A bird’s-eye view of deep learning in bioimage analysis

Computational and Structural Biotechnology Journal, Journal Year: 2020, Volume and Issue: 18, P. 2312 - 2325

Published: Jan. 1, 2020

Deep learning of artificial neural networks has become the de facto standard approach to solving data analysis problems in virtually all fields science and engineering. Also biology medicine, deep technologies are fundamentally transforming how we acquire, process, analyze, interpret data, with potentially far-reaching consequences for healthcare. In this mini-review, take a bird’s-eye view at past, present, future developments learning, starting from large, biomedical imaging, bioimage particular.

Language: Английский

---

Using deep convolutional neural networks for multi-classification of thyroid tumor by histopathology: a large-scale pilot study

Annals of Translational Medicine, Journal Year: 2019, Volume and Issue: 7(18), P. 468 - 468

Published: Sept. 1, 2019

Background: To explore whether deep convolutional neural networks (DCNNs) have the potential to improve diagnostic efficiency and increase level of interobserver agreement in classification thyroid nodules histopathological slides. Methods: A total 11,715 fragmented images from 806 patients' original histological were divided into a training dataset test dataset. Inception-ResNet-v2 VGG-19 trained using tested determine efficiencies different histologic types nodules, including normal tissue, adenoma, nodular goiter, papillary carcinoma (PTC), follicular (FTC), medullary (MTC) anaplastic (ATC). Misdiagnoses further analyzed. Results: The for each pathology type at ratio 5:1. Using set, yielded better average accuracy than did (97.34% vs. 94.42%, respectively). model applied 7 showed fragmentation 88.33% 98.57% ATC, 98.89% FTC, 100% MTC, 97.77% PTC, goiter 92.44% adenoma. It achieved excellent all malignant types. Normal tissue adenoma most challenging classify. Conclusions: DCNN models, especially VGG-19, satisfactory accuracies on task differentiating tumors by histopathology. Analysis misdiagnosed cases revealed that differentiate, while classifications efficiencies. results indicate models may facilitating histopathologic disease diagnosis.

Language: Английский

---

Weakly supervised annotation‐free cancer detection and prediction of genotype in routine histopathology

The Journal of Pathology, Journal Year: 2021, Volume and Issue: 256(1), P. 50 - 60

Published: Sept. 25, 2021

Deep learning is a powerful tool in computational pathology: it can be used for tumor detection and predicting genetic alterations based on histopathology images alone. Conventionally, prediction of are two separate workflows. Newer methods have combined them, but require complex, manually engineered pipelines, restricting reproducibility robustness. To address these issues, we present new method simultaneous alterations: The Slide-Level Assessment Model (SLAM) uses single off-the-shelf neural network to predict molecular directly from routine pathology slides without any manual annotations, improving upon previous by automatically excluding normal non-informative tissue regions. SLAM requires only standard programming libraries conceptually simpler than approaches. We extensively validated clinically relevant tasks using large multicentric cohorts colorectal cancer patients, Darmkrebs: Chancen der Verhütung durch Screening (DACHS) Germany Yorkshire Cancer Research Bowel Improvement Programme (YCR-BCIP) the UK. show that yields reliable slide-level classification presence with an area under receiver operating curve (AUROC) 0.980 (confidence interval 0.975, 0.984; n = 2,297 1,281 slides). In addition, detect microsatellite instability (MSI)/mismatch repair deficiency (dMMR) or stability/mismatch proficiency AUROC 0.909 (0.888, 0.929; 2,039 patients) BRAF mutational status 0.821 (0.786, 0.852; 2,075 patients). improvement respect was external testing cohort which MSI/dMMR detected 0.900 (0.864, 0.931; 805 provides human-interpretable visualization maps, enabling analysis multiplexed predictions human experts. summary, simple could applied multiple disease contexts. © 2021 Authors. Journal Pathology published John Wiley & Sons, Ltd. behalf Pathological Society Great Britain Ireland.

Language: Английский

---

Artificial intelligence and pathology: From principles to practice and future applications in histomorphology and molecular profiling

Seminars in Cancer Biology, Journal Year: 2021, Volume and Issue: 84, P. 129 - 143

Published: Feb. 22, 2021

Language: Английский

---

The prognostic value and molecular properties of tertiary lymphoid structures in oesophageal squamous cell carcinoma

Clinical and Translational Medicine, Journal Year: 2022, Volume and Issue: 12(10)

Published: Oct. 1, 2022

Tertiary lymphoid structures (TLSs) play key roles in tumour adaptive immunity. However, the prognostic value and molecular properties of TLSs oesophageal squamous cell carcinoma (ESCC) patients have not been studied.The values presence maturation status tumour-associated were determined 394 256 ESCC from Sun Yat-sen University Cancer Center (Centre A) Hospital Shantou Medical College B), respectively. A deep-learning (DL) TLS classifier was established with haematoxylin eosin (H&E)-stained slides using an inception-resnet-v2 neural network. Digital spatial profiling performed to determine cellular tissues.TLSs observed 73.1% ESCCs Centre via pathological examination H&E-stained primary slides, among which 42.9% TLS-mature 30.2% TLS-immature tumours. The DL yielded favourable sensitivities specificities for patient identification evaluation, 55.1%, 39.5% 5.5% B identified as TLS-mature, TLS-negative Multivariate analyses proved that mature independent factor both cohorts (p < .05). Increased proportions proliferative B, plasma CD4+ T helper (Th) cells increased memory Th17 signatures compared immature ones. Intratumoural CD8+ infiltration tissues TLS-absent tissues. combination high associated best survival patients.Mature improve prognosis who underwent complete resection. use would facilitate precise efficient evaluation offer a novel probability treatment individualization.

Language: Английский

---

Artificial Intelligence and Machine Learning Technologies in Cancer Care: Addressing Disparities, Bias, and Data Diversity

Cancer Discovery, Journal Year: 2022, Volume and Issue: 12(6), P. 1423 - 1427

Published: June 2, 2022

Summary: Artificial intelligence (AI) and machine learning (ML) technologies have not only tremendous potential to augment clinical decision-making enhance quality care precision medicine efforts, but also the worsen existing health disparities without a thoughtful, transparent, inclusive approach that includes addressing bias in their design implementation along cancer discovery continuum. We discuss applications of AI/ML tools provide recommendations for mitigating with AI ML while promoting equity.

Language: Английский

---

Digital Pathology: Transforming Diagnosis in the Digital Age

Cureus, Journal Year: 2023, Volume and Issue: unknown

Published: Sept. 3, 2023

In the context of rapid technological advancements, narrative review titled "Digital Pathology: Transforming Diagnosis in Digital Age" explores significant impact digital pathology reshaping diagnostic approaches. This delves into various effects field, including remote consultations and artificial intelligence (AI)-assisted analysis, revealing ongoing transformation taking place. The investigation process digitizing traditional glass slides, which aims to improve accessibility facilitate sharing. Additionally, it addresses complexities associated with data security standardization challenges. Incorporating AI enhances pathologists' capabilities accelerates analytical procedures. Furthermore, highlights growing importance collaborative networks facilitating global knowledge It also emphasizes this technology on medical education patient care. provide an overview pathology's transformative innovative potential, highlighting its disruptive nature practices.

Language: Английский

---

Drug repositioning based on weighted local information augmented graph neural network

Briefings in Bioinformatics, Journal Year: 2023, Volume and Issue: 25(1)

Published: Nov. 22, 2023

Abstract Drug repositioning, the strategy of redirecting existing drugs to new therapeutic purposes, is pivotal in accelerating drug discovery. While many studies have engaged modeling complex drug–disease associations, they often overlook relevance between different node embeddings. Consequently, we propose a novel weighted local information augmented graph neural network model, termed DRAGNN, for repositioning. Specifically, DRAGNN firstly incorporates attention mechanism dynamically allocate coefficients and disease heterogeneous nodes, enhancing effectiveness target collection. To prevent excessive embedding limited vector space, omit self-node aggregation, thereby emphasizing valuable homogeneous information. Additionally, average pooling neighbor aggregation introduced enhance while maintaining simplicity. A multi-layer perceptron then employed generate final association predictions. The model’s repositioning supported by 10-times 10-fold cross-validation on three benchmark datasets. Further validation provided through analysis predicted associations using multiple authoritative data sources, molecular docking experiments analysis, laying solid foundation future

Language: Английский

---

Artificial intelligence in accelerating vaccine development - current and future perspectives

Frontiers in Bacteriology, Journal Year: 2023, Volume and Issue: 2

Published: Oct. 9, 2023

Tackling antimicrobial resistance requires the development of new drugs and vaccines. Artificial intelligence (AI) assisted computational approaches offer an alternative to traditionally empirical drug vaccine discovery pipelines. In this mini review, we focus on increasingly important role that AI now plays in vaccines provide reader with methods used identify candidate candidates for selected multi-drug resistant bacteria.

Language: Английский

---

Artificial intelligence and machine learning in ocular oncology: Retinoblastoma

Indian Journal of Ophthalmology, Journal Year: 2023, Volume and Issue: 71(2), P. 424 - 430

Published: Feb. 1, 2023

Purpose: This study was done to explore the utility of artificial intelligence (AI) and machine learning in diagnosis grouping intraocular retinoblastoma (iRB). Methods: It a retrospective observational using AI Machine learning, Computer Vision (OpenCV). Results: Of 771 fundus images 109 eyes, 181 had no tumor 590 displayed iRB based on review by two independent ocular oncologists (with an interobserver variability <1%). The sensitivity, specificity, positive predictive value, negative value trained model were 85%, 99%, 99.6%, 67%, respectively. for detection RB 96%, 94%, 97%, 91%, these, eyes normal (n = 31) or belonged groupA (n=1), B (n=22), C (n=8), D (n=23),and E (n=24) 0%). 100%, 100% group A; 82%, 20 21 98%, 90%, 96% B; 63%, 83%, 97% C; 78%, 94% D, 92%, 73%, 98% E, Conclusion: Based our study, we conclude that is highly sensitive with high specificity classification iRB.

Language: Английский

---