Advances in artificial intelligence for meibomian gland evaluation: A comprehensive review.

Meibomian gland dysfunction (MGD) is increasingly recognized as a critical contributor to evaporative dry eye, significantly impacting visual quality. With a global prevalence estimated at 35.8 %, it presents substantial challenges for clinicians. Conventional manual evaluation techniques for MGD face limitations characterized by inefficiencies, high subjectivity, limited big data processing capabilities, and a dearth of quantitative analytical tools. With rapidly advancing artificial intelligence (AI) techniques revolutionizing ophthalmology, studies are now leveraging sophisticated AI methodologies--including computer vision, unsupervised learning, and supervised learning--to facilitate comprehensive analyses of meibomian gland (MG) evaluations. These evaluations employ various techniques, including slit lamp examination, infrared imaging, confocal microscopy, and optical coherence tomography. This paradigm shift promises enhanced accuracy and consistency in disease evaluation and severity classification. While AI has achieved preliminary strides in meibomian gland evaluation, ongoing advancements in system development and clinical validation are imperative. We review the evolution of MG evaluation, juxtapose AI-driven methods with traditional approaches, elucidate the specific roles of diverse AI technologies, and explore their practical applications using various evaluation techniques. Moreover, we delve into critical considerations for the clinical deployment of AI technologies and envisages future prospects, providing novel insights into MG evaluation and fostering technological and clinical progress in this arena.

Meibomian glands segmentation in infrared images with limited annotation.

AIM: To investigate a pioneering framework for the segmentation of meibomian glands (MGs), using limited annotations to reduce the workload on ophthalmologists and enhance the efficiency of clinical diagnosis. METHODS: Totally 203 infrared meibomian images from 138 patients with dry eye disease, accompanied by corresponding annotations, were gathered for the study. A rectified scribble-supervised gland segmentation (RSSGS) model, incorporating temporal ensemble prediction, uncertainty estimation, and a transformation equivariance constraint, was introduced to address constraints imposed by limited supervision information inherent in scribble annotations. The viability and efficacy of the proposed model were assessed based on accuracy, intersection over union (IoU), and dice coefficient. RESULTS: Using manual labels as the gold standard, RSSGS demonstrated outcomes with an accuracy of 93.54%, a dice coefficient of 78.02%, and an IoU of 64.18%. Notably, these performance metrics exceed the current weakly supervised state-of-the-art methods by 0.76%, 2.06%, and 2.69%, respectively. Furthermore, despite achieving a substantial 80% reduction in annotation costs, it only lags behind fully annotated methods by 0.72%, 1.51%, and 2.04%. CONCLUSION: An innovative automatic segmentation model is developed for MGs in infrared eyelid images, using scribble annotation for training. This model maintains an exceptionally high level of segmentation accuracy while substantially reducing training costs. It holds substantial utility for calculating clinical parameters, thereby greatly enhancing the diagnostic efficiency of ophthalmologists in evaluating meibomian gland dysfunction.

Effects of orthokeratology lenses on tear film and tarsal glands and control of unilateral myopia in children.

Introduction: To investigate the effects of an orthokeratology lens on the tear film and tarsal glands and myopia control in children with unilateral myopia using an intelligent analysis model. Methods: We retrospectively reviewed the medical records from November 2020 to November 2022 of 68 pediatric patients with unilateral myopia in Fujian Provincial Hospital who had been wearing an orthokeratology lens for more than 1 year. The 68 myopic eyes were included in the treatment group, while the 68 healthy, untreated contralateral eyes were included in the control group. Tear film break-up times (TBUTs) were compared between the two groups at various intervals, and an intelligent analysis model was used to compare the deformation coefficients of 10 meibomian glands in the central area and the different positions of the glands in the two groups after 12 months of treatment. Changes in axial length and equivalent spherical power were also compared between the groups before and after 12 months of treatment. Results: In the treatment group, TBUTs differed significantly between 1 and 12 months after treatment, although no significant differences from baseline were observed at 3 or 6 months. No significant differences in TBUTs were observed at any time point in the control group. After 12 months of treatment, significant between-group differences were observed for glands 2, 3, 4, 5, 6, 7, 8, and 10 (numbered from the temporal to nasal regions). The treatment group also exhibited significant differences in deformation coefficients at different detection positions in the central region, with glands 5 and 6 exhibiting the highest deformation coefficients. Increases in axial length and equivalent spherical power were significantly greater in the control group than in the treatment group after 12 months of treatment. Discussion: Wearing orthokeratology lenses at night can effectively control myopia progression in children with unilateral myopia. However, long-term use of these lenses may lead to meibomian gland deformation and impact tear film function, and the extent of deformation may vary at different positions in the central region.

Metabolomics and Network Analyses Reveal Phenylalanine and Tyrosine as Signatures of Anthracycline-Induced Hepatotoxicity.

The chemotherapy drug doxorubicin (DOX) is an anthracycline with over 30% incidence of liver injury in breast cancer patients, yet the mechanism of its hepatotoxicity remains unclear. To identify potential biomarkers for anthracycline-induced hepatotoxicity (AIH), we generated clinically-relevant mouse and rat models administered low-dose, long-term DOX. These models exhibited significant liver damage but no decline in cardiac function. Through untargeted metabolic profiling of the liver, we identified 27 differential metabolites in a mouse model and 28 in a rat model. We then constructed a metabolite-metabolite network for each animal model and computationally identified several potential metabolic markers, with particular emphasis on aromatic amino acids, including phenylalanine, tyrosine, and tryptophan. We further performed targeted metabolomics analysis on DOX-treated 4T1 breast cancer mice for external validation. We found significant (p < 0.001) reductions in hepatic levels of phenylalanine and tyrosine (but not tryptophan) following DOX treatment, which were strongly correlated with serum aminotransferases (ALT and AST) levels. In summary, the results of our study present compelling evidence supporting the use of phenylalanine and tyrosine as metabolic signatures of AIH.

Quantification of 10 B vitamins in mouse colon by LC-MS/MS: Application on breast cancer mice treated with doxorubicin.

B vitamins play important roles in various physiological processes, including cell metabolism and DNA synthesis. The intestine is critical for the absorption and utilization of B vitamins, but few analytical methods for detecting intestinal B vitamins are currently available. In this study, we developed a novel liquid chromatography-tandem mass spectrometry (LC-MS/MS) method for the simultaneous quantification of 10 B vitamins in mouse colon tissue, including thiamin (B1), riboflavin (B2), nicotinic acid (B3), niacinamide (B3-AM), pantothenic acid (B5), pyridoxine (B6), pyridoxal 5'-phosphate (B6-5P), biotin (B7), folic acid (B9), and cyanocobalamin (B12). The method was thoroughly validated following the U.S. Food and Drug Administration (FDA) guidelines and yielded good results in terms of linearity (r2 > 0.9928), lower limit of quantification (40-600 ng/g), accuracy (88.9-119.80 %) and precision (relative standard deviation ≤ 19.71 %), recovery (87.95-113.79 %), matrix effect (91.26-113.78 %), and stability (85.65-114.05 %). Furthermore, we applied our method to profile B vitamins in the colons of mice with breast cancer after doxorubicin chemotherapy treatment, which revealed that the doxorubicin treatment led to significant colon damage and accumulation of several B vitamins including B1, B2 and B5. We also confirmed the capability of this method for quantifying B vitamins in other intestinal tissues like the ileum, jejunum, and duodenum. The newly developed method is simple, specific, and useful for targeted profiling of B vitamins in mouse colon, with a potential for future studies on the role of these micronutrients in healthy and diseased states.

Identification of region-specific amino acid signatures for doxorubicin-induced chemo brain.

Doxorubicin (DOX) is a cornerstone of chemotherapy for solid tumors and leukemias. DOX-induced cognitive impairment, termed chemo brain, has been reported in cancer survivors, whereas its mechanism remains poorly understood. Here we initially evaluated the cognitive impairments of mice treated with clinically relevant, long-term, low-dosage of DOX. Using HILIC-MS/MS-based targeted metabolomics, we presented the changes of 21 amino acids across six anatomical brain regions of mice with DOX-induced chemo brain. By mapping the altered amino acids to the human metabolic network, we constructed an amino acid-based network module for each brain region. We identified phenylalanine, tyrosine, methionine, and γ-aminobutyric acid as putative signatures of three regions (hippocampus, prefrontal cortex, and neocortex) highly associated with cognition. Relying on the reported mouse brain metabolome atlas, we found that DOX might perturb the amino acid homeostasis in multiple brain regions, similar to the changes in the aging brain. Correlation analysis suggested the possible indirect neurotoxicity of DOX that altered the brain levels of phenylalanine, tyrosine, and methionine by causing metabolic disorders in the liver and kidney. In summary, we revealed the region-specific amino acid signatures as actionable targets for DOX-induced chemo brain, which might provide safer treatment and improve the quality of life among cancer survivors.

Plasma Pharmacokinetics and Tissue Distribution of Doxorubicin in Rats following Treatment with Astragali Radix.

Doxorubicin (DOX) is an essential component in chemotherapy, and Astragali Radix (AR) is a widely used tonic herbal medicine. The combination of DOX and AR offers widespread, well-documented advantages in treating cancer, e.g., reducing the risk of adverse effects. This study mainly aims to uncover the impact of AR on DOX disposition in vivo. Rats received a single intravenous dose of 5 mg/kg DOX following a single-dose co-treatment or multiple-dose pre-treatment of AR (10 g/kg × 1 or × 10). The concentrations of DOX in rat plasma and six tissues, including heart, liver, lung, kidney, spleen, and skeletal muscle, were determined by a fully validated LC-MS/MS method. A network-based approach was further employed to quantify the relationships between enzymes that metabolize and transport DOX and the targets of nine representative AR components in the human protein−protein interactome. We found that short-term (≤10 d) AR administration was ineffective in changing the plasma pharmacokinetics of DOX in terms of the area under the concentration−time curve (AUC, 1303.35 ± 271.74 µg/L*h versus 1208.74 ± 145.35 µg/L*h, p > 0.46), peak concentrations (Cmax, 1351.21 ± 364.86 µg/L versus 1411.01 ± 368.38 µg/L, p > 0.78), and half-life (t1/2, 31.79 ± 5.12 h versus 32.05 ± 6.95 h, p > 0.94), etc. Compared to the isotype control group, DOX concentrations in six tissues slightly decreased under AR pre-administration but only showed statistical significance (p < 0.05) in the liver. Using network analysis, we showed that five of the nine representative AR components were not localized to the vicinity of the DOX disposition-associated module. These findings suggest that AR may mitigate DOX-induced toxicity by affecting drug targets rather than drug disposition.

Network-driven targeted analysis reveals that Astragali Radix alleviates doxorubicin-induced cardiotoxicity by maintaining fatty acid homeostasis.

ETHNOPHARMACOLOGICAL RELEVANCE: Astragali Radix (AR) is a popular traditional Chinese medicine that has been used for more than 2000 years. It is a well-known tonic for weak people with chronic diseases, such as heart failure and cerebral ischemia. Previous studies have reported that AR could support the "weak heart" of cancer patients who suffered from doxorubicin (DOX)-induced cardiotoxicity (DIC). However, the underlying mechanism remains unclear. AIM OF THE STUDY: This study aimed to uncover the critical pathways and molecular determinants for AR against DIC by fully characterizing the network-based relationship. MATERIALS AND METHODS: We integrated ultra-high-performance liquid chromatography-high-resolution mass spectrometry (UHPLC-HRMS) profiling, database and literature searching, and the human protein-protein interactome to discover the specific network module associated with AR against DIC. To validate the network-based findings, a low-dose, long-term DIC mouse model and rat cardiomyoblast H9c2 cells were employed. The levels of potential key metabolites and proteins in hearts and cells were quantified by the LC-MS/MS targeted analysis and western blotting, respectively. RESULTS: We constructed one of the most comprehensive AR component-target network described to date, which included 730 interactions connecting 64 unique components and 359 unique targets. Relying on the network-based evaluation, we identified fatty acid metabolism as a putative critical pathway and peroxisome proliferator-activated receptors (PPARα and PPARγ) as potential molecular determinants. We then confirmed that DOX caused the accumulation of fatty acids in the mouse failing heart, while AR promoted fatty acid metabolism and preserved heart function. By inhibiting PPARγ in H9c2 cells, we further found that AR could alleviate DIC by activating PPARγ to maintain fatty acid homeostasis. CONCLUSIONS: Our findings imply that AR is a promising drug candidate that treats DIC by maintaining fatty acid homeostasis. More importantly, the network-based method developed here could facilitate the mechanism discovery of AR therapy and help catalyze innovation in its clinical application.

Bioinformatic identification of potential biomarkers and therapeutic targets in carotid atherosclerosis and vascular dementia.

Background: Vascular disease is the second most common cause of dementia. The prevalence of vascular dementia (VaD) has increased over the past decade. However, there are no licensed treatments for this disease. Carotid atherosclerosis (CAS) is highly prevalent and is the main cause of ischemic stroke and VaD. We studied co-expressed genes to understand the relationships between CAS and VaD and further reveal the potential biomarkers and therapeutic targets of CAS and VaD. Methods: CAS and VaD differentially expressed genes (DEGs) were identified through bioinformatic analysis Gene Expression Omnibus (GEO) datasets GSE43292 and GSE122063, respectively. Furthermore, a variety of target prediction methods and network analysis approaches were used to assess the protein-protein interaction (PPI) networks, the Gene Ontology (GO) terms, and the pathway enrichment for DEGs, and the top 7 hub genes, coupled with corresponding predicted miRNAs involved in CAS and VaD, were assessed as well. Result: A total of 60 upregulated DEGs and 159 downregulated DEGs were identified, of which the top 7 hub genes with a high degree of connectivity were selected. Overexpression of these hub genes was associated with CAS and VaD. Finally, the top 7 hub genes were coupled with corresponding predicted miRNAs. hsa-miR-567 and hsa-miR-4652-5p may be significantly associated with CAS and VaD.

Metabolic network-based identification of plasma markers for non-small cell lung cancer.

Metabolic markers, offering sensitive information on biological dysfunction, play important roles in diagnosing and treating cancers. However, the discovery of effective markers is limited by the lack of well-established metabolite selection approaches. Here, we propose a network-based strategy to uncover the metabolic markers with potential clinical availability for non-small cell lung cancer (NSCLC). First, an integrated mass spectrometry-based untargeted metabolomics was used to profile the plasma samples from 43 NSCLC patients and 43 healthy controls. We found that a series of 39 metabolites were altered significantly. Relying on the human metabolic network assembled from Kyoto Encyclopedia of Genes and Genomes (KEGG) database, we mapped these differential metabolites to the network and constructed an NSCLC-related disease module containing 23 putative metabolic markers. By measuring the PageRank centrality of molecules in this module, we computationally evaluated the network-based importance of the 23 metabolites and demonstrated that the metabolism pathways of aromatic amino acids and long-chain fatty acids provided potential molecular targets of NSCLC (i.e., IL4l1 and ACOT2). Combining network-based ranking and support-vector machine modeling, we further found a panel of eight metabolites (i.e., pyruvate, tryptophan, and palmitic acid) that showed a high capability to differentiate patients from controls (accuracy > 97.7%). In summary, we present a meaningful network method for metabolic marker discovery and have identified eight strong candidate metabolites for NSCLC diagnosis.

Attenuation of doxorubicin-induced oxidative damage in rat brain by regulating amino acid homeostasis with Astragali Radix.

The nervous system disorders caused by doxorubicin (DOX) are among the severe adverse effects that dramatically reduce the quality of life of cancer survivors. Astragali Radix (AR), a popular herbal drug and dietary supplement, is believed to help treat brain diseases by reducing oxidative stress and maintaining metabolic homeostasis. Here we show the protective effects of AR against DOX-induced oxidative damage in rat brain via regulating amino acid homeostasis. By constructing a clinically relevant low-dose DOX-induced toxicity rat model, we first performed an untargeted metabolomics analysis to discover specific metabolic features in the brain after DOX treatment and AR co-treatment. It was found that the amino acid (AA) metabolism pathways altered most significantly. To accurately characterize the brain AA profile, we established a sensitive, fast, and reproducible hydrophilic interaction chromatography-tandem mass spectrometry method for the simultaneous quantification of 22 AAs. The targeted analysis further confirmed the changes of AAs between different groups of rat brain. Specifically, the levels of six AAs, including glutamate, glycine, serine, alanine, citrulline, and ornithine, correlated (Pearson |r| > 0.47, p < 0.05) with the brain oxidative damage that was caused by DOX and rescued by AR. These findings present that AAs are among the regulatory targets of DOX-induced brain toxicity, and AR is a promising therapeutic agent for it.

Reconstruction Algorithm-Based Ultrasonic and Spiral CT Images in Evaluating the Effects of Dexmedetomidine Anesthesia for Acute Abdomen.

OBJECTIVE: The study focused on the application value of iteration reconstruction algorithm-based ultrasound and spiral computed tomography (CT) examinations, and the safety of dexmedetomidine anesthesia in acute abdominal surgery. METHODS: 80 cases having the acute abdomen surgery were selected as the research subjects. They were divided into group A (40 cases) and group B (40 cases) according to the anesthetic drugs used in the later period. The experimental group was injected with propofol, remifentanil, and atracurium combined with dexmedetomidine; the control group was injected with propofol, remifentanil, and atracurium only. After the operation, the patient was for observed for the pain, agitation, adverse reactions, heart rate (HR), and blood pressure. All patients received ultrasound and spiral CT examinations, and based on the characteristics of the back-projection algorithm, an accelerated algorithm was established and used to process the image, and according to which, the patient's condition and curative effects were evaluated. RESULTS: After image reconstruction, the ultrasound and spiral CT images were clearer with less noise and more prominent lesions than before reconstruction. Before image reconstruction, the accuracy rates of ultrasound and spiral CT in diagnosing acute abdomen were 92.3% and 91.1%, respectively. After reconstruction, the corresponding numbers were 96.3% and 98.1%, respectively. After reconstruction, the accuracy of the two methods in diagnosing acute abdomen was significantly improved compared with that before reconstruction, and the difference was statistically significant (P < 0.05). The Ramsay score of the experimental group was significantly higher than that of the control group at each time period, P < 0.05; the agitation score and visual analogue scale (VAS) score of the experimental group were significantly lower than the control group at each time period after waking up, P < 0.05. CONCLUSION: Reconstruction algorithm-based ultrasound and spiral CT images have high application value in the diagnosis of patients with acute abdomen, and dexmedetomidine has good safety in anesthesia surgery.

Phage-display library biopanning and bioinformatic analysis yielded a high-affinity peptide to inflamed vascular endothelium both in vitro and in vivo.

Vascular inflammation is considered the primary pathological condition occurring in many chronic diseases. To detect the inflamed endothelium via imaging analysis or guide the drug to target lesions is therefore important for early diagnosis and treatment of vascular inflammatory diseases. In this study, we obtained a novel peptide NTTTH through high throughout biopanning and bioinformatic analysis. In vitro studies indicated that NTTTH homologs could especially target inflamed vascular endothelial cells, as imaging quantitative analysis indicated that the mean of integrated optical density (MIOD) and mean of stained area (MSA) were significantly higher versus control (P<0.05). In vivo studies showed that, after intravenous injection of enhanced green fluorescent protein (EGFP)-labeled NTTTH homologs into the lipopolysaccharide (LPS)-inflamed mice for 30min, NTTTH homologs were distributed in highly vascularized and inflamed organs like liver and kidney. As a control, little fluorescence could be detected in mice injected with EGFP alone. Cryosection showed that NTTTH homologs especially targeted inflamed vasculatures but not normal ones. We did not detect fluorescence signal in either normal or inflamed mice which were injected with EGFP alone. The results suggested the role of NTTTH homologs in guiding the targeted binding of EGFP to inflamed vasculature and the potential usage for imaging detection and drug delivery.