Accurate Image Research Articles

PDCdb: the biological activity and pharmaceutical information of peptide-drug conjugate (PDC).

Peptide-drug conjugates (PDCs) have emerged as a promising class of targeted therapeutics with substantial pharmaceutical advantages and market potentials, which is a combination of a peptide (selective to the disease-relevant target), a linker (stable in circulation but cleavable at target site)and a cytotoxic/radioactive drug (efficacious/traceable for disease). Among existing PDCs, those based on radiopharmaceuticals (a.k.a. radioactive drugs) are valued due to their accurate imaging and targeted destruction of disease sites. It's demanded to accumulate the biological activity and pharmaceutical information of PDCs. Herein, a database PDCdb was thus constructed to systematically describe these valuable data. Particularly, biological activities for 2036 PDCs were retrieved from literatures, which resulted in 1684, 613and 2753 activity data generated based on clinical trial, animal modeland cell line, respectively. Furthermore, the pharmaceutical information for all 2036 PDCs was collected, which gave the diverse data of (a) ADME property, plasma half-lifeand administration approach of a PDC and (b) chemical modification, primary target, mode of action, conjugating feature of the constituent peptide/linker/drug. In sum, PDCdb systematically provided the biological activities and pharmaceutical information for the most comprehensive list of PDCs among the available databases, which was expected to attract broad interest from related communities and could be freely accessible at: https://idrblab.org/PDCdb/.

Comparison between the intraoperative CT O-arm and C-arm related to radiation dose and accuracy in scoliosis surgery: A review study

Pedicle screw placement is a vital scoliosis therapy operation that aims to address spinal abnormalities while avoiding consequences including neurological damage caused by misplaced screws. Accurate imaging is required for precise screw insertion, with intraoperative CT (O-arm) and fluoroscopy (C-arm) being the principal modalities used. This review assesses various modalities based on their radiation dosages and accuracy. It comprises research published between 2012 and 2020 that evaluates both effective dose and accuracy for pediatric and adult patients. According to the review, the O-arm outperforms the C-arm in terms of accuracy, resulting in shorter procedure times for both pediatric and adult patients. The O-arm's sophisticated 3-D imaging capacity enhances screw insertion accuracy, reducing the likelihood of problems and the need for additional procedures. Furthermore, using pediatric low-dose protocols and techniques like the Detachable Pedicle Marker and Probe (DPMPs) with pulsed fluoroscopy can reduce radiation exposure while still ensuring precise screw placement. The review suggests that the O-arm has considerable advantages over the C-arm, such as improved image quality and lower radiation risk. This leads to enhanced patient safety and procedural outcomes. Overall, the O-arm's capacity to produce high-quality 3-D pictures makes it an invaluable tool in pedicle screw insertion, resulting in more accurate surgical outcomes and reducing the risk of problems associated with spinal deformity repairs.

Evaluating T1-weighted MRI techniques for fetal gastrointestinal diagnostics: A comparative study

PurposeIn clinical practice, fetal gastrointestinal magnetic resonance imaging (MRI) encounters significant challenges. T1-weighted images are particularly susceptible to the effects of fetal and maternal movements compared to other weighted images, complicating the acquisition of satisfactory results. This study aimed to compare three fast 3D-T1 weighted gradient echo (GRE) sequences—free-breathing stack-of-stars VIBE (STAR-VIBE), breath-hold VIBE (BH-VIBE), and free-breathing multi-average VIBE (MA-VIBE)—for fetal gastrointestinal MRI in fetuses with both normal and abnormal gastrointestinal tracts between 21 and 36 weeks of gestation. MethodsThis study enrolled 67 pregnant women who underwent fetal abdominal MRI at our hospital between October 2022 and October 2023, during their gestational period of 21–36 weeks. Among these participants, 22 were suspected of having fetal gastrointestinal anomalies based on ultrasound findings, while the remaining 45 were considered to have normal fetal gastrointestinal development. All subjects underwent True fast imaging with steady-state precession sequence scanning along with three T1-weighted imaging techniques on a Siemens 1.5-T Aera scanner: STAR-VIBE, BH-VIBE, and MA-VIBE. Two radiologists evaluated image quality, intestinal clarity, and lesion conspicuity using a five-point scale where higher scores indicated superior performance for each technique; they were blinded to the acquisition schemes used. Interobserver variability assessments were also conducted. ResultsThe free-breathing MA-VIBE sequence demonstrated significantly better performance than both STAR-VIBE and BH-VIBE in terms of fetal gastrointestinal MRI quality (3.81 ± 0.40 vs. 3.35 ± 0.70 vs. 2.90 ± 0.64; p < .05). The STAR-VIBE and BH-VIBE sequences exhibited moderate consistency (kappa = 0.586 and kappa = 0.527 respectively; P < .05), whereas the MA-VIBE sequence showed higher consistency (kappa = 0.712; P < .05). ConclusionThe free-breathing MA-VIBE sequence provided superior visualization for assessing fetal intestinal conditions compared to other methods employed in this study. On a 1.5 T MRI device, T1-weighted images based on the free-breathing MA-VIBE sequence can effectively overcome motion artifacts and compensate for the reduced signal-to-noise ratio caused by the application of acceleration techniques, thus significantly improving the quality of T1-weighted images.

Clinical Implication of the New AI-TIRADS Classification of Thyroid Nodules; Our Real Clinical Experience

Background: Ultrasound and fine needle aspiration biopsy (FNAB) play a primary role in determining the nature of nodules. Objectives: The aim of our study is to determine the diagnostic accuracy, sensitivity, and specificity of artificial intelligence-thyroid imaging reporting and data system (AI-TIRADS) in differentiating malignant from benign nodules to establish the treatment plan for patients and avoid unnecessary FNAB. Methods: This cross-sectional study was conducted between September 2020 and September 2021 (one year). Patients (n = 133) who had an indication for surgery due to nodular thyroid disease and were referred to the therapeutic training center of 5 Azar and Falsafi private section hospitals, Gorgan, Iran, were included. The diagnostic accuracy of FNAB and AI-TIRADS was calculated. Results: The mean age of patients with malignant pathology (39.65 ± 11.69) was lower than that of patients with benign pathology (44.14 ± 13.54), which was statistically significant (P = 0.042). Fifty-two out of 72 nodules less than 25 mm were malignant, demonstrating a higher prevalence of benign pathology in larger nodules. A pooled analysis of thyroid reporting (TR) = 2 and TR points ≥ 7 for their power in predicting the final pathology outcome showed figures of 0.97, 0.93, 0.97, and 0.97 for sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV), respectively, with an overall pooled accuracy of 0.97 for these two groups of sonographic results. Conclusions: We concluded that AI-TIRADS has a high sensitivity and diagnostic accuracy in differentiating benign from suspicious nodules.

B-037 Zeon original coating technology COP microplates for cell imaging

Abstract Background Zeon has developed high quality microplate made of Cyclo Olefin Polymer (COP) and special coat reagent for cell culture for Aurora microplates. The new 96 well-microplates have best-in-class properties for analysis of cultured cells can reduce the background of fluorescence images absolutely and excellent bottom flatness for observation of target. This enables to obtain high-precision data for cell analysis. Also, we developed Zeon original coating (ZOC) reagent specialized for COP surface. It is suitable for various cells including neurons derived from human induced pluripotent stem cells (iPSC), primary cells, general cell lines, and so on. ZOC coated COP microplates (ZOC plates) are fully-precoated and ready-to-use. Methods We evaluated the following points regarding ZOC plates. The cells used were derived from human iPSC. [1. Storage stability] ZOC plates were stored at room temperature for a long time. After that, cardiomyocyte cells were cultured for 3 days on the plates. The cell adhesion and beat rate were evaluated. [2. Neurite extension] Dopaminergic neurons were cultured for 14 days, then the neurite extension was evaluated. For comparison, a conventional poly-D-lysine* (PDL) coating was used. [3. Lot-to-lot variation] COP microplates coated with different lots of ZOC reagents were prepared. Glutamatergic neurons were cultured for 14 days on those plates, then the neurite extension and cell morphology were evaluated. *The typical coating reagent applied for neuronal culture. Results [1.] ZOC plates stored over 12 months at room temperature showed excellent cell adhesion, and the beat rate was equivalent to prepared just before use. [2.] ZOC plates tended to extend neurites more than PDL coated plates. [3.] No variation between lots was observed on ZOC plates regarding the neurite extension and cell morphology. Conclusions We suggest applying the combination of high accuracy for cell imaging using COP microplates and high performance for cell culture using ZOC can provide high precision data at bioassay research. This would be a significant value for development of drug screening processes at pharmaceutical company.

MFH‐Net: A Hybrid CNN‐Transformer Network Based Multi‐Scale Fusion for Medical Image Segmentation

ABSTRACTIn recent years, U‐Net and its variants have gained widespread use in medical image segmentation. One key aspect of U‐Net's design is the skip connection, facilitating the retention of detailed information and leading to finer segmentation results. However, existing research often concentrates on enhancing either the encoder or decoder, neglecting the semantic gap between them, and resulting in suboptimal model performance. In response, we introduce Multi‐Scale Fusion module aimed at enhancing the original skip connections and addressing the semantic gap. Our approach fully incorporates the correlation between outputs from adjacent encoder layers and facilitates bidirectional information exchange across multiple layers. Additionally, we introduce Channel Relation Perception module to guide the fused multi‐scale information for efficient connection with decoder features. These two modules collectively bridge the semantic gap by capturing spatial and channel dependencies in the features, contributing to accurate medical image segmentation. Building upon these innovations, we propose a novel network called MFH‐Net. On three publicly available datasets, ISIC2016, ISIC2017, and Kvasir‐SEG, we perform a comprehensive evaluation of the network. The experimental results show that MFH‐Net exhibits higher segmentation accuracy in comparison with other competing methods. Importantly, the modules we have devised can be seamlessly incorporated into various networks, such as U‐Net and its variants, offering a potential avenue for further improving model performance.

HyFormer: a hybrid transformer-CNN architecture for retinal OCT image segmentation

Optical coherence tomography (OCT) has become the leading imaging technique in diagnosing and treatment planning for retinal diseases. Retinal OCT image segmentation involves extracting lesions and/or tissue structures to aid in the decisions of ophthalmologists, and multi-class segmentation is commonly needed. As the target regions often spread widely inside the retina, and the intensities and locations of different categories can be close, good segmentation networks must possess both global modeling capabilities and the ability to capture fine details. To address the challenge in capturing both global and local features simultaneously, we propose HyFormer, an efficient, lightweight, and robust hybrid network architecture. The proposed architecture features parallel Transformer and convolutional encoders for independent feature capture. A multi-scale gated attention block and a group positional embedding block are introduced within the Transformer encoder to enhance feature extraction. Feature integration is achieved in the decoder composed of the proposed three-path fusion modules. A class activation map-based cross-entropy loss function is also proposed to improve segmentation results. Evaluations are performed on a private dataset with myopic traction maculopathy lesions and the public AROI dataset for retinal layer and lesion segmentation with age-related degeneration. The results demonstrate HyFormer's superior segmentation performance and robustness compared to existing methods, showing promise for accurate and efficient OCT image segmentation. .

Performance and Clinical Implications of VI-RADS in detecting muscle invasion in bladder tumors: A prospective observational study

ABSTRACT Introduction: Bladder cancer staging is pivotal for guiding therapeutic decisions. In this study, the primary objective was to assess the diagnostic accuracy of multiparametric magnetic resonance imaging (mpMRI), The study aimed to classify bladder tumors as either nonmuscle-invasive bladder cancers or muscle-invasive bladder cancers (MIBC) using the Vesical Imaging Reporting and Data System (VI-RADS) scoring. A secondary objective of the study focused on the accuracy of biparametric magnetic resonance imaging (bpMRI) in comparison to mpMRI. Methods: Thirty-three patients with bladder tumors were enrolled and underwent both mpMRI and bpMRI scoring assessments. VI-RADS scores were assigned and subsequently compared with histopathological findings posttransurethral resection of bladder tumor., Statistical measures included sensitivity, specificity, positive predictive value, negative predictive value, and diagnostic accuracy for VI-RADS scores at cutoff thresholds of ≥4 and 5. Results: MpMRI at a VI-RADS cutoff of ≥4 yielded an 83.33% sensitivity and 100% specificity, with a diagnostic accuracy of 90.91%. At a cutoff of 5, sensitivity was 100%, specificity was 77.27%, and diagnostic accuracy was 84.85%. bpMRI at a cutoff of ≥4 showed an 80% sensitivity and 100% specificity, with diagnostic accuracy matching mpMRI at 90.91%. Conclusions: This prospective analysis demonstrates that VI-RADS scoring with mpMRI provides reliable diagnostic accuracy for bladder cancer staging. mpMRI exhibits high sensitivity and specificity at a cutoff of ≥ 4, making it a robust tool for MIBC detection. bpMRI is an effective alternative in select patients. The study validates the use of VI-RADS scoring in clinical practice for effective treatment planning.

Role Of Digital Image Processing In Education And Medical Field

Digital image processing is a technique that uses computer algorithms to analyze images and transform images into better ones.It is an important branch of Telecommunication engineering that deals with the improvisation of reliability and accuracy of the digital communication by employing multiple techniques. Image processing involves many processes such as taking images, preprocessing, image enhancement, transformation and analysis. Variations in pixel values are one of the major challenges faced by image processing. Digital Signal Processing is at the core of virtually all of today's information technology, and its impact is felt everywhere in telecommunications, medical technology, radar and sonar, and in seismic data analysis.Digital Image Processing plays an important role in medical field and education. Various researches has been done and currently popularly used. In education, pictures can be used to enhance student learning and to develop students’ self-esteem, or their utilization can be of little or no value and, if used carelessly, can, in fact, undermine students' self-confidence. Digital health technologies include both hardware and software solutions and services, including telemedicine, wearable devices, augmented reality, and virtual reality. Generally, digital health interconnects health systems to improve the use of computational technologies, smart devices, computational analysis techniques, and communication media to aid healthcare professionals and their patients manage illnesses and health risks, as well as promote health and wellbeing. Digital imaging gives healthcare providers better and more accurate images, which leads to establishing diagnoses sooner, making treatment of the patient more successful.

Engineering AIEgens-Tethered Gold Nanoparticles with Enzymatic Dual Self-Assembly for Amplified Cancer-Specific Phototheranostics.

Accurate imaging and precise treatment are critical to controlling the progression of pancreatic cancer. However, current approaches for pancreatic cancer theranostics suffer from limitations in tumor specificity and invasive surgery. Herein, a pancreatic cancer-specific phototheranostic modulator (AuHQ) dominated by aggregation-induced emission (AIE) luminogens-tethered gold nanoparticles is meticulously designed to facilitate prominent fluorescence-photoacoustic bimodal imaging-guided photothermal immunotherapy. Once reaching the pancreatic tumor microenvironment (TME), the peptide Ala-Gly-Phe-Ser-Leu-Pro-Ala-Gly-Cys (AGFSLPAGC) linkage within AuHQ can be specifically cleaved by the overexpressed enzyme Cathepsin E (CTSE), triggering the dual self-assembly of AuNPs and AIE luminogens. The aggregation of AuNPs mediated by enzymatic cleavage results in potentiated photothermal therapy (PTT) under near-infrared (NIR) laser irradiation, induced immunogenic cell death (ICD), and enhanced photoacoustic imaging. Simultaneously, AIE luminogen aggregates formed by hydrophobic interaction can generate AIE fluorescence, enabling real-time and specific fluorescence imaging of pancreatic cancer. Furthermore, coadministration of an indoleamine 2,3-dioxygenase 1 (IDO1) inhibitor with AuHQ can address the limitations of PTT efficacy imposed by the immunosuppressive TME and leverage the synergistic potential to activate systemic antitumor immunity. Thus, this well-designed phototheranostic modulator AuHQ facilitates the intelligent enzymatic dual self-assembly of imaging and therapeutic agents, providing an efficient and precise approach for pancreatic cancer theranostics.

Body image at the trunk: An investigation into externally referenced width perception and picture mapping.

Body image is a conscious representation of the body, encompassing how our body feels to us. Body image can be measured in a variety of ways, including metric and depictive measures. This study sought to assess body image at the trunk by investigating, and comparing, a metric and depictive measure. Sixty-nine healthy participants estimated their thorax, waist, and hip width by externally referencing mechanical calipers. Participants were also asked to select the true image of their trunk from a random display of nine images containing the true image and incrementally shrunken or enlarged images. Participants demonstrated evidence of thorax and waist width overestimation in the width perception task, with no evidence for hip misestimation. For the picture mapping task, the majority of participants were inaccurate. In participants who were inaccurate, approximately equal proportions underestimated and overestimated their trunk width. The two tasks were found to be independent of each other. Distortions, or inaccuracies, were apparent in a metric measure, and inaccuracies also present in a depictive measure, of body image at the trunk for healthy participants. An overestimation bias was apparent in the metric, but not depictive, task. No relationship was found between tasks..

Instrumental Diagnosis of Placenta Accreta and Obstetric and Perinatal Outcomes: Literature Review and Observational Study.

Placenta accreta (PA) is a condition where the placenta is pathologically adherent to the uterus due to a defect in the basal decidua with myometrium invasion by chorionic villi and is classified based on the depth of myometrial invasion by histology. However, ultrasound and magnetic resonance imaging have excellent accuracy. In this study, we investigated clinical benefits of early instrumental diagnosis of PA, especially in reducing maternal-fetal complications and improving perinatal outcomes. We also evaluated diagnostic accuracy of ultrasound and magnetic resonance imaging on placental invasiveness assessment. In this review and observational retrospective study, risk factors of PA were collected, and pregnant women underwent third-trimester ultrasound and magnetic resonance imaging (MRI) to evaluate the degree of infiltration. Imaging results compared to histological findings and surgical evaluation. A total of 38 patients were diagnosed with at the University Hospital "San Giovanni di Dio and Ruggi d'Aragona", Salerno, Italy, by second-trimester ultrasound with high sensitivity (100%) and accuracy (86%). Moreover, 37 of them performed MRI and 60.5% were diagnosed with Accreta, 7.9% increta, 10.5% percreta, and 21.1% not accrue with high sensitivity (100%), specificity (88.9%), and accuracy (97.4%). Histological assay confirmed MRI findings in 96.7% of cases. Risk factors of PA were age >35 years and previous CT scans. In unborn babies, mean 1-min Apgar was 4.3 (range, 3-6), and mean 5-min Apgar was 7.13 (range, 7-9). MRI could be a not-invasive, specific, sensitive, and accurate diagnostic tool for assessing the degree of infiltration in PA, and could guide clinical decisions, such as delivery plan, thus reducing perioperative and fetal complications.

Accuracy, repeatability, and reproducibility of water-fat magnetic resonance imaging in a phantom and healthy volunteer

Bone marrow (BM) damage due to chemoradiotherapy can increase BM fat in cervical cancer patients. Water-fat magnetic resonance (MR) scans were performed on a phantom and a healthy female volunteer to validate proton density fat fraction accuracy, reproducibility, and repeatability across different vendors, field strengths and protocols. Phantom measurements showed a high accuracy, high repeatability, and excellent reproducibility. Volunteer measurements had an excellent intra- and interreader reliability, good repeatability, and moderate to good reproducibility. Water-fat MRI show potential for quantification of longitudinal vertebral BM fat changes. Further studies are needed to validate and extend these findings for broader clinical applicability.

Exploring the Potential of the Hyperspectral Remote Sensing Data China Orbita Zhuhai-1 in Land Cover Classification

Responding to the shortcomings of China's civil remote sensing data in land cover classification, such as the difficulty of data acquisition and the low utilization rate, we used Landsat-8, China Orbita Zhuhai-1 hyperspectral remote sensing (OHS) data, and Landsat-8 + OHS data combined with band (red, green, and blue) and vegetation index features to classify land cover using maximum likelihood (ML), Mahalanobis distance (MD), and support vector machine (SVM). The results show that Landsat-8 + OHS data have the highest classification accuracy in SVM, with an overall accuracy of 83.52% and a kappa coefficient of 0.71, and this result is higher than that of Landsat-8 images and OHS images separately. In addition, the classification accuracy of OHS images was higher than that of Landsat-8 images. The results of the study provide a reference for the use of civil satellite remote sensing data in China.

Accuracy of Myocardial Perfusion Imaging (MPI) in Predicting Post-transplant Cardiovascular Events among Asymptomatic Diabetic Kidney Transplant Candidates

Accuracy of Myocardial Perfusion Imaging (MPI) in Predicting Post-transplant Cardiovascular Events among Asymptomatic Diabetic Kidney Transplant Candidates

Assessing Transducer Parameters for Accurate Medium Sound Speed Estimation and Image Reconstruction.

The influence of the transducer lens on image reconstruction is often overlooked. Lenses usually exhibit a lower sound speed than soft biological tissues. In academic research, the exact lens sound speed and thickness are typically unknown. Here, we present a simple and nondestructive method to characterize the lens sound speed and thickness as well as the time to peak of the round-trip ultrasound waveform, another key parameter for optimal image reconstruction. We applied our method to three transducers with center frequencies of 2.5, 7.5, and 15 MHz. We estimated the three parameters with an element-by-element transmission sequence that records internal reflections within the lens. We validated the retrieved parameters using an autofocusing approach that estimates sound speed in water. We show that the combination of our parameters estimation method with two-layer ray tracing outperforms standard image reconstruction. For all transducers, we successfully improved the accuracy of medium sound speed estimation, spatial resolution, and contrast. The proposed method is simple and robust and provides an accurate estimation of the transducer lens parameters and the time to peak of the ultrasound waveform, which leads to improved ultrasound image quality.

Integrated detection for open and closed surface fatigue cracks utilizing scanning laser source-induced Rayleigh wave fields with self-reference peak-to-peak features

In this study, the integrated detection method for open and closed surface fatigue cracks is investigated. Firstly, the finite element simulation models are established to investigate the interaction between scanning laser source-induced Rayleigh waves and open and closed surface fatigue cracks, and the laser ultrasonic detection experiments are conducted for fatigue specimens containing fatigue cracks in different states. The simulation and experimental results indicate that the variation patterns in the peak-to-peak values of Rayleigh waves with horizontal scanning positions for open and closed cracks exhibit both similarities and differences. Subsequently, an integrated detection method based on self-referenced peak-to-peak features is proposed, which utilizes the similarities and differences to detect and distinguish open and closed cracks, respectively. Furthermore, this proposed method is experimentally validated, indicating that it can achieve accurate integrated imaging of open and closed cracks that have a high degree of agreement with the SEM images of the cracks. Additionally, the proposed method achieves a detection error of 2 % for the vertical lengths of the fatigue cracks. This study can provide guidance for integrated real-time detection of open and closed surface fatigue cracks of mechanical components in service.

Enhanced probabilistic damage localization for composite plate based on joint recurrence plots

Localizing damage is a key aspect of monitoring composite laminate structural health. To address the problem of generating the damage index (DI) of composite laminates using Lamb waves, a reconstruction approach for probabilistic damage inspection (RAPID) based on joint recurrence plots (JRP) was proposed. First, the dynamics of the Lamb wave signals are analyzed using phase space reconstruction. Second, the damage information in each sensing path is represented using JRP, and the DI based on the JRP is built with variational auto-encoder (VAE). Finally, the relative magnitude of the DI was employed to modify RAPID’s weight distribution function, allowing for composite plate damage localization detection. The findings of finite element modeling and experiment data reveal that the approach can achieve accurate localization imaging of the composite plate’s interior damage while also properly identifying the damage under 20 dB and 10 dB noise interference. The approach does not require an understanding of Lamb wave dispersion characteristics and involve the extraction of complicated wave packet signals, making it promising for structural health monitoring based on damage index.

Optimization and correction of breast dynamic optical imaging projection data based on deep learning

Breast cancer poses a significant health threat to women, necessitating advancements in diagnostic technologies. Breast dynamic optical imaging (DOI) technology, recognized for its non-invasive and radiation-free properties, is extensively utilized for the early screening and quantitative analysis of breast tumors. The integration of deep learning, a robust technology for automatic image feature extraction, with breast DOI has the potential to enhance tumor detection and diagnosis significantly. This paper introduces a deep learning-enhanced image optimization approach to overcome challenges such as poor image quality and distorted projection data commonly encountered in existing DOI methods. The approach utilizes convolutional neural networks (CNNs) to extract features from raw images and employs generative adversarial networks (GANs) to enhance these images, thereby improving their quality and contrast. Additionally, a novel correction algorithm is developed to address projection data distortion, enabling the reconstruction and correction of this data for more accurate and reliable imaging results. Experimental findings confirm that the proposed method markedly enhances both image quality and projection data accuracy in breast DOI, offering a reliable foundation for clinical diagnosis. This study not only provides a new perspective and methodology for the early screening and diagnosis of breast cancer but also holds substantial clinical importance and prospective applications.

A Distance Transformation Deep Forest Framework With Hybrid-Feature Fusion for CXR Image Classification.

Detecting pneumonia, especially coronavirus disease 2019 (COVID-19), from chest X-ray (CXR) images is one of the most effective ways for disease diagnosis and patient triage. The application of deep neural networks (DNNs) for CXR image classification is limited due to the small sample size of the well-curated data. To tackle this problem, this article proposes a distance transformation-based deep forest framework with hybrid-feature fusion (DTDF-HFF) for accurate CXR image classification. In our proposed method, hybrid features of CXR images are extracted in two ways: hand-crafted feature extraction and multigrained scanning. Different types of features are fed into different classifiers in the same layer of the deep forest (DF), and the prediction vector obtained at each layer is transformed to form distance vector based on a self-adaptive scheme. The distance vectors obtained by different classifiers are fused and concatenated with the original features, then input into the corresponding classifier at the next layer. The cascade grows until DTDF-HFF can no longer gain benefits from the new layer. We compare the proposed method with other methods on the public CXR datasets, and the experimental results show that the proposed method can achieve state-of-the art (SOTA) performance. The code will be made publicly available at https://github.com/hongqq/DTDF-HFF.