Explainable ensemble learning method for OCT detection with transfer learning.

The accuracy and interpretability of artificial intelligence (AI) are crucial for the advancement of optical coherence tomography (OCT) image detection, as it can greatly reduce the manual labor required by clinicians. By prioritizing these aspects during development and application, we can make significant progress towards streamlining the clinical workflow. In this paper, we propose an explainable ensemble approach that utilizes transfer learning to detect fundus lesion diseases through OCT imaging. Our study utilized a publicly available OCT dataset consisting of normal subjects, patients with dry age-related macular degeneration (AMD), and patients with diabetic macular edema (DME), each with 15 samples. The impact of pre-trained weights on the performance of individual networks was first compared, and then these networks were ensemble using majority soft polling. Finally, the features learned by the networks were visualized using Grad-CAM and CAM. The use of pre-trained ImageNet weights improved the performance from 68.17% to 92.89%. The ensemble model consisting of the three CNN models with pre-trained parameters loaded performed best, correctly distinguishing between AMD patients, DME patients and normal subjects 100% of the time. Visualization results showed that Grad-CAM could display the lesion area more accurately. It is demonstrated that the proposed approach could have good performance of both accuracy and interpretability in retinal OCT image detection.

A Novel N-Gram-Based Image Classification Model and Its Applications in Diagnosing Thyroid Nodule and Retinal OCT Images.

Imbalanced classes and dimensional disasters are critical challenges in medical image classification. As a classical machine learning model, the n-gram model has shown excellent performance in addressing this issue in text classification. In this study, we proposed an algorithm to classify medical images by extracting their n-gram semantic features. This algorithm first converts an image classification problem to a text classification problem by building an n-gram corpus for an image. After that, the algorithm was based on the n-gram model to classify images. The algorithm was evaluated by two independent public datasets. The first experiment is to diagnose benign and malignant thyroid nodules. The best area under the curve (AUC) is 0.989. The second experiment is to diagnose the type of fundus lesion. The best result is that it correctly identified 86.667% of patients with dry age-related macular degeneration (AMD), 93.333% of patients with diabetic macular edema (DME), and 93.333% of normal individuals.

Amyloid-Related Imaging Abnormalities in 2 Phase 3 Studies Evaluating Aducanumab in Patients With Early Alzheimer Disease.

Importance: The EMERGE and ENGAGE phase 3 randomized clinical trials of aducanumab provide a robust data set to characterize amyloid-related imaging abnormalities (ARIA) that occur with treatment with aducanumab, an amyloid-ß (Aß)-targeting monoclonal antibody, in patients with mild cognitive impairment due to Alzheimer disease or mild Alzheimer disease dementia. Objective: To describe the radiographic and clinical characteristics of ARIA that occurred in EMERGE and ENGAGE. Design, Setting, and Participants: Secondary analysis of data from the EMERGE and ENGAGE trials, which were 2 double-blind, placebo-controlled, parallel-group, phase 3 randomized clinical trials that compared low-dose and high-dose aducanumab treatment with placebo among participants at 348 sites across 20 countries. Enrollment occurred from August 2015 to July 2018, and the trials were terminated early (March 21, 2019) based on a futility analysis. The combined studies consisted of a total of 3285 participants with Alzheimer disease who received 1 or more doses of placebo (n = 1087) or aducanumab (n = 2198; 2752 total person-years of exposure) during the placebo-controlled period. Primary data analyses were performed from November 2019 to July 2020, with additional analyses performed through July 2021. Interventions: Participants were randomly assigned 1:1:1 to high-dose or low-dose intravenous aducanumab or placebo once every 4 weeks. Dose titration was used as a risk-minimization strategy. Main Outcomes and Measures: Brain magnetic resonance imaging was used to monitor patients for ARIA; associated symptoms were reported as adverse events. Results: Of 3285 included participants, the mean (SD) age was 70.4 (7.45) years; 1706 participants (52%) were female, 2661 (81%) had mild cognitive impairment due to Alzheimer disease, and 1777 (54%) used symptomatic medications for Alzheimer disease. A total of 764 participants from EMERGE and 709 participants from ENGAGE were categorized as withdrawn before study completion, most often owing to early termination of the study by the sponsor. Unless otherwise specified, all results represent analyses from the 10-mg/kg group. During the placebo-controlled period, 425 of 1029 patients (41.3%) experienced ARIA, with serious cases occurring in 14 patients (1.4%). ARIA-edema (ARIA-E) was the most common adverse event (362 of 1029 [35.2%]), and 263 initial events (72.7%) occurred within the first 8 doses of aducanumab; 94 participants (26.0%) with an event exhibited symptoms. Common associated symptoms among 103 patients with symptomatic ARIA-E or ARIA-H were headache (48 [46.6%]), confusion (15 [14.6%]), dizziness (11 [10.7%]), and nausea (8 [7.8%]). Incidence of ARIA-E was highest in aducanumab-treated participants who were apolipoprotein E ε4 allele carriers. Most events (479 of 488 [98.2%]) among those with ARIA-E resolved radiographically; 404 of 488 (82.8%) resolved within 16 weeks. In the placebo group, 29 of 1076 participants (2.7%) had ARIA-E (apolipoprotein E ε4 carriers: 16 of 742 [2.2%]; noncarriers, 13 of 334 [3.9%]). ARIA-microhemorrhage and ARIA-superficial siderosis occurred in 197 participants (19.1%) and 151 participants (14.7%), respectively. Conclusions and Relevance: In this integrated safety data set from EMERGE and ENGAGE, the most common adverse event in the 10-mg/kg group was ARIA-E, which occurred in 362 of the 1029 patients (35.2%) in the 10-mg/kg group with at least 1 postbaseline MRI scan, with 94 patients (26.0%) experiencing associated symptoms. The most common associated symptom was headache. Trial Registrations: ClinicalTrials.gov Identifiers: NCT02484547, NCT02477800.

Measure synchronization in a two-species bosonic Josephson junction.

Measure synchronization (MS) in a two-species bosonic Josephson junction (BJJ) is studied based on semiclassical theory. Six different scenarios for MS, including two in the Josephson oscillation regime (the zero-phase mode) and four in the self-trapping regime (the π-phase mode), are clearly shown. Systematic investigations of the common features behind these different scenarios are performed. We show that the average energies of the two species merge at the MS transition point. The scaling of the power law near the MS transition is verified and the critical exponent is 1/2 for all of the different scenarios for MS. We also illustrate MS in a three-dimensional phase space; from this illustration, more detailed information on the dynamical process can be obtained. In particular, by analyzing the Poincaré sections with changing interspecies interactions, we find that the two-species BJJ exhibits separatrix crossing behavior at the MS transition point and such behavior depicts the general mechanism behind the different scenarios for the MS transitions. The new critical behavior found in a two-species BJJ is expected to be found in real systems of atomic Bose gases.

Homolog interaction during meiotic prophase I in Arabidopsis requires the SOLO DANCERS gene encoding a novel cyclin-like protein.

Interactions between homologs in meiotic prophase I, such as recombination and synapsis, are critical for proper homolog segregation and involve the coordination of several parallel events. However, few regulatory genes have been identified; in particular, it is not clear what roles the proteins similar to the mitotic cell cycle regulators might play during meiotic prophase I. We describe here the isolation and characterization of a new Arabidopsis mutant called solo dancers that exhibits a severe defect in homolog synapsis, recombination and bivalent formation in meiotic prophase I, subsequently resulting in seemingly random chromosome distribution and formation of abnormal meiotic products. We further demonstrate that the mutation affects a meiosis-specific gene encoding a novel protein of 578 amino acid residues with up to 31% amino acid sequence identity to known cyclins in the C-terminal portion. These results argue strongly that homolog interactions during meiotic prophase I require a novel meiosis-specific cyclin in Arabidopsis.

The excess microsporocytes1 gene encodes a putative leucine-rich repeat receptor protein kinase that controls somatic and reproductive cell fates in the Arabidopsis anther.

Cell differentiation is essential for the development of multicellular organisms. In flowering plants, the haploid male gametophytes (pollen grains) are generated in the anther from reproductive cells called microsporocytes. Several types of somatic cells ensure successful pollen development, and thus reproduction. However, it is not clear what genes regulate the differentiation of these diverse, highly specialized cells in the anther. We report here the isolation and characterization of a novel Arabidopsis thaliana male sterile mutant, excess microsporocytes1 (ems1), that produces excess microsporocytes, lacks tapetal cells, and abnormally maintains middle layer cells. Although the meiotic nuclear division in the ems1 mutant is normal, the microsporocytes do not undergo cytokinesis, resulting in failed microsporogenesis and male sterility. The EMS1 gene encodes a putative leucine-rich repeat receptor protein kinase (LRR-RPK), and its expression is associated with the differentiation of the microsporocytes and tapetal cells, suggesting that EMS1 mediates signals that control the fate of reproductive cells and their contiguous somatic cells.

Genome-wide analysis of the cyclin family in Arabidopsis and comparative phylogenetic analysis of plant cyclin-like proteins.

Cyclins are primary regulators of the activity of cyclin-dependent kinases, which are known to play critical roles in controlling eukaryotic cell cycle progression. While there has been extensive research on cell cycle mechanisms and cyclin function in animals and yeasts, only a small number of plant cyclins have been characterized functionally. In this paper, we describe an exhaustive search for cyclin genes in the Arabidopsis genome and among available sequences from other vascular plants. Based on phylogenetic analysis, we define 10 classes of plant cyclins, four of which are plant-specific, and a fifth is shared between plants and protists but not animals. Microarray and reverse transcriptase-polymerase chain reaction analyses further provide expression profiles of cyclin genes in different tissues of wild-type Arabidopsis plants. Comparative phylogenetic studies of 174 plant cyclins were also performed. The phylogenetic results imply that the cyclin gene family in plants has experienced more gene duplication events than in animals. Expression patterns and phylogenetic analyses of Arabidopsis cyclin genes suggest potential gene redundancy among members belonging to the same group. We discuss possible divergence and conservation of some plant cyclins. Our study provides an opportunity to rapidly assess the position of plant cyclin genes in terms of evolution and classification, serving as a guide for further functional study of plant cyclins.