum CYFRA21-1 levels between the AIS group and the BPN/HC group (P> 0.05). Serum IL-6 and CEA levels for AIS patients showed an area under the curve (AUC) of 0.622 with 23.1% sensitivity at 90.7% specificity, and an AUC of 0.672 with 24.6% sensitivity at 97.6% specificity, respectively. The combination of serum IL-6 and CEA presented an AUC of 0.739, with 60.0% sensitivity at 95.4% specificity. The combination of serum IL-6 and CEA showed an AUC of 0.767 for AIS patients, with 57.1% sensitivity at 91.4% specificity in the validation set. IL-6 shows potential as a prospective serum biomarker for the diagnosis of AIS, and the combination of serum IL-6 with CEA may contribute to increased accuracy in AIS diagnosis. However, it is worth noting that further research is still necessary to validate and optimize the diagnostic efficacy of these biomarkers and to address potential sensitivity limitations.

How Accurate Are Anatomical Surface Topography Parameters in Indicating the Presence of a Scoliosis?

Retrospective analysis of a longitudinal cohort. To identify the presence of scoliosis from surface data. Identifying AIS can be difficult. Screening is not universal for reasons including high false positive and negative rates. These difficulties can lead to some adolescents missing out on bracing. Logistic regression analysis of ISIS2 surface topography images only was performed. The x,y positions of the shoulders (Sh), axillae (Ax), waist (waist) and the x,y,z positions of the most prominent points over the posterior torso (scap) were used for the thoracic, thoracolumbar/lumbar and whole spine. The models were used to identify the presence of a 20-degree or larger scoliosis. Differences in the position of the landmarks were analyzed comparing left and right, with the suffix "Ht" representing a difference in the y coordinate, "Off" the x coordinate, and "Depth," the z coordinate. Model accuracy was assessed as both percentages and ROC curves with the coefficients as odds ratios. There were 1283 images (1015 females and 268 males) all with a diagnosis of AIS. The models identified scoliosis in the thoracic spine with an 83% accuracy (AUC 0.91), thoracolumbar/lumbar spine with 74% accuracy (AUC 0.76), and whole spine with 80% accuracy (AUC 0.88). Significant parameters were AxDiffHt, AxDiffOff, WaistDiffHt, ScapDiffOff, and ScapDiffHt for the thoracic curves, AxDiffHt, AxDiffOff, WaistDiffHt for the thoracolumbar/lumbar curves, and AxDiffHt, AxDiffOff, WaistDiffHt and ScapDiffHt for the whole spine. The use of fixed anatomical points around the torso, analyzed using logistic regression, has a high accuracy for identifying curves in the thoracic, thoracolumbar/lumbar, and whole spines. While coming from surface topography images, the results raise the future use of digital photography as a tool for the identification of small scoliosis without using other imaging techniques.

A Multimode Microfiber Specklegram Biosensor for Measurement of CEACAM5 through AI Diagnosis.

Carcinoembryonic antigen (CEACAM5), as a broad-spectrum tumor biomarker, plays a crucial role in analyzing the therapeutic efficacy and progression of cancer. Herein, we propose a novel biosensor based on specklegrams of tapered multimode fiber (MMF) and two-dimensional convolutional neural networks (2D-CNNs) for the detection of CEACAM5. The microfiber is modified with CEA antibodies to specifically recognize antigens. The biosensor utilizes the interference effect of tapered MMF to generate highly sensitive specklegrams in response to different CEACAM5 concentrations. A zero mean normalized cross-correlation (ZNCC) function is explored to calculate the image matching degree of the specklegrams. Profiting from the extremely high detection limit of the speckle sensor, variations in the specklegrams of antibody concentrations from 1 to 1000 ng/mL are measured in the experiment. The surface sensitivity of the biosensor is 0.0012 (ng/mL)-1 within a range of 1 to 50 ng/mL. Moreover, a 2D-CNN was introduced to solve the problem of nonlinear detection surface sensitivity variation in a large dynamic range, and in the search for image features to improve evaluation accuracy, achieving more accurate CEACAM5 monitoring, with a maximum detection error of 0.358%. The proposed fiber specklegram biosensing scheme is easy to implement and has great potential in analyzing the postoperative condition of patients.

Intraoperative navigation increases the projected lifetime cancer risk in patients undergoing surgery for adolescent idiopathic scoliosis

BACKGROUND CONTEXTAdolescent idiopathic scoliosis (AIS) is a common condition, often requiring surgical correction. Computed tomography (CT) based navigation technologies, which rely on ionizing radiation, are increasingly being utilized for surgical treatment. Although this population is highly vulnerable to radiation, given their age and female predominance, there is little available information elucidating modeled iatrogenic cancer risk. PURPOSETo model lifetime cancer risk associated with the use of intraoperative CT-based navigation for surgical treatment of AIS. STUDY DESIGN/SETTINGThis retrospective cross-sectional study took place in a quaternary care academic pediatric hospital in the United States. PATIENT SAMPLEAdolescents aged 10–18 who underwent posterior spinal fusion for a diagnosis of AIS between July 2014 and December 2019. OUTCOMES MEASURESEffective radiation dose and projected lifetime cancer risk associated with intraoperative doses of ionizing radiation. METHODSClinical and radiographic parameters were abstracted, including total radiation dose during surgery from flat plate radiographs, fluoroscopy, and intraoperative CT scans. Multivariable regression analysis was used to assess differences in radiation exposure between patients treated with conventional radiography versus intraoperative navigation. Radiation exposure was translated into lifetime cancer risk using well-established algorithms. RESULTSIn total, 245 patients were included, 119 of whom were treated with navigation. The cohort was 82.9% female and 14.4 years of age. The median radiation exposure (in millisieverts, mSv) for fluoroscopy, radiography, and navigation was 0.05, 4.14, and 8.19 mSv, respectively. When accounting for clinical and radiographic differences, patients treated with intraoperative navigation received 8.18 mSv more radiation (95%CI: 7.22–9.15, p<.001). This increase in radiation projects to 0.90 iatrogenic malignancies per 1,000 patients (95%CI 0.79–1.01). CONCLUSIONSOurs is the first work to define cancer risk in the setting of radiation exposure for navigated AIS surgery. We project that intraoperative navigation will generate approximately one iatrogenic malignancy for every 1,000 patients treated. Given that spine surgery for AIS is common and occurs in the context of a multitude of other radiation sources, these data highlight the need for radiation budgeting protocols and continued development of lower radiation dose technologies. LEVEL OF EVIDENCETherapeutic, III.

Hybrid architecture based intelligent diagnosis assistant for GP

As the first point of contact for patients, General Practitioners (GPs) play a crucial role in the National Health Service (NHS). An accurate primary diagnosis from the GP can alleviate the burden on specialists and reduce the time needed to re-confirm the patient’s condition, allowing for more efficient further examinations. However, GPs have broad but less specialized knowledge, which limits the accuracy of their diagnosis. Therefore, it is imperative to introduce an intelligent system to assist GPs in making decisions. This paper introduces two data augmentation methods, the Complaint Symptoms Integration Method and Symptom Dot Separating Method, to integrate essential information into the Integration dataset. Additionally, it proposes a hybrid architecture that fuses the features of words from different representation spaces. Experiments demonstrate that, compared to commonly used pre-trained attention-based models, our hybrid architecture delivers the best classification performance for four common neurological diseases on the enhanced Integration dataset. For example, the classification accuracy of the BERT+CNN hybrid architecture is 0.897, which is a 5.1% improvement over both BERT and CNN with 0.846. Finally, this paper develops an AI diagnosis assistant web application that leverages the superior performance of this architecture to help GPs complete primary diagnosis efficiently and accurately.

Cost‐effectiveness simulation of AI screening for DR in Japan

Aims: We investigated the efficacy of AI‐based screenings in preventing blindness, specifically assessing cost‐effectiveness when two further detection opportunities are incorporated: “DR screening during national health checkups” and “DR screening during diabetes clinic visits at a physician's office”, or a combination of the two. We also examined potential additional costs associated with AI screening under various conditions.Methods: We employed a Markov Model in this simulation study. We constructed the modelling of diabetes detection, treatment, and retinopathy treatment based on transition probabilities derived from epidemiological and clinical studies. These covered incidence and progression rates of diabetes and diabetic retinopathy, treatment options and their efficacy, and visual prognosis. The age range for the model was set at 40–90 years with a 1‐year cycle, assuming a population of 500 000 individuals with a discount rate of 2%. Simulations were conducted using TreeAge. Costs were calculated using the direct costs of general health examinations and insurance treatments. We assumed diabetic retinopathy would result in a consistent visual acuity following detection, progression, and treatment, providing utility values based on the severity of DR.Results: DR screening via AI in health check‐up programs proved not to be cost‐effective, as it led to a 0.9% increase in cumulative blindness and the ICERs were dominated. The addition of an annual AI screening at physicians' clinics, with referrals to ophthalmology based on the results, resulted in an ICER of 8192751JPY/QALY, which was not cost‐effective. However, assuming that 75% of patients received AI screening during medical examinations and that 75% of these underwent ophthalmological examinations per the results, AI screening proved to be dominant and cost‐effective, reducing the cumulative number of blindness by 2.8%.Conclusions: Even though automated AI diagnosis of DR is becoming a reality, our findings suggest that merely introducing AI screening in health checkups and physicians' clinics in Japan will not be cost‐effective. However, cost‐effectiveness can be improved by increasing the rate of screening and the follow‐up rate of ophthalmological examinations. If this is achieved, there might also be a possibility of incurring additional costs due to AI screening. Our study underlines the importance of a seamless link from diabetes detection to ophthalmologic care for successful DR screening with AI systems.

Application of artificial intelligence-based dual source CT scanning in the differentiation of lung adenocarcinoma in situ and minimally invasive adenocarcinoma.

Lung adenocarcinoma is the most common type of lung cancer with highly incidence and mortality. Due to the overlap of morphological features, it is difficult to distinguish clinically between preinvasive lesions (in situ adenocarcinoma, AIS) and invasive lesions (minimally invasive adenocarcinoma, MIA), which appear as ground glass cloudy nodules. This study was performed to probe the application value of artificial intelligence (AI)-based dual source CT scanning in the differentiation of AIS as well as MIA. The clinical data of 136 patients in Shanghai Baoshan Hospital of Integrated Traditional Chinese and Western Medicine from January 2019 to January 2022 were retrospectively analyzed. The accuracy of AI in distinguishing lung AIS (n=76) and MIA (n=60) were analyzed. The effectiveness of AI in detecting nodules and its diagnostic efficacy for AIS and MIA were explored. The proportion of patients with clear and regular lesion boundaries in AIS was higher than that in MIA. The mean lesion diameter of AIS patients was shorter than MIA patients. There was no difference in the CT value between AIS and MIA in the ground glass nodule density area of pure ground glass nodule and mixed ground glass nodule, but the CT value of the solid nodule density area in AIS was lower. The occurrence of pulmonary vascular abnormality, air bronchogram sign, and pleural depression in AIS patients were lower than MIA patients. The detection rate of AI for lung adenocarcinoma with nodule diameter ≤ 5 mm, complete solid nodules and ground glass nodules was significantly higher than radiologists. The sensitivity, specificity, positive prediction rate, negative prediction rate and accuracy of AI detection were significantly higher than radiologists. AI-based dual source CT scanning can clearly show the morphological characteristics of lung adenocarcinoma, which is helpful for the differential diagnosis of lung AIS as well as MIA.

STGA-MS: AI diagnosis model of regional wall motion abnormality based on 2D transthoracic echocardiography

Regional wall motion abnormality (RWMA) is a common manifestation of ischemic heart disease detected through echocardiography. Currently, RWMA diagnosis heavily relies on visual assessment by doctors, leading to limitations in experience-based dependence and suboptimal reproducibility among observers. Several RWMA diagnosis models were proposed, while RWMA diagnosis with more refined segments can provide more comprehensive wall motion information to better assist doctors in the diagnosis of ischemic heart disease. In this paper, we proposed the STGA-MS model which consists of three modules, the spatial-temporal grouping attention (STGA) module, the segment feature extraction module, and the multiscale downsampling module, for the diagnosis of RWMA for multiple myocardial segments. The STGA module captures global spatial and temporal information, enhancing the representation of myocardial motion characteristics. The segment feature extraction module focuses on specific segment regions, extracting relevant features. The multiscale downsampling module analyzes myocardial motion deformation across different receptive fields. Experimental results on a 2D transthoracic echocardiography dataset show that the proposed STGA-MS model achieves better performance compared to state-of-the-art models. It holds promise in improving the accuracy and reproducibility of RWMA diagnosis, assisting clinicians in diagnosing ischemic heart disease more reliably.

Identification of the cuproptosis-related ceRNA network and risk model in acute ischemic stroke by integrated bioinformatics analysis

BackgroundAcute ischemic stroke (AIS) is one of the leading contributors to death and disability in adults. And cuproptosis is a novel type of cell death. Yet, its role in AIS is still unknown.MethodsThe mRNA, miRNA, and circRNA expression data were downloaded from the Gene Expression Omnibus database. We explored differentially expressed circRNAs (DEcircRNAs), microRNAs (DEmiRNAs), and cuproptosis-related genes (DECuRGs) after AIS. With the target prediction tools, we constructed a cuproptosis-related competitive endogenous RNA (ceRNA) network mediated by circRNAs in AIS. Afterward, functional enrichment analysis, cytoHubba plugin, protein–protein interaction, weighted gene co-expression network analysis, and unsupervised clustering analysis were performed to determine the critical genes and relevant pathways. Machine learning techniques were used to identify the optimal risk model. The CIBERSORT was applied to explore the immune-infiltrating characteristics in AIS samples. Finally, two independent datasets were employed to verify the predictive value of the risk model.ResultsAltogether, 26 DECuRGs were identified in this study. Enrichment analysis revealed that they participated in the reactive oxygen metabolism, inflammatory responses, and corresponding cuproptosis-related biological processes. Of the DECuRGs, MTF1 and UBE2D2 were included in the ceRNA network, comprising three circRNA-miRNA and two miRNA-mRNA interaction pairs. Hub gene analysis determined the hub regulatory axis in the process of cuproptosis, namely, MTF1-miR-765-circ_0040760/0068531. We finally constructed a 5-gene risk model (C10orf32, NUCB1, AX748267, MRPL28, and PPP1R15A) by multiple analyses, which was validated by two independent datasets (AUC = 0.958 and 0.668). Besides, significant differences in immune cell infiltration were observed between AIS patients and normal controls. The levels of neutrophils were correlated with most of the DECuRGs. The ceRNA axis identified in this study was also associated with the immune microenvironment of AIS patients.ConclusionThe findings revealed that cuproptosis might be associated with AIS and that the key nodes, including the regulatory axes, might exert critical roles in the process of AIS. The risk model provided new insights into the early diagnosis and treatment of AIS.

Progress in the Application of Artificial Intelligence in Ultrasound Diagnosis of Breast Cancer

The mammary gland is an important human organ that secretes milk and feeds offspring, while breast tumors are benign or malignant tumors that occur in the breast tissue. There are many causes of breast cancer, and the incidence continues to rise, which is an important killer that threatens women's health. In recent years, a large number of researchers have been interested in the study of AI diagnosis of breast cancer. Artificial intelligence uses a specific algorithm to intelligently process ultrasound images, and develops a high-precision and high-efficiency breast cancer recognition model through training and optimization of the algorithm. At present, the application of computer-aided detection methods in breast cancer ultrasound has been gradually promoted, and the combined application of artificial intelligence has played an advantageous role in the field of breast disease ultrasound diagnosis, such as shortening the examination time, effectively improving the detection rate and diagnostic accuracy rate. The main reasons are: the accuracy of traditional AI diagnosis model of breast cancer based on machine learning is not high; The aim of this paper is to improve the accuracy of early diagnosis of breast diseases effectively and reduce the misdiagnosis rate caused by overwork of doctors on the basis of clear medical images and computer-aided diagnosis technology.

Automatic axial vertebral rotation estimation on radiographs for adolescent idiopathic scoliosis by deep learning

PurposeAdolescent idiopathic scoliosis (AIS) is a three-dimensional spine deformity governed by lateral curvature and axial vertebral rotation (AVR). Estimating AVR is important for treatment decisions and the prediction of AIS progression. However, manual AVR measurements have intra-observer and inter-observer errors, and therefore this paper proposes an automatic AVR measurement method to address the low precision issue of manual measurement. MethodWe develop an improved feature extraction module for vertebral landmark detection and pedicle segmentation. The improved coordinate convolution layer combined Polarized Self-Attention mechanism is applied in the feature extraction module to help the High-resolution Network to extract coordinate information. Based on vertebral landmark detection and pedicle segmentation, we propose an automatic AVR measurement algorithm to estimate the AVR. ResultsThe mean radial error (MRE) of vertebral landmark detection is 2.70 mm pixels, and the dice coefficient of pedicle segmentation is 72.45%. Compared with the original model, the MRE decreases by 2.13 mm, and the dice coefficient improves by 4.61%. For the AVR measurement results, the testing set contains 37 spine radiographs, including 481 vertebrae and 962 AVR measurements (481 vertebrae by two observers). The average measurement error is lower than 5°, which is within the standard of clinical measurement error. ConclusionThe results demonstrate that the proposed network performs well in vertebral landmark detection and pedicle segmentation, and the proposed AVR measurement is adopted for clinic diagnosis of AIS. SignificanceOur method achieves automatic AVR measurement, reducing the error introduced by manual measurement and improving the efficiency of orthopedists.

Moderate scoliosis continues to progress at 30-year follow-up: a call for concern?

This natural history study reports long-term radiographic and clinical outcomes of patients with diagnosis of AIS with curves between 30° and 50°. Our purpose was to determine if any intervention in the natural history is warranted. This was a longitudinal descriptive study at a single institution. We reviewed patient factors, radiographic parameters, and patient-reported outcomes at 20- and 30-year follow-up. A total of 31 patients were included. At skeletal maturity (which was the initial point of measurement), the median age was 17years (range 12-21), the thoracic Cobb angle was 35° ± 5° (maximum-minimum 27°-47°), and the lumbar Cobb angle was 33° ± 7° (maximum-minimum 18°-45°). The median final follow-up was 35years (median age 52, range 32-61) when the thoracic Cobb angle was 47° ± 12° (maximum-minimum 31°-74°) and the lumbar Cobb angle was 40° ± 17° (maximum-minimum 19°-69°). At final follow-up, 9 (29%) patients had a structural curve > 50°. Ten (32%) patients had a curve from 40° to 49° and 11 (35%) patients had a curve < 40°. The thoracic Cobb angle had progressed from < 40° to > 50° in 5 patients. Thoracolumbar and lumbar Cobb angles progressed from < 40° to greater than > 50° in 1 and 3 patients, respectively. Few patients had functional limitations according to Roland-Morris, Oswestry, and SF36 scores. Pain scores were minimal at final follow-up. All AIS curves between 30° and 50° at skeletal maturity tend to progress. Thoracic curves progress more than lumbar curves during the first 20years and then progression slows down. The opposite happens with lumbar curves. Therefore, the rate of progression decreases with thoracic curves and increases with lumbar curves. Nevertheless, few patients have functional limitations. Further follow-up is necessary to define the true long-term outcome of moderate curves at maturity.

Clinical characteristics and treatment efficacy of immune checkpoint inhibitors (ICIs) in patients with ICIs-induced Adrenal insufficiency

BackgroundAdrenal insufficiency (AI) caused by immune checkpoint inhibitors (ICIs) is an extremely rare immune-related adverse event (irAE). The detailed clinical characteristics and outcomes of patients with ICI-induced AI are unavailable. This study aimed to explore the clinical characteristics and efficacy of treatment in patients with ICI-induced AI. MethodsWe retrospectively collected information on patients diagnosed with AI caused by ICIs at LiShui Municipal Central Hospital and Zhejiang Cancer Hospital, including baseline characteristics, laboratory results, symptoms, treatment outcomes of AI, and hormone use. Survival outcomes were calculated using the Kaplan–Meier method and stratified according to the different situations. ResultsFrom December 2020 to February 2023, among 1014 patients treated with ICI therapy, a total of twenty patients were diagnosed with ICI-induced AI. Most of the patients were men (80%, n = 16), with a performance status (PS) of 0 − 1 (95%, n = 19). The median (range) age was 65.9 (49−80) years and 14 patients (70%) were treated with ICIs as first-line therapy. The majority of the patients (70%, n = 14) experienced grade 3 − 4 AI. All patients received corticosteroid replacement therapy, and only 7 patients recovered. The median time to the diagnosis of AI after starting ICI therapy was 5.2 (3.0 − 7.5) months. The objective response rate was 70% and median progression-free survival in these patients was 16.0 months (95% confidence interval: 11.7 − 20.3 months). ConclusionsICI-induced AI is a rare irAE, and close monitoring of cortisol levels is important. Patients diagnosed with AI after receiving immunotherapy seem to have a favorable outcome.

2D/3D ultrasound diagnosis of pediatric distal radius fractures by human readers vs artificial intelligence

Wrist trauma is common in children and generally requires radiography for exclusion of fractures, subjecting children to radiation and long wait times in the emergency department. Ultrasound (US) has potential to be a safer, faster diagnostic tool. This study aimed to determine how reliably US could detect distal radius fractures in children, to contrast the accuracy of 2DUS to 3DUS, and to assess the utility of artificial intelligence for image interpretation. 127 children were scanned with 2DUS and 3DUS on the affected wrist. US scans were then read by 7 blinded human readers and an AI model. With radiographs used as the gold standard, expert human readers obtained a mean sensitivity of 0.97 and 0.98 for 2DUS and 3DUS respectively. The AI model sensitivity was 0.91 and 1.00 for 2DUS and 3DUS respectively. Study data suggests that 2DUS is comparable to 3DUS and AI diagnosis is comparable to human experts.