Relevant Images Research Articles

Content-based image retrieval for medical diagnosis using fuzzy clustering and deep learning

Brain tumors are one of the most threatening causes of death worldwide. Neuroradiologists widely use Magnetic Resonance Imaging as a golden standard non-invasive imaging method for diagnosing Glioma grades. Brain tumor diagnosis based on radiology images is tedious and highly subjective to intra and inter-observer variability; it adversely affects therapeutic planning. Better medical treatment can be provided relating to similar past cases if relevant images are retrieved from an extensive medical image database. Content-Based Image Retrieval systems are potent image tools for dealing with such massive datasets. In a CBIR system, accurate classification and retrieval of similar pathological images can be effectively automated by utilising Convolutional Neural Network-based feature extraction methods. This paper presents a Content-Based Medical Image Retrieval pipeline on a medical domain using the CNN model for feature extraction and the clustering method used to index the feature map database. The proposed system applies a Multi-level Gain-based feature selection to reduce the dimensionality of the feature vectors obtained from the pre-trained CNN models. The experiment follows five-fold cross-validation on BraTS 2018 and 2020 datasets. The proposed system achieves state-of-the-art MRI brain image retrieval performance with a mean Average Precision and Precision@10 of 98.15% and 97.62%, respectively. The results prove the proposed method’s effectiveness and validate its clinical application.

Multimodal Ophthalmic Imaging in Spinocerebellar Ataxia Type 7.

The aim of this case series and narrative literature review is to highlight the importance of multimodal imaging in the ophthalmological examination of patients with spinocerebellar ataxia type 7 and provide a summary of the most relevant imaging techniques. Three patients with SCA7 were included in this case series. A literature review revealed twenty-one publications regarding ocular manifestations of SCA7, and the most relevant aspects are summarized. The role of different imaging techniques in the follow-up of SCA7 patients is analyzed, including color vision testing, corneal endothelial topography, color fundus photography (CFP) and autofluorescence, near infrared reflectance imaging, spectral domain optical coherence tomography (SDOCT), visual field examination, and electrophysiological tests. SDOCT provides a rapid and non-invasive imaging evaluation of disease progression over time. Additional examination including NIR imaging can provide further information on photoreceptor alteration and subtle disruption of the RPE, which are not evident with CFP at an early stage. Electrophysiological tests provide essential results on the state of cone and rod dystrophy, which could be paramount in guiding future genetic therapies. Multimodal imaging is a valuable addition to comprehensive ophthalmological examination in the diagnosis and management of patients with SCA7.

The patient with an acute painful wrist.

Patients with wrist pain commonly present to primary care and emergency departments. A detailed history and examination, alongside relevant imaging, will help find the correct diagnosis and ensure that patients receive the correct treatment in a timely manner. This article summarises the key points in history and examination and the role of imaging, including suggestions of which modality should be requested. Finally, important acute pathologies are highlighted, including fractures, soft tissue injuries and infection, with examples of their relevant imaging.

Abstract 083: Initial Angiogenesis in Vein of Galen Malformations Is Associated with Increased Symptomatic Burden

Introduction Vein of Galen Malformations (VOGM) are congenital brain arteriovenous malformations which often present during the neonatal or infantile periods. While endovascular embolization is the recommended management for this pathology, a subset of VOGM patients present with pathologic angiogenesis which complicates endovascular therapy. Previous studies have shown that these patients require a greater number of embolizations to achieve cure, but it is unclear whether this radiographic finding is associated with differences in clinical presentation. Methods A single‐center retrospective review identified all patients with a Vein of Galen Malformation from January 2004 to May 2023. Patients were stratified into three groups based on the development and timing of angiogenesis during their treatment course: pre‐intervention angiogenesis, post‐intervention angiogenesis, and no angiogenesis. Clinical presentations, relevant imaging, radiographic data, and procedural details were reviewed. Results 178 patients with VOGM were identified. 38 patients were in the pre‐intervention angiogenesis group, 51 patients were in the post‐intervention angiogenesis group, and 89 patients were in the no angiogenesis group. Pre‐intervention angiogenesis patients were significantly older on presentation for initial angiogram than post‐intervention angiogenesis or no angiogenesis patients (p<0.05). At initial clinical presentation, patients with pre‐intervention angiogenesis demonstrated significantly greater developmental delay (p<0.05). This difference persisted after controlling for age at initial presentation. Patients presenting with angiogenesis either pre‐intervention or post‐intervention required more embolizations to achieve cure compared to patients who did not develop angiogenesis during their treatment course (p<0.05). Conclusion Patients with VOGM who present with pre‐intervention angiogenesis demonstrate significantly more serious neurologic symptoms on initial presentation, even after controlling for age at presentation. Given that patients with angiogenesis are often older on initial presentation and require more embolizations to be cured, early and timely treatment is likely necessary to mitigate the further progression of both the angiogenic network and any associated neurological symptoms.

Extraocular Orbital and Peri-Orbital Masses.

In this article, we will describe relevant anatomy and imaging findings of extraocular and orbital rim pathologic conditions. We will highlight important clinical and imaging pearls that help in differentiating these lesions from one another, and provide a few practical tips for challenging cases.

Pilot study of HER2 targeted 64 Cu-DOTA-tagged PET imaging in gastric cancer patients.

Human epidermal growth factor receptor 2 (HER2) is an important biomarker for targeted gastric cancer (GC) immunotherapy. However, heterogeneous HER2 overexpression in GC, loss of HER2 expression during therapy, and inability to non-invasively identify HER2 overexpressing tumors impede effective targeting therapies. Improved HER2-specific functional imaging can address these challenges. Trastuzumab is a HER2-directed mAb to treat HER2 overexpressing cancers. The 64 Cu-DOTA-trastuzumab radiotracer is used to detect HER2+ metastatic breast cancer. We aimed to develop 64 Cu-DOTA-trastuzumab PET-CT to detect and characterize tumor uptake in HER2+ or - GC patients. We conducted a single-arm phase II pilot study exploring the feasibility of 64 Cu-DOTA-trastuzumab for PET imaging of HER2 overexpressing GC compared to HER2 non-expressing tumors. Eight patients with biopsy-confirmed gastric adenocarcinoma were included. Immunohistochemistry was used to evaluate primary tumor biopsies for HER2 overexpression. Patients were injected with 45 mg of cold trastuzumab followed by 5 mg of 64 Cu-DOTA-trastuzumab. PET-CT scans were performed 24-48 h post radiotracer injection and compared to standard staging CT scans. We observed limited toxicity following 64 Cu-DOTA-trastuzumab injections. While there was uptake of the radiotracer in portions of HER2+ lesions, there was no statistically significant distinction between tumor and background by standardized uptake value analysis. Despite the potential of 64 Cu-DOTA-trastuzumab PET imaging of HER2+ metastatic breast cancer, a 5 mg dose of this radiotracer injected 24-48 h before imaging was insufficient to identify HER2+ GC. These results inform future GC imaging studies to optimize biomarker-targeted therapies based on dosage and timing for more clinically relevant imaging.

Mono- and multimeric PSMA-targeting small molecule-thorium-227 conjugates for optimized efficacy and biodistribution in preclinical models

PurposePSMA (prostate-specific membrane antigen) is highly expressed on prostate cancer (PrCa) cells and extensively used as a homing target for PrCa treatment. Most prominently, PSMA-targeting conjugate PSMA-617, carrying a DOTA chelator and labeled with therapeutic radionuclides like beta-emitting lutetium-177 or alpha-emitting actinium-225, has shown clinical activity in PrCa patients. We sought to develop PSMA-targeting small molecule (SMOL) conjugates that show high uptake in PSMA-expressing tumors and fast clearance, and can easily be labeled with the alpha emitter thorium-227 (half-life 18.7 days).MethodsA novel linker motif with improved competition against 3H-PSMA-617 on PSMA-expressing LNCaP cells was identified. A 2,3-hydroxypyridinone chelator modified with carboxyl groups (carboxy-HOPO) with increased hydrophilicity and robust labeling with thorium-227 was developed and allowed the synthesis of mono-, di-, tri-, and tetrameric conjugates. The resulting monomeric and multimeric PSMA SMOL-TTCs (targeted thorium conjugate) were evaluated for cellular binding, internalization, and antiproliferative activity. The in vivo antitumor efficacy of the PSMA SMOL-TTCs was determined in ST1273 and KUCaP-1 PrCa models in mice, and their biodistribution was assessed in cynomolgus monkeys, minipigs, and mice.ResultsThe monomeric and multimeric PSMA SMOL conjugates were readily labeled with thorium-227 at room temperature and possessed high stability and good binding, internalization, and antiproliferative activity in vitro. In vivo, the monomeric, dimeric, and trimeric PSMA SMOL-TTCs showed fast clearance, potent antitumor efficacy, and high uptake and retention in prostate tumors in mice. No major uptake or retention in other organs was observed beyond kidneys. Low uptake of free thorium-227 into bone confirmed high complex stability in vivo. Salivary gland uptake remained inconclusive as mini pigs were devalidated as a relevant model and imaging controls failed in cynomolgus monkeys.ConclusionMonomeric and multimeric PSMA SMOL-TTCs show high tumor uptake and fast clearance in preclinical models and warrant further therapeutic exploration.

Abnormal Rhomboid Lip and Choroid Plexus Should be Valued in Microvascular Decompression for Vestibulocochlear Diseases

Surgical outcomes for functional vestibulocochlear diseases vary, and the influencing factors are not fully understood. The role of a rhomboid lip (RL) and choroid plexus (CP) in microvascular decompression (MVD) of the vestibulocochlear nerve has not been studied. This study aims to evaluate the surgical efficacy of MVD for vestibulocochlear diseases, with and without addressing the RL and CP, to enhance our understanding of their etiology. A total of 15 patients who underwent MVD for the vestibulocochlear nerve between 2013 and 2022 were retrospectively identified and followed up. The patients were classified into 4 categories: vestibular paroxysmia (VP), benign positional paroxysmal vertigo (BPPV), and Meniere disease (MD). The fourth was a "tinnitus" group. The relief of symptoms, recurrence, satisfaction after surgery, available relevant imaging studies, and intraoperative observation data were evaluated. Following MVD, 6 of the 7 patients in the VP group, the 1 patient in the BPPV group, and 1 of 2 patients in the MD group were completely relieved of vertigo. The seventh VP patient showed significant improvement. The 5 patients in the "tinnitus" group remained unchanged. Retrospectively, 4 patients from the VP, BPPV, and MD groups who underwent RL incision and CP excision were also free of vertigo, although vascular compression was not confirmed in these cases. MVD is generally considered a useful treatment for VP and could also be effective in managing recurrent vertigo caused by BPPV and MD. Our results highlight the potential role of an abnormal RL and CP in the development of vertigo symptoms. Patients presenting with "tinnitus" require further investigation and might not be suitable for MVD.

Radiological findings in patients with isolated acute-onset ocular motility disorders

ObjectivesTo image or not to image is a critical question in patients with isolated acute-onset ocular motility disorders (AOMD), without any neurological symptoms/signs. Imaging may help in diagnosis; but is associated with risks and costs. The study aimed to evaluate radiological findings in patients with apparently isolated AOMD on clinical exam, and identify patient/disease characteristics more likely to be associated with positive imaging. MethodsA retrospective review of patient charts with isolated AOMD (<3 months onset), who were examined by an ophthalmologist and had imaging at a tertiary-care center over a period of 18 months, was conducted. Ophthalmology exam and diagnostic imaging findings were recorded. Radiological findings were classified as “clinically relevant”, “non-relevant”, and “normal/no positive” findings. Results46 patients were included (3–91 years; 27 males, 19 females). 19 were clinically classified as nerve palsy (1 third, 10 fourth, and 8 sixth nerve), 14 acute-onset esotropia, 3 acute-onset exotropia, 7 acute-onset vertical strabismus inconsistent with nerve palsy, and 3 limitation of elevation/suspected dorsal midbrain syndrome. We found 8/46 (17%) with clinically relevant imaging findings; of these, 5/46 (11%) had positive neuro-imaging findings (mass, malformation, aneurysm, infarct) and 3/46 (7%) had positive orbital imaging findings (thyroid eye disease). Positive neuro-imaging was more common in patients with certain symptoms (headache), in certain clinical diagnostic entities (dorsal midbrain syndrome spectrum, acute-onset esotropia), and in younger patients (<40 years). ConclusionsPositive neuro-imaging findings may be seen even in patients with apparently isolated ocular symptoms/signs on clinical exam. This data may help institutions with decision-making and policy formulation for imaging patients with isolated AOMD.

Manta Ray Foraging Optimizer with Deep Learning-based Fundus Image Retrieval and Classification for Diabetic Retinopathy Grading

Diabetic Retinopathy (DR) is a major source of sightlessness and permanent visual damage. Manual Analysis of DR is a labor-intensive and costly task that requires skilled ophthalmologists to observe and evaluate DR utilizing digital fundus images. The images can be employed for analysis and disease screening. This laborious task can gain a great advantage in automated detection by exploiting Artificial Intelligence (AI) techniques. Content-Based Image Retrieval (CBIR) approaches are utilized to retrieve related images in massive databases and are helpful in many application regions and most healthcare systems. With this motivation, this article develops the new Manta Ray Foraging Optimizer with Deep Learning-based Fundus Image Retrieval and Classification (MRFODL-FIRC) approach for the grading of DR. The suggested MRFODL-FIRC model investigates the retinal fundus imaging effectively to retrieve the relevant images and identify class labels. To achieve this, the MRFODL-FIRC technique uses Median Filtering (MF) as a pre-processing step. The Capsule Network (CapsNet) model is used to produce feature vectors with the MRFO algorithm as a hyperparameter optimizer. For the image retrieval process, the Manhattan distance metric is used. Finally, the Variational Autoencoder (VAE) model is used for recognizing and classifying DR. The investigational assessment of the MRFODL-FIRC technique is accomplished on medical DR and the outputs highlighted the improved performance of the MRFODL-FIRC algorithm over the current approaches.

Scraping Relevant Images from Web Pages without Download

Automatically scraping relevant images from web pages is an error-prone and time-consuming task, leading experts to prefer manually preparing extraction patterns for a website. Existing web scraping tools are built on these patterns. However, this manual approach is laborious and requires specialized knowledge. Automatic extraction approaches, while a potential solution, require large training datasets and numerous features, including width, height, pixels, and file size, that can be difficult and time-consuming to obtain. To address these challenges, we propose a semi-automatic approach that does not require an expert, utilizes small training datasets, and has a low error rate while saving time and storage. Our approach involves clustering web pages from a website and suggesting several pages for a non-expert to annotate relevant images. The approach then uses these annotations to construct a learning model based on textual data from the HTML elements. In the experiments, we used a dataset of 635,015 images from 200 news websites, each containing 100 pages, with 22,632 relevant images. When comparing several machine learning methods for both automatic approaches and our proposed approach, the AdaBoost method yields the best performance results. When using automatic extraction approaches, the best f-Measure that can be achieved is 0.805 with a learning model constructed from a large training dataset consisting of 120 websites (12,000 web pages). In contrast, our approach achieved an average f-Measure of 0.958 for 200 websites with only six web pages annotated per website. This means that a non-expert only needs to examine 1,200 web pages to determine the relevant images for 200 websites. Our approach also saves time and storage space by not requiring the download of images and can be easily integrated into currently available web scraping tools, because it is based on textual data.

THU007 Improving Completeness Of Endocrinology Referrals Using Structured Templates

Abstract Disclosure: K. Arao: None. V. Vimalananda: None. Objective: Endocrinologists rely on information in the referral to guide the focus and scope of their evaluations. However, endocrinology referrals frequently lack important clinical information, which may increase the risk of inefficiency, errors, and adverse outcomes. This quality improvement project aimed to improve the completeness of endocrinology referrals by utilizing structured referral templates for common endocrine conditions including diabetes, hypogonadism, osteoporosis, thyroid dysfunction (hyperthyroidism and hypothyroidism), and thyroid nodule. Methods: The project was implemented at a large Veterans Affairs Medical Center (VAMC) in New England. Structured referral templates were designed based on existing templates at other VA health systems, the American College of Physicians High Value Care Coordination Toolkit (developed in partnership with The Endocrine Society), and input from endocrinologists and primary care providers at the VAMC. Templates included fields for essential elements: a consult question, relevant clinical history and past medical history, type of visit preference, pertinent over-the-counter medications, and dates of relevant laboratory and imaging studies. We used functions including auto-population of data, checkboxes, and drop-down menus to ease usability. The templates were embedded into the electronic referral system. We conducted a retrospective chart review to compare referrals for a cohort of patients referred between 1/1/2021-9/1/2021 (pre-intervention) vs. a cohort referred between 10/1/2021-9/30/2022 (post-intervention). Each referral was rated using a scoring system derived from the criteria in the data abstraction tool formulated by the investigators. Results: On average, pre-intervention referrals included 52% of the essential elements and post-intervention referrals included 92%. Post-intervention referrals for diabetes, hypogonadism, osteoporosis, thyroid dysfunction (hyperthyroidism and hypothyroidism), and thyroid nodule improved from baseline by 44%, 45%, 39%, 39%, and 36% respectively. The greatest improvement was for the element “type of visit requested (one-time consultation, shared care, transfer of care, and consultant’s preference)”. A separate analysis excluding “type of visit” showed an average improvement from 64% of essential elements included pre-intervention to 92% included post-intervention. Conclusion: Structured referral templates embedded into an electronic referral system improves availability of essential information for the endocrine consultant. Our next steps are to modify the templates with the ongoing input of both endocrinologists and primary care providers. Future work should examine the effect of referral templates on specialist and primary care provider user experience, patient experience of care, efficiency, and clinical outcomes. Presentation: Thursday, June 15, 2023

Comparison of Venous Thromboembolism Outcomes after COVID-19 and Influenza Vaccinations.

Background Published data on the risk of venous thromboembolism (VTE) with coronavirus disease 2019 (COVID-19) vaccines are scarce and inconclusive, leading to an unmet need for further studies. Methods A retrospective, multicentered study of adult patients vaccinated for one of the three approved COVID-19 vaccines in the United States of America and a pre-COVID-19 cohort of patients vaccinated for influenza at two institutions: Mayo Clinic Enterprise sites and the Medical College of Wisconsin, looking at rate of VTE over 90 days. VTE was identified by applying validated natural language processing algorithms to relevant imaging studies. Kaplan-Meier curves were used to evaluate rate of VTE and Cox proportional hazard models for incident VTE after vaccinations. Sensitivity analyses were performed for age, sex, outpatient versus inpatient status, and type of COVID-19 vaccine. Results A total of 911,381 study subjects received COVID-19 vaccine (mean age: 56.8 [standard deviation, SD: 18.3] years, 55.3% females) and 442,612 received influenza vaccine (mean age: 56.5 [SD: 18.3] years, 58.7% females). VTE occurred within 90 days in 1,498 (0.11%) of the total 1,353,993 vaccinations: 882 (0.10%) in the COVID-19 and 616 (0.14%) in the influenza vaccination cohort. After adjusting for confounding variables, there was no difference in VTE event rate after COVID-19 vaccination compared with influenza vaccination (adjusted hazard ratio: 0.95 [95% confidence interval: 0.85-1.05]). No significant difference in VTE rates was observed between the two cohorts on sensitivity analyses. Conclusion In this large cohort of COVID-19-vaccinated patients, risk of VTE at 90 days was low and no different than a pre-COVID-19 cohort of influenza-vaccinated patients.

GuaRD: Guaranteed robustness of image retrieval system under data distortion turbulence.

Image search systems could be endangered by adversarial attacks and data perturbations. The image retrieval system can be compromised either by distorting the query or hacking the ranking system. However, existing literature primarily discusses attack methods, whereas the research on countermeasures to defend against such adversarial attacks is rare. As a defense mechanism against the intrusions, quality assessment can complement existing image retrieval systems. "GuaRD" is proposed as an end-to-end framework that uses the quality metric as a weighted-regularization term. Proper utilization and balance of the two features could lead to reliable and robust ranking; the original image is assigned a higher rank while the distorted image is assigned a relatively lower rank. Meanwhile, the primary goal of the image retrieval system is to prioritize searching the relevant images. Therefore, the use of leveraged features should not compromise the accuracy of the system. To evaluate the generality of the framework, we conducted three experiments on two image quality assessment(IQA) benchmarks (Waterloo and PieAPP). For the first two tests, GuaRD achieved enhanced performance than the existing model: the mean reciprocal rank(mRR) value of the original image predictions increased by 61%, and the predictions for the distorted input query decreased by 18%. The third experiment was conducted to analyze the mean average precision (mAP) score of the system to verify the accuracy of the retrieval system. The results indicated little deviation in performance between the tested methods, and the score was not effected or slightly decreased by 0.9% after the GuaRD was applied. Therefore, GuaRD is a novel and robust framework that can act as a defense mechanism for data distortions.

High-resolution proton metabolic mapping of the human brain at 7 T using free induction decay rosette spectroscopic imaging.

Magnetic resonance spectroscopic imaging (MRSI) provides information about the spatial distribution of metabolites in the brain. These metabolite maps can be valuable in diagnosing central nervous system pathology. However, MRSI generally suffers from a long acquisition time, poor spatial resolution, and a low metabolite signal-to-noise ratio (SNR). Ultrahigh field strengths (≥ 7 T) can benefit MRSI with an improved SNR and allow high-resolution metabolic mapping. Non-Cartesian spatial-spectral encoding techniques, such as rosette spectroscopic imaging, can efficiently sample spatial and temporal domains, which significantly reduces the imaging time and enables high-resolution metabolic mapping in a clinically relevant scan time. In the current study, high-resolution (in-plane resolution of 2 × 2 mm2 ) mapping of proton (1 H) metabolites in the human brain at 7 T, is demonstrated. Five healthy subjects participated in the study. Using a time-efficient rosette trajectory and short TR/TE free induction decay MRSI, high-resolution maps of 1 H metabolites were obtained in a clinically relevant imaging time (6 min). Suppression of the water signal was achieved with an optimized water suppression enhanced through T1 effects approach and lipid removal was performed using L2 -regularization in the postprocessing. Spatial distributions of N-acetyl-aspartate, total choline, creatine, N-acetyl-aspartyl glutamate, myo-inositol, and glutamate were generated with Cramer-Rao lower bounds of less than 20%.

MD-YOLO: Multi-scale Dense YOLO for small target pest detection

The detection of pests plays a crucial role in intelligent early warning systems of injurious insects and diseases in precision agriculture. However, pests strong concealment and mobility pose significant challenges to their timely detection. In this paper, we propose a novel approach called Multi-scale Dense YOLO (MD-YOLO) for detecting three typical small target lepidopteran pests on sticky insect boards. In MD-YOLO, we design three key components: the image feature extraction part, the feature fusion network, and the prediction module. To enhance the utilization of feature maps and mitigate information loss, we incorporate DenseNet blocks and an adaptive attention module (AAM) into the feature extraction part. The AAM helps capture relevant image details and improves the model’s ability to exploit feature representations effectively. For effective feature integration, our feature fusion network incorporates both a feature extraction path and a feature aggregation path. This enables the deep network to leverage spatial location information from the shallower network, thereby enhancing the detection accuracy. Experimental results demonstrate the effectiveness of MD-YOLO, with detection results achieving an mAP@.5 value of 86.2%, an F1 score of 79.1%, and an IoU value of 88.1%. We conduct extensive experiments to compare MD-YOLO with state-of-the-art models, and the results showcase its superiority. Furthermore, we design an Internet of Things (IoT) system that demonstrates MD-YOLO’s performance in real-world field scenes, highlighting its practical applicability.

Automatic Weight of Color, Texture, and Shape Features in Content-Based Image Retrieval Using Artificial Neural Network

Image retrieval is the process of finding images in the database that are similar to the query image by measuring how close the feature values of the query image are to other images. Image retrieval is currently dominated by approaches that combine several different representations or features. The optimal weight of each feature is needed in combining the image features such as color features, texture features, and shape features. In this study, we use a multi-layer perceptron artificial neural network (MLP) method to obtain feature weights automatically and simultaneously look for optimal weights. The color moment is used to find nine color features, Gray Level Co-occurrence Matrix (GLCM) to find four texture features, and Hu Moment to find seven shape features totaling 20 neurons and all of these features become the input layer in our MLP network. Three neurons in output layers become the automatic weight of each feature. These weights are used to combine each feature's role in obtaining the relevant image. Euclidean Distance is used to measure similarity. The average precision values obtained using automatic feature weights are 93.94% for the synthetic dataset, 91.19% for the Coil-100 dataset, and 54.31% for the Wang dataset. These results have an average difference of 5.06% with the target so automatic feature weighting works well. This value is obtained at a hidden layer size of 11 and a learning rate of 0.1. In addition, the use of automatic feature weighting gives more accurate results compared to manual feature weighting.

An artificial intelligence-based prognostic prediction model for hemorrhagic stroke

PurposeThe prognosis following a hemorrhagic stroke is usually extremely poor. Rating scales have been developed to predict the outcomes of patients with intracerebral hemorrhage (ICH). To date, however, the prognostic prediction models have not included the full range of relevant imaging features. We constructed a clinic-imaging fusion model based on convolutional neural networks (CNN) to predict the short-term prognosis of ICH patients. Materials and methodsThis was a multi-center retrospective study, which included 1990 patients with ICH. Two CNN-based deep learning models were constructed to predict the neurofunctional outcomes at discharge; these were validated using a nested 5-fold cross-validation approach. The models’ predictive efficiency was compared with the original ICH scale and the ICH grading scale. Poor neurological outcome was defined as a Glasgow Outcome Scale (GOS) score of 1–3. ResultsThe training and test sets included 1599 and 391 patients, respectively. For the test set, the clinic-imaging fusion model had the highest area under the curve (AUC = 0.903), followed by the imaging-based model (AUC = 0.886), the ICH scale (AUC = 0.777), and finally the ICH grading scale (AUC = 0.747). ConclusionThe CNN prognostic prediction model based on neuroimaging features was more effective than the ICH scales in predicting the neurological outcomes of ICH patients at discharge. The CNN model’s predictive efficiency slightly improved when clinical data were included.

A novel hash based feature descriptors for content based image retrieval in large database

For retrieving the relevant images from the internet, CBIRs (content based image retrievals) techniques are most globally utilized. However, the traditional image retrieval techniques are unable to represent the image features semantically. The CNNs (convolutional neural networks) and DL has made the retrieval task simpler. But, it is not adequate to consider only the finalized aspect vectors from the completely linked layers to fill the semantic gap. In order to alleviate this problem, a novel Hash Based Feature Descriptors (HBFD) method is proposed. In this method, the most significant feature vectors from each block are considered. To reduce the number of descriptors, pyramid pooling is used. To improve the performance in huge databases, the hash code like function is introduced in each block to represent the descriptors. The proposed method has been evaluated in Oxford 5k, Paris 6k, and UKBench datasets with the accuracy level of 80.6%, 83.9% and 92.14% respectively and demonstrated better recall value than the existing methods.

Optimization Technique Based Approach for Image Segmentation.

The study's goal was to diagnose the condition at an earlier stage by employing the optimization-based technique for image segmentation to find deformities in MRI and Aura images Methods: Our methodology was based on two case studies. The diseased data set of MRI images obtained from the UCI data set and Aura images from Bio-Well were taken into consideration. Using the Relevance Feedback Mechanism (RFM), the sick images that are most pertinent are determined. The optimization-based Cuckoo Search (CS) algorithm is used to find the best features. The resulting model utilising the Truncated Gaussian Mixture Model (TGMM) is used to compare the extracted characteristics. The most relevant images are chosen based on the likely hood estimation. The suggested methodology is tested using 150 retrieved Aura images, 50 trained photos, and processing of the input image utilizing morphological techniques like dilation, erosion, opening, and closing to improve the image quality. Together with segmentation quality measurements including Global Consistency Error (GCE), Probability Random Index (PRI), and Volume of Symmetry(VOS), the results are assessed using image quality metrics such as Average Difference (AD), Maximum Difference (MD), and Image Fidelity (IF). The TGMM algorithm is used to conduct the experiment. The outcomes demonstrate the effectiveness of the suggested approaches in locating various injured tissues inside medical images obtained using MRI technology as well as in locating high-intensity energy zones in which a potential deformity is associated in Aura images. The outcomes reveal a respectable recognition accuracy of about 93%.