Towards more precise automatic analysis: a systematic review of deep learning-based multi-organ segmentation

BioMedical Engineering OnLine, Journal Year: 2024, Volume and Issue: 23(1)

Published: June 8, 2024

Abstract Accurate segmentation of multiple organs in the head, neck, chest, and abdomen from medical images is an essential step computer-aided diagnosis, surgical navigation, radiation therapy. In past few years, with a data-driven feature extraction approach end-to-end training, automatic deep learning-based multi-organ methods have far outperformed traditional become new research topic. This review systematically summarizes latest this field. We searched Google Scholar for papers published January 1, 2016 to December 31, 2023, using keywords “multi-organ segmentation” “deep learning”, resulting 327 papers. followed PRISMA guidelines paper selection, 195 studies were deemed be within scope review. summarized two main aspects involved segmentation: datasets methods. Regarding datasets, we provided overview existing public conducted in-depth analysis. Concerning methods, categorized approaches into three major classes: fully supervised, weakly supervised semi-supervised, based on whether they require complete label information. achievements these terms accuracy. discussion conclusion section, outlined current trends segmentation.

Language: Английский

---

A multi-scale self-calibrating lung nodule detection based on SPC-UNet

Biomedical Signal Processing and Control, Journal Year: 2025, Volume and Issue: 104, P. 107515 - 107515

Published: Jan. 18, 2025

Language: Английский

---

A Survey on Deep Active Learning: Recent Advances and New Frontiers

IEEE Transactions on Neural Networks and Learning Systems, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 21

Published: Jan. 1, 2024

Active learning seeks to achieve strong performance with fewer training samples. It does this by iteratively asking an oracle label newly selected samples in a human-in-the-loop manner. This technique has gained increasing popularity due its broad applicability, yet survey papers, especially for deep active (DAL), remain scarce. Therefore, we conduct advanced and comprehensive on DAL. We first introduce reviewed paper collection filtering. Second, formally define the DAL task summarize most influential baselines widely used datasets. Third, systematically provide taxonomy of methods from five perspectives, including annotation types, query strategies, model architectures, paradigms, processes, objectively analyze their strengths weaknesses. Then, comprehensively main applications natural language processing (NLP), computer vision (CV), data mining (DM), so on. Finally, discuss challenges perspectives after detailed analysis current studies. work aims serve as useful quick guide researchers overcoming difficulties hope that will spur further progress burgeoning field.

Language: Английский

---

Cf-Wiad: Consistency Fusion with Weighted Instance and Adaptive Distribution for Enhanced Semi-Supervised Skin Lesion Classification

Published: Jan. 1, 2025

Language: Английский

---

Dynamic blinking feature extraction for automated facial nerve paralysis detection

Computers in Biology and Medicine, Journal Year: 2025, Volume and Issue: 187, P. 109722 - 109722

Published: Feb. 5, 2025

Language: Английский

---

DeepPath: Overcoming data scarcity for protein transition pathway prediction using physics-based deep learning

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2025, Volume and Issue: unknown

Published: March 2, 2025

The structural dynamics of proteins play a crucial role in their function, yet most experimental and deep learning methods produce only static models. While molecular (MD) simulations provide atomistic insight into conformational transitions, they remain computationally prohibitive, particularly for large-scale motions. Here, we introduce DeepPath, deep-learning-based framework that rapidly generates physically realistic transition pathways between known protein states. Unlike conventional supervised approaches, DeepPath employs active to iteratively refine its predictions, leveraging mechanical force fields as an oracle guide pathway generation. We validated on three biologically relevant test cases: SHP2 activation, CdiB H1 secretion, the BAM complex lateral gate opening. accurately predicted all cases, reproducing key intermediate structures transient interactions observed previous studies. Notably, also inwardand outward-open states closely aligns with experimentally hybrid-barrel structure (TMscore = 0.91). Across achieved accurate predictions within hours, showcasing efficient alternative MD exploring transitions.

Language: Английский

---

Improving gastric lesion detection with synthetic images from diffusion models

Smart Health, Journal Year: 2025, Volume and Issue: unknown, P. 100569 - 100569

Published: March 1, 2025

Language: Английский

---

MDAL: Modality-difference-based active learning for multimodal medical image analysis via contrastive learning and pointwise mutual information

Computerized Medical Imaging and Graphics, Journal Year: 2025, Volume and Issue: unknown, P. 102544 - 102544

Published: April 1, 2025

Language: Английский

---

Active learning in computational pathology with noise detection empowered by loss-based prior and feature analysis

Biomedical Signal Processing and Control, Journal Year: 2025, Volume and Issue: 108, P. 107953 - 107953

Published: May 1, 2025

Language: Английский

---

Multiple deep learning models based on MRI images in discriminating glioblastoma from solitary brain metastases: a multicentre study

BMC Medical Imaging, Journal Year: 2025, Volume and Issue: 25(1)

Published: May 19, 2025

Development of a deep learning model for accurate preoperative identification glioblastoma and solitary brain metastases by combining multi-centre multi-sequence magnetic resonance images comparison the performance different models. Clinical data MR total 236 patients with pathologically confirmed single were retrospectively collected from January 2019 to May 2024 at Provincial Hospital Shandong First Medical University, randomly divided into training set test according ratio 8:2, in which contained 197 cases 39 cases; preprocessed labeled tumor regions. The pre-processed regions, MRI sequences input individually or combination train 3D ResNet-18, optimal sequence combinations obtained five-fold cross-validation enhancement inputs models Vision Transformer (3D Vit), DenseNet, VGG; working characteristic curves (ROCs) subjects plotted, area under curve (AUC) was calculated. (AUC), accuracy, precision, recall F1 score used evaluate discriminative In addition, 48 2020 December 2022 Affiliated Cancer University as an external compare performance, robustness generalization ability four effect sequences, three T1-CE, T2, T2-Flair gained effect, accuracy AUC values 0.8718 0.9305, respectively; after inputted aforementioned combinations, validation ResNet-18 0.8125, respectively, 0.8899, all are highest among A can efficiently identify preoperatively, has efficacy identifying two types tumours.

Language: Английский

---