Image Analysis System Research Articles

A Federated Learning System for Histopathology Image Analysis with an Orchestral Stain-Normalization GAN.

Currently, data-driven based machine learning is considered one of the best choices in clinical pathology analysis, and its success is subject to the sufficiency of digitized slides, particularly those with deep annotations. Although centralized training on a large data set may be more reliable and more generalized, the slides to the examination are more often than not collected from many distributed medical institutes. This brings its own challenges, and the most important is the assurance of privacy and security of incoming data samples. In the discipline of histopathology image, the universal stain-variation issue adds to the difficulty of an automatic system as different clinical institutions provide distinct stain styles. To address these two important challenges in AI-based histopathology diagnoses, this work proposes a novel conditional Generative Adversarial Network (GAN) with one orchestration generator and multiple distributed discriminators, to cope with multiple-client based stain-style normalization. Implemented within a Federated Learning (FL) paradigm, this framework well preserves data privacy and security. Additionally, the training consistency and stability of the distributed system are further enhanced by a novel temporal self-distillation regularization scheme. Empirically, on large cohorts of histopathology datasets as a benchmark, the proposed model matches the performance of conventional centralized learning very closely. It also outperforms state-of-the-art stain-style transfer methods on the downstream Federated Learning image classification task, with an accuracy increase of over 20.0% in comparison to the baseline classification model.

Interpretation of stripping at the bitumen–aggregate interface based on fluorescence tracing method

Asphalt pavement field performance largely depends on the stripping resistance at the bitumen–aggregate interface. However, the conventional stripping resistance evaluation method is unsuitable for direct and precise quantification of the retained bitumen film area on the aggregate surface that designates the stripping resistance between bitumen and aggregate phases under the action of vehicle wheels or moisture damage. This study proposed a novel fluorescence tracing method to quantify the exact area of aggregate coating loss in accordance with the bituminous-coated aggregate image processing technique. The fluorescence tracing image acquisition and analysis system for precisely calculating the bitumen stripping ratio of the bituminous-coated aggregate were established after the bitumen stripping test. The thresholding manner of the two-mode threshold algorithm was devised to segment the fluorescence-traced bituminous-coated aggregate image with high quality. The results demonstrated that the appropriate threshold value for segmenting the fluorescence-traced image mainly falls in the range of 60–70 for limestone, 90–110 for basalt, and 100–110 for both granites; the ability of aggregates based on their chemisorption capacity with bitumen is in the order of limestone, basalt, and granite; the fluorescence tracing method eliminated nearly 92.3% of the undesirable reflection on bitumen phase; the detected bitumen stripping ratio, raised by 65.87% on average and by 37.82% in the median value in contrast to the conventional visual detection manner. This research introduces a practical framework from an engineering perspective for visually quantifying the interpretation of the adhesive bond at the bitumen-aggregate interface with reasonable accuracy and confidence.

Different patterns of mast cell distribution in B-cell non-Hodgkin lymphomas

Tumor growth, progression, and metastatic capability in non-Hodgkin lymphomas (NHLs) are influenced by different component of tumor microenvironment, including inflammatory cells. Among these latter, mast cells play a crucial role. The spatial distribution of mast cells inside the tumor stroma of different types of B-cell NHLs has not yet been investigated. The aim of this study is to analyze the pattern of distribution of mast cells in biopsy samples obtained from three different types of B-cell NHLs by utilizing an image analysis system and a mathematical model to allow a quantitative estimation to characterize their spatial distribution. As concerns the spatial distributions exhibited by mast cells in diffuse large B cell lymphoma (DLBCL), some clustering was detected in both activated B-like (ABC) and germinal center B-like (GBC) groups. In follicular lymphoma (FL), mast cell spatial distribution tends to uniformly fill the tissue space as far as the grade of the pathology increases. Finally, in marginal lymphoma tissue (MALT) lymphoma, mast cells maintain a significantly clustered spatial distribution, suggesting a lower tendency of the cells to fill the tissue space in this pathological condition. Overall, the data of this study confirm that the analysis of the spatial distribution of the tumor cells is of particular significance for the knowledge of the biological processes occurring in tumor stroma and for the development of parameters to characterize the morphologic organization of the cellular patterns in different types of tumors.

P-170 A sibling oocyte study using a real-time image analysis system to identify optimal puncture positions in Piezo-ICSI

Abstract Study question Is the newly developed image analysis system capable of visualizing the shape features of the oolemma in real-time, useful for reducing oocyte degeneration during Piezo-ICSI? Summary answer Unintentional membrane rupture (UMR) can be reduced by this real-time image analysis, which lowers the risk of oocyte degeneration following Piezo-ICSI. What is known already UMR in the puncturing process of ICSI often predisposes oocyte degeneration. Identifying the appropriate puncturing positions may decrease the likelihood of UMR and thus degeneration, but this cannot be visualized via a microscope. We have reported that visualization of the shape features of the oolemma using moving image analysis during Piezo-ICSI can locate areas on the oolemma where UMR is likely to occur. It is feasible to assess the incidence of UMR and oocyte degeneration by performing ICSI while selecting appropriate puncturing positions with the aid of the newly developed imaging system. Study design, size, duration Our team have developed an image analysis system called ICSI POSITION DETECTOR (IPD), which can identify areas where rupture is likely to occur and visualize it in real-time using a video monitor. We prospectively evaluated the usefulness of IPD in a sibling oocyte study. From January 2020 to August 2021, a total of 1268 oocytes obtained in 225 oocyte retrieval cycles (average maternal age: 38.3 ± 4.7 years old) was included. Participants/materials, setting, methods The oocytes were randomly assigned to two groups: IPD-using vs. non-IPD-using. In the IPD-using group, Piezo-ICSI was performed at the “appropriate” position with a low chance of UMR indicated by IPD. In the non-IPD-using group, Piezo-ICSI was performed blindly. The rates of UMR, oocyte degeneration, fertilization and embryonic development were compared between the two groups. In addition, in the non-IPD-using group, moving images were recorded during Piezo-ICSI and analyzed retrospectively using IPD. Main results and the role of chance The rates of UMR and degeneration were significantly lower in the IPD-using group compared to the non-IPD-using group (6.0% vs. 11.9%, P < 0.001 and 1.6% vs. 4.6%, P < 0.01, respectively). The rates of fertilization (83.7% vs. 79.7%), blastocyst formation (51.9% vs. 51.0%), and good-quality blastocyst (24.7% vs. 24.3%) were not significantly different. Retrospective analysis of moving images using IPD on the non-IPD-using group showed that in 45.3% of oocytes (286 out of 632) ICSI was performed at a position with a high chance of UMR (inappropriate position). When ICSI was performed at the appropriate positions using IPD, the rates of UMR (6.0% vs. 18.2%, P < 0.001) and degeneration (1.6% vs. 7.3%, P < 0.001) were significantly lower, while the rates of fertilization (83.7% vs. 74.5%, P < 0.01) and blastocyst formation (51.9% vs. 43.0%, P < 0.05) were significantly higher than those when ICSI was performed at an inappropriate position. The rate of good-quality blastocyst (24.7% vs. 19.7%) was also higher, but not statistically significant. These results indicate that IPD can reduce the risk of UMR, thereby lower the degeneration rate. Furthermore, the embryonic development was better when using IPD to identify the appropriate position to perform ICSI, suggesting that UMR seems to be associated with poor ICSI outcome. Limitations, reasons for caution In this study, Piezo-ICSI was performed by only two designated embryologists in a single center. It did not assess clinical outcomes. Further research involving more embryologists is needed. Moreover, it would be feasible to investigate whether the usage of IPD is similarly effective on conventional-ICSI outcome. Wider implications of the findings The application of IPD to perform ICSI at the appropriate position can significantly avoid UMR, and thereby reduce oocyte degeneration and provide better embryonic development. We therefore, consider IPD to be a highly clinically useful tool, which contributes to the production of more embryos that can be used for treatment. Trial registration number not applicable

AI image-based diagnosis systems: how to implement them?

Artificial intelligence (AI) is starting to be widely used in the medical field and has great potential benefits to help doctors and patients. However, it also raises new challenges and problems. This paper analyzed the existing capacities of AI to make a diagnosis and assessed the legal consequences. We present AI medical image analysis systems developed in Belarus. International practice on how AI-systems are implemented in medicine is analyzed. Russian experience in developing standards to test and use AI systems in hospitals is described. Finally, the paper put forward some suggestions on how to improve the legal framework of AI systems using in medicine.

Molecular characterization of Fasciola hepatica in endemic regions of Colombia.

Fasciola hepatica is a zoonotic trematode that affects a wide range of hosts, including cattle, sheep, and goats. The economic impact of the parasite on the cattle industry is significant, with high losses reported worldwide. While its impact on human health was previously underestimated, recent years have seen a rise in fascioliasis cases, leading to increased interest among researchers globally. To characterize the genetic diversity and intraspecific variation of this parasite in South America, specifically in Colombia, we collected 105 adult parasites from cattle bile ducts in seven Colombian departments (Antioquia, Boyacá, Santander, Cauca, Cundinamarca, Nariño, Norte de Santander, and Santander) to assess the parasite's phenotypic analyses, genetic diversity, and population structure. A computer image analysis system (CIAS) was applied based on standardized morphological measurements. Liver-fluke size was studied by principal component analysis (PCA). DNA sequences were obtained for nuclear markers such as the 28S, β-tubulin 3, ITS1, ITS2, and the mitochondrial marker Cytochrome Oxidase I (COI). Multiple statistical tests were performed, and the parasite's population structure was analyzed. Maximum Likelihood (ML) phylogenetic reconstructions were carried out using the sequences obtained herein and sequences available in GenBank. Morphological results revealed that all the obtained individuals matched F. hepatica's morphology. There was no evidence of high genetic diversity, and the absence of genetic structure at the country-level was notable, possibly caused by a demographic expansion of this trematode in Colombia or the low resolution of the molecular markers employed. Future studies are still needed to unveil the genetic population structure of F. hepatica across the country.

The evaluation of IgG4 and IgG expression in cutaneous Rosai-Dorfman disease

ObjectiveThe authors investigated the expression of IgG4 and IgG in cutaneous Rosai-Dorfman Disease (CRDD) to further improve the understanding of this disease. MethodsThe authors retrospectively reviewed the clinicopathological features of 23 CRDD patients. The authors diagnosed CRDD by the presence of emperipolesis and immunohistochemical (IHC) staining of histiocytes consisting of S-100(+)/CD68(+)/CD1a(-) cells. The expressions of IgG and IgG4 in cutaneous specimens were assessed by IHC (EnVision) and quantitatively calculated by a medical image analysis system. ResultsAll 23 patients, including 14 males and 9 females, were confirmed to have CRDD. Their ages ranged from 17 to 68 years (mean 47.91 ± 14.16). The most frequently affected skin regions were the face, followed by the trunk, ears, neck, limbs, and genitals. In 16 of these cases, the disease presented as a single lesion. IHC staining of sections showed that IgG was positive (≥ 10 cells/High-Power Field [HPF]) in 22 cases, while IgG4 was positive (≥ 10 cells/HPF) in 18 cases. Moreover, the IgG4/IgG proportion ranged from 1.7% to 85.7% (mean 29.50 ± 24.67%, median 18.4%) in the 18 cases. Study limitationsIn the majority of studies, as well as in the current study, the design. RDD is a rare disease, so the sample size is small. In the next studies to come, the authors will expand the sample for multi-center verification and in-depth study. ConclusionThe positive rates of IgG4 and IgG and the IgG4/IgG ratio assessed through IHC staining may be important in understanding the pathogenesis of CRDD.

Ganoderenic acid D-loaded functionalized graphene oxide-based carrier for active targeting therapy of cervical carcinoma

Ganoderenic acid D (GAD), extracted from the Chinese herb Ganoderma lucidum, was loaded onto a graphene oxide-polyethylene glycol-anti-epidermal growth factor receptor (GO-PEG-EGFR) carrier to develop a targeting antitumor nanocomposite (GO-PEG@GAD). The carrier was fabricated from PEG and anti-EGFR aptamer modified GO. Targeting was mediated by the grafted anti-EGFR aptamer, which targets the membrane of HeLa cells. Physicochemical properties were characterized by transmission electron microscopy, dynamic light scattering, X-ray powder diffraction, and Fourier transform infrared spectroscopy. High loading content (77.3 % ± 1.08 %) and encapsulation efficiency (89.1 % ± 2.11 %) were achieved. Drug release continued for approximately 100 h. The targeting effect both in vitro and in vivo was confirmed by confocal laser scanning microscopy (CLSM) and imaging analysis system. The mass of the subcutaneous implanted tumor was significantly decreased by 27.27 ± 1.23 % after treatment with GO-PEG@GAD compared with the negative control group. Moreover, the in vivo anti-cervical carcinoma activity of this medicine was due to activation of the intrinsic mitochondrial pathway.

EARLY INDICES OF SUBCLINICAL CARDIOVASCULAR DAMAGE IN PATIENTS WITH ESSENTIAL HYPERTENSION AND PRIMARY ALDOSTERONISM. A PRELIMINARY OBSERVATIONAL STUDY

Objective: In hypertensive patients, the activation of the renin-angiotensin-aldosterone system is associated with an increase of subclinical vascular and cardiac damage. Our study aimed to compare some early markers of cardiovascular organ damage between patients with essential hypertension and patients with secondary hypertension, in particular primary aldosteronism. Design and method: Fifty-seven hypertensive patients (mean age: 44,4±13,6 years, 60.7% males) were included, 47 with essential hypertension and 10 with primary hyperaldosteronism. Peripheral blood pressure (pBP), central blood pressure (cBP) and pulse wave velocity (PWV) were assessed by the SphygmoCorXCEL. The carotid Intima-Media Thickness (cIMT) and the carotid Distensibility Coefficient (cCD) were measured by a scanner connected with an image acquisition and analysis system (CardiovascularSuite). Left ventricular mass index (LVMI) and relative wall thickness (RWT) were evaluated by transthoracic echocardiography. Results: Patients with primary aldosteronism have higher cIMT and PWV values compared to patients with essential hypertension (0.7±0.1mm vs. 0.6±0.1 mm, p = 0.02 and 8.6±2.5 m/s vs. 7.2±1.5 m/s, p = 0.02; respectively). By linear regression, age was the only variable independently associated with cIMT. On the other hand, various factors such as age, the diagnosis of primary aldosteronism, pSBP and cSBP were independently associated with PWV. Although not statistically significant, LVMI and RWT tended to be higher in the group with primary aldosteronism in comparison to patients with essential hypertension (14.3% vs. 0% exhibited concentric hypertrophy and 42.8% vs. 28.1% showed findings consistent with concentric cardiac remodelling). No differences were found in cDC values between the two groups. Conclusions: The results of this study confirm a possible deleterious impact of aldosteronism on cardiovascular damage in hypertensive patients, more evident for PWV than for cIMT or DC. This fact could suggest a possible different effect of the abnormal hormone secretion at different vascular districts.

SUBCLINICAL TARGET ORGAN DAMAGE IN PATIENTS WITH TYPE 1 DIABETES MELLITUS. A PROSPECTIVE PRELIMINARY STUDY

Objective: Early atherosclerosis is the main cause of cardiovascular damage and death in patients with type 1 Diabetes Mellitus (T1DM). Some vascular indexes can help to identify and stratify patients before clinical events. This study aims to evaluate the trend over time of some markers of subclinical vascular damage to identify possible associations with anthropometric parameters, blood pressure (BP) and metabolic variables. Design and method: Fifty-six patients with T1DM (mean age: 17.5±2.5 years; mean duration of diabetes: 8.9±4.0 years) were assessed in two occasions (mean follow-up time: 2.7±0.9 years). Peripheral systolic and diastolic BP (pSBP, pDBP), central systolic and diastolic BP (cSBP, cDBP) and Pulse Wave Velocity (PWV) were measured by the SphygmoCorXCEL. Carotid Intima-Media Thickness (cIMT) and the carotid Distensibility Coefficient (cCD) were assessed by ultrasound connected with an image acquisition and analysis system (CardiovascularSuite). Biochemical and anthropometric data were also recorded. Results: At follow-up, PWV and BP values increased significantly while cDC further decreased with the 80.4% of patients that presented a pathological cDC (lower than 40x10 -3 KPa or the 5th percentile). BP values at baseline correlated with all vascular indexes at follow-up: in particular, cSBP was associated with cDC (r = -0.46, p< 0.001), cIMT (r = 0.39, p = 0.004) and PWV (r = 0.32, p = 0.01). These associations remained significant even after adjustment for metabolic (including glycated hemoglobin) and anthropometric parameters. Moreover, baseline arterial stiffness significantly correlated with BP values at follow-up: particularly cDC and PWV were associated with cDBP (r = -0.354, p = 0.009 and r = 0.40, p = 0.002; respectively) even after multiple adjustments. Regarding metabolic parameters, non-HDL-to-HDL-cholesterol ratio correlated with both cDC (r = -0.36, p = 0.007) and PWV (r = 0.30, p = 0.03). Conclusions: The most important finding of this study is the strict and independent association between BP parameters, especially cSBP at baseline, and indexes of vascular damage at follow-up, even after multiple adjustments. Indeed, cDC at baseline is also associated with increased cDBP at follow-up. Our data support the hypothesis of a vicious cycle between BP and arterial stiffness, which may worsen both parameters over time.

SUBCLINICAL TARGET ORGAN DAMAGE IN PATIENTS WITH RESTLESS LEGS SYNDROME. A PRELIMINARY OBSERVATIONAL STUDY

Objective: Restless legs syndrome (RLLS) is a frequent and often underdiagnosed type of insomnia. Since other types of insomnia present a strict association with cardiovascular disease, our study aimed to evaluate the possible association of RLLS with some early markers of cardiovascular organ damage. Design and method: Patients with other types of insomnia and those with central nervous system diseases were excluded. Twenty-eight patients (mean age: 63,1±13,9 years; 28.6% males) participated in the study. Fourteen subjects with diagnosis of RLLS according to International diagnostic criteria (Allen RP et al. 2014 Sleep Medicine) (disease mean duration 9.6±9.2 years; International Restless Legs Scale score 22.9±5.2) were matched for sex, age and major cardiovascular risk factors with 14 controls. In both groups, peripheral Blood Pressure (pBP), central Blood Pressure (cBP) and Pulse Wave Velocity (PWV) were assessed by the SphygmoCorXCEL. Carotid Intima-Media Thickness (cIMT) and the carotid Distensibility Coefficient (cCD) were assessed by ultrasound connected with an image acquisition and analysis system (CardiovascularSuite). Results: We did not find any significant difference between the two groups in the indices of subclinical vascular damage. In the RLS group, blood pressure measurements significantly correlated with all the markers of subclinical vascular damage, particularly pSBP was associated with PWV (r = 0.53, p = 0.009), Pulse Pressure correlated with cIMT (r = 0.58, p = 0.01) and cDBP was inversely related to cDC (r = -0.53; p = 0.03). Conclusions: Our data did not show significant differences in markers of early vascular damage between patients with RLS and matched controls. As expected, the markers of vascular damage were associated with BP. The present is only a pilot study; further studies with a larger sample size are required to confirm our findings.

Retinal Microvascular Diameters are Associated with Diabetic Kidney Disease in Patients with Type 2 Diabetes Mellitus.

To investigate the association between retinal microvascular diameters and diabetic kidney disease (DKD) in patients with type 2 diabetes mellitus (T2DM). A total of 690 patients with T2DM were included in this retrospective study. Patients were divided into DKD and non-DKD groups according to urine microalbumin/creatinine ratio and estimated glomerular filtration rate. Retinal microvascular diameters were measured by the automated retinal image analysis system. Multivariate logistic regression analysis and restricted cubic splines were used to assess the relationships between the retinal microvascular diameters and DKD in patients with T2DM. Multivariate logistic regression showed that widened diameters of retinal venules and narrowed diameters of retinal arterioles were associated with DKD after adjusting for potential confounding variables. There was a significant linear trend between the diameters of superior temporal retinal venula (P for trend < 0.001, P for non-linearity = 0.080), inferior temporal retinal venula (P for trend < 0.001, P for non-linearity = 0.111) and central retinal venular equivalent (CRVE) (P for trend < 0.001, P for non-linearity = 0.392) and risk of DKD in patients with T2DM. The restricted cubic splines showed that narrowed retinal arteriolar diameters, superior and inferior nasal retinal venulas were associated with the risk of DKD in a non-linear fashion (all P for non-linearity < 0.001). Wider retinal venular diameters and narrower retinal arteriolar diameters were associated with an increased risk of DKD in patients with T2DM. Widened retinal venular diameters, especially CRVE, superior and inferior temporal retinal venula, were positively associated with an increased risk of DKD in a linear fashion. In contrast, narrowed retinal arteriolar diameters were associated with the risk of DKD in a non-linear fashion.

A new AI assisted medical molecular image diagnostic model

In recent times, ML algorithms that plays a significant role right from drug discovery to clinical decision making. The recent advances in DL technologies contribute towards improved performance for carrying out computer aided medical image analysis and disease diagnosis. The key benefit of AI in processing of medical big data offers spectacular insights into the hierarchal relationships that exist among data which can be algorithmically explored thus replacing the tedious manual processes to extract and localize specific areas of interests in medical images thus considerably changing the way medicine has been practiced so far. In bio medical related clinical applications, there is a constant demand pertaining the research and development with respect to deploying AI as a mainstream tool to perform several medical imaging activities like analysis, diagnosis, segmentation as well as classification. The increased usage of electronic health records and medical images being its integral component the need for appropriate and efficient AI assisted medical image analysis system that takes care of accurate and automated decision making could be of great help to radiologists and medical practitioners. Molecular image analysis is a dynamic field that makes use of ML and DL algorithms that utilizes labeled and structured information which also proves to be helpful to the patients as they serve as an initial interface before further diagnosis and treatments. Thus our research aims to offer a novel and efficient AI based medical analysis system that can assist clinical practitioners to focus on enhancing the disease diagnosis through DL based medical image analysis and decision making. In addition, we also address specific challenges related to disease diagnosis and propose novel GAN model for improved diagnosis and implementation. Our proposed technique can also be generalized to generate synthetic data for further issues related to molecular image analysis in the field of medicine and help towards building a better disease diagnosis model.

Skin moisturizing and anti-acne effect of acne moisturizer in healthy adult subjects with mild-to-moderate acne-an open label, single arm clinical study

Objectives: Acne vulgaris is an extremely common condition that occurs mostly during adolescence. It often causes non-inflammatory lesions (open and closed comedones), inflammatory lesions (papules, pustules, and nodules), and varying degrees of scarring that leads to a negative impact on quality of life. This open label, single arm, and clinical study was conducted to evaluate the effectiveness of acne moisturizer on skin moisturization in adult subjects with mild-to-moderate acne. Material and Methods: An open label, single-arm, and clinical study was conducted on 36 enrolled healthy male and females aged 18–35 years having acne-prone skin and with mild-to-moderate acne. Acne moisturizer (Venusia Acne) was applied twice daily and safety and effectiveness of test product was assessed before application to post-application on day 15, day 30, day 45, and day 60 at the end of study. All enrolled subjects had undergone clinical evaluation by dermatologist, instruments evaluation, and subjective evaluation on various parameters. Results: Thirty-one (n = 31) subjects completed all study visits. After using Acne Moisturizer for 60 days, statistically significant (P &lt; 0.0001) improvement from baseline was observed in skin moisturization. The product was also found effective in protecting skin barrier function (P &lt; 0.0001) by reducing trans-epidermal water loss. Assessment of prevention of appearance of new acne and acne reduction using Investigator’s Global Assessment (IGA) Scale for acne vulgaris by dermatologist shows significant improvement from baseline to each study visit and at the end of study. Statistically significant reduction (P &lt; 0.0001) in acne severity assessed by IGA was also observed. Moreover, significant reduction (P &lt; 0.0001) in skin sebum level, skin blemishes, skin redness, skin pigmentation, and facial pores was also observed as assessed by instruments and 3D image analysis system. Assessment of the subject satisfaction questionnaire and subject response index was also found in parity of the outcome. Furthermore, no adverse event was recorded during the study conduction. Conclusion: The results of this clinical study suggest that acne moisturizer helps in significantly improving skin moisturization when applied for 15 days onward. It is also safe and effective in aiding reduction of skin sebum level, skin blemishes, skin redness, skin pigmentation, and facial pores; it can be effectively used as an adjunct for management of mild-to-moderate acne prevention and acne reduction, while providing sufficient moisturization to the skin.

Vision Transformers for Lung Segmentation on CXR Images.

Accurate segmentation of the lungs in CXR images is the basis for an automated CXR image analysis system. It helps radiologists in detecting lung areas, subtle signs of disease and improving the diagnosis process for patients. However, precise semantic segmentation of lungs is considered a challenging case due to the presence of the edge rib cage, wide variation of lung shape, and lungs affected by diseases. In this paper, we address the problem of lung segmentation in healthy and unhealthy CXR images. Five models were developed and used in detecting and segmenting lung regions. Two loss functions and three benchmark datasets were employed to evaluate these models. Experimental results showed that the proposed models were able to extract salient global and local features from the input CXR images. The best performing model achieved an F1 score of 97.47%, outperforming recent published models. They proved their ability to separate lung regions from the rib cage and clavicle edges and segment varying lung shape depending on age and gender, as well as challenging cases of lungs affected by anomalies such as tuberculosis and the presence of nodules.

Deceptive Tricks in Artificial Intelligence: Adversarial Attacks in Ophthalmology.

The artificial intelligence (AI) systems used for diagnosing ophthalmic diseases have significantly progressed in recent years. The diagnosis of difficult eye conditions, such as cataracts, diabetic retinopathy, age-related macular degeneration, glaucoma, and retinopathy of prematurity, has become significantly less complicated as a result of the development of AI algorithms, which are currently on par with ophthalmologists in terms of their level of effectiveness. However, in the context of building AI systems for medical applications such as identifying eye diseases, addressing the challenges of safety and trustworthiness is paramount, including the emerging threat of adversarial attacks. Research has increasingly focused on understanding and mitigating these attacks, with numerous articles discussing this topic in recent years. As a starting point for our discussion, we used the paper by Ma et al. "Understanding Adversarial Attacks on Deep Learning Based Medical Image Analysis Systems". A literature review was performed for this study, which included a thorough search of open-access research papers using online sources (PubMed and Google). The research provides examples of unique attack strategies for medical images. Unfortunately, unique algorithms for attacks on the various ophthalmic image types have yet to be developed. It is a task that needs to be performed. As a result, it is necessary to build algorithms that validate the computation and explain the findings of artificial intelligence models. In this article, we focus on adversarial attacks, one of the most well-known attack methods, which provide evidence (i.e., adversarial examples) of the lack of resilience of decision models that do not include provable guarantees. Adversarial attacks have the potential to provide inaccurate findings in deep learning systems and can have catastrophic effects in the healthcare industry, such as healthcare financing fraud and wrong diagnosis.

Application of an AI image analysis and classification approach to characterise dissolution and precipitation events in the flow through apparatus

Imaging and artificial intelligence (AI) approaches have been used with increasing frequency in pharmaceutical industry in recent years. Characterisation of processes such as drug dissolution and precipitation is vital in quality control testing and drug manufacture. To support existing techniques like in vitro dissolution testing, novel process analytical technologies (PATs) can give an insight into these processes. The aim of this study was to create and explore the potential of an automated image classification model based on image analysis to identify events (dissolution and precipitation) occurring in the flow-through apparatus (FTA) test cell, and the ability to characterise a dissolution process over time. Several precipitation conditions were tested in a USP 4 FTA test cell with images recorded during early (plume formation) and late (particulate re-formation) stages of precipitation. An available MATLAB code was used as a base to develop and validate an anomaly classification model able to detect different events occurring during the precipitation process in the dissolution cell. Two variants of the model were tested on images from a dissolution test in the FTA, with a view to application of the image analysis system to quantitative characterization of the dissolution process over time. It was found that the classification model is highly accurate (>90%) in detecting events occurring in the FTA test cell. The model showed potential to be used to characterise the stages of dissolution and precipitation processes, and as a proof of concept demonstrates potential for deep machine learning image analysis to be applied to kinetics of other pharmaceutical processes.

A fully-automatic semi-supervised deep learning model for difficult airway assessment

BackgroundDifficult airway conditions represent a substantial challenge for clinicians. Predicting such conditions is essential for subsequent treatment planning, but the reported diagnostic accuracies are still quite low. To overcome these challenges, we developed a rapid, non-invasive, cost-effective, and highly-accurate deep-learning approach to identify difficult airway conditions through photographic image analysis. MethodsFor each of 1000 patients scheduled for elective surgery under general anesthesia, images were captured from 9 specific and different viewpoints. The collected image set was divided into training and testing subsets in the ratio of 8:2. We used a semi-supervised deep-learning method to train and test an AI model for difficult airway prediction. ResultsWe trained our semi-supervised deep-learning model using only 30% of the labeled training samples (with the remaining 70% used without labels). We evaluated the model performance using metrics of accuracy, sensitivity, specificity, F1-score, and the area under the ROC curve (AUC). The numerical values of these four metrics were found to be 90.00%, 89.58%, 90.13%, 81.13%, and 0.9435, respectively. For a fully-supervised learning scheme (with 100% of the labeled training samples used for model training), the corresponding values were 90.50%, 91.67%, 90.13%, 82.25%, and 0.9457, respectively. When three professional anesthesiologists conducted comprehensive evaluation, the corresponding results were 91.00%, 91.67%, 90.79%, 83.26%, and 0.9497, respectively. It can be seen that the semi-supervised deep learning model trained by us with only 30% labeled samples can achieve a comparable effect with the fully supervised learning model, but the sample labeling cost is smaller. Our method can achieve a good balance between performance and cost. At the same time, the results of the semi-supervised model trained with only 30% labeled samples were very close to the performance of human experts. ConclusionsTo the best of our knowledge, our study is the first one to apply a semi-supervised deep-learning method in order to identify the difficulties of both mask ventilation and intubation. Our AI-based image analysis system can be used as an effective tool to identify patients with difficult airway conditions. Clinical trial registrationChiCTR2100049879 (URL: http://www.chictr.org.cn).

Efficient segmentation algorithm for complex cellular image analysis system

AbstractCellular image analysis system is a complex system that plays a critical role in disease diagnosis and pharmaceutical research. The analysis of image data is one of the most critical aspects of the system. However, there are differences in the distribution of cellular images, including cell morphology, cell density etc. This often requires careful algorithm customization, strict parameter tuning, or even inefficient manual processing, leading to low levels of automation. In this work, an efficient end‐to‐end cell segmentation algorithm, ECS‐Net, is proposed that can handle detection, segmentation, and counting tasks simultaneously. Two modules, proposal focus module (PFM) and enhance mask feature head (EMFH), are introduced to improve the segmentation accuracy. The proposed algorithm achieves better detection and segmentation accuracy with fewer parameters and computational cost, thus improving cellular image analysis systems. Furthermore, considering the medical IoT scenario, the scaled‐down model with only 5.8M parameters has only a small decrease in accuracy which has significant application value.

GRAIN SIZE OF ELECTROCORUNDUM SINTERED FROM DISPERSED WASTE ALUMINUM GRADE AD0E

To develop technologies for obtaining electrocorundum by spark plasma sintering of aluminum waste dispersed by electroerosion and to evaluate the effectiveness of their practical application, comprehensive the oretical and experimental studies are required. The purpose of this work was the grain size of electrocorundum sintered from dispersed aluminum waste of the AD0E brand. Electrodispersion of aluminum waste of the AD0E brand was carried out in distilled water at the original installation with a capacity of 65.5 UF discharge capacitors, a voltage at the electrodes of 200 V, a pulse repetition frequency of 200 Hz. As a result of exposure to short-term electrical discharges, particles of electroerosive powder of various sizes were formed in the water. Further, the fusion of the electroerosion charge was carried out in the spark plasma fusion system SPS 25-10 "Thermal Technology" (USA) at a pressure of 30 MPa, a temperature of 560 °C and a holding time of 3 minutes. Analysis of the microstructure of sintered electrocorundum showed that it has a fine-grained structure, without inclusions, uniform phase distribution and the absence of significant pores, cracks and discontinuities. The grain size of the studied alloys, determined using the SIMAGIS Photolab automated image analysis system and the OLYMPUS GX51 optical inverted microscope, was about 0.45 microns. The small grain size of the resulting electrocorundum is associated with the high dispersion of the initial electroerosion charge and the effect of "grain growth suppression" during spark plasma fusion due to the short working cycle time, high pressure and uniform heat distribution over the sample when exposed to pulsed electric current and the so-called "spark discharge plasma effect". The conducted studies will allow to perform resource conservation and import substitution in the production of electrocorundum from aluminum waste of the AD0E brand.