Improving Generation and Evaluation of Long Image Sequences for Embryo Development Prediction

Electronics, Journal Year: 2024, Volume and Issue: 13(3), P. 476 - 476

Published: Jan. 23, 2024

Generating synthetic time series data, such as videos, presents a formidable challenge complexity increases when it is necessary to maintain specific distribution of shown stages. One case embryonic development, where prediction and categorization are crucial for anticipating future outcomes. To address this challenge, we propose Siamese architecture based on diffusion models generate predictive long-duration development videos an evaluation method select the most realistic video in non-supervised manner. We validated model using standard metrics, Fréchet inception distance (FID), (FVD), structural similarity (SSIM), peak signal-to-noise ratio (PSNR), mean squared error (MSE). The proposed generates up 197 frames with size 128×128, considering real input images. Regarding quality all results showed improvements over default (FID = 129.18, FVD 802.46, SSIM 0.39, PSNR 28.63, MSE 97.46). On coherence stages, global stage 9.00 was achieved versus 13.31 59.3 methods. technique produces more accurate successfully removes cases that display sudden movements or changes.

Language: Английский

---

Breast Cancer Diagnosis Method Based on Cross-Mammogram Four-View Interactive Learning

Tomography, Journal Year: 2024, Volume and Issue: 10(6), P. 848 - 868

Published: June 1, 2024

Computer-aided diagnosis systems play a crucial role in the and early detection of breast cancer. However, most current methods focus primarily on dual-view analysis single breast, thereby neglecting potentially valuable information between bilateral mammograms. In this paper, we propose Four-View Correlation Contrastive Joint Learning Network (FV-Net) for classification mammogram images. Specifically, FV-Net focuses extracting matching features across four views mammograms while maximizing both their similarities dissimilarities. Through Cross-Mammogram Dual-Pathway Attention Module, feature is achieved, capturing consistency complementary effectively reducing misalignment. reconstituted maps derived from mammograms, Bilateral-Mammogram module performs associative contrastive learning positive negative sample pairs within each local region. This aims to maximize correlation similar enhance differentiation dissimilar representations. Our experimental results test set comprising 20% combined Mini-DDSM Vindr-mamo datasets, as well INbreast dataset, show that our model exhibits superior performance cancer compared competing methods.

Language: Английский

---

Advancements in Medical Imaging

Advances in healthcare information systems and administration book series, Journal Year: 2024, Volume and Issue: unknown, P. 106 - 123

Published: June 5, 2024

Medical imaging holds a pivotal role in modern healthcare, facilitating early disease identification, treatment planning, and patient progress monitoring. The integration of machine learning (ML) has significantly transformed medical imaging, offering automated analysis, pattern recognition, diagnostic support. However, notable paradigm shift emerged recent times, highlighting the ascendancy deep (DL) techniques, heralding new era this field. This exploration scrutinizes dynamic evolution within accentuating departure from conventional methods toward more advanced domain learning. It foundational principles as applied shedding light on constraints that prompted adoption methodologies. Furthermore, chapter explores efficacy models across diverse modalities encompassing MRI, CT scans, X-rays, ultrasound.

Language: Английский

---

TRANSFORMING BRAIN TUMOR DIAGNOSIS: VISION TRANSFORMERS COMBINED WITH ENSEMBLE TECHNIQUES

Published: July 26, 2024

Brain Tumor (BT) is widely recognized as one of the most prevalent illnesses worldwide, affecting approximately 24,810 people in year 2023. Most suffering from brain tumor disease belong to Southeast Asian and Western Pacific regions. Medical diagnostics using artificial intelligence deep learning models demonstrate efficacy addressing critical health challenges initial detection intervention BT. In this paper, we proposed ViT along with ensemble for multiclass classification detection. The work aims provide novel best solution problem a approach. Ensemble Learning obtained 96% accuracy loss 0.13 an F1-score, precision, recall 0.96. comparative result shows that Vision Transformer 90% 0.30 0.89 on MRI dataset containing 7023 images, which further divided into train test. promising results showcase potential system early accurate can be used tumors.

Language: Английский

---

Skin Cancer Diagnosis Using VGG16 and Transfer Learning: Analyzing the Effects of Data Quality over Quantity on Model Efficiency

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(17), P. 7447 - 7447

Published: Aug. 23, 2024

The recent increase in the prevalence of skin cancer, along with its significant impact on individuals’ lives, has garnered attention many researchers field deep learning models, especially following promising results observed using these models medical field. This study aimed to develop a system that can accurately diagnose one three types cancer: basal cell carcinoma (BCC), melanoma (MEL), and nevi (NV). Additionally, it emphasizes importance image quality, as studies focus quantity images used learning. In this study, transfer was employed pre-trained VGG-16 model alongside dataset sourced from Kaggle. Three were trained while maintaining same hyperparameters script ensure fair comparison. However, data train each varied observe specific effects hypothesize about quality within highest validation score selected for further testing separate test dataset, which had not seen before, evaluate model’s performance accurately. work contributes existing body research by demonstrating critical role enhancing diagnostic accuracy, providing comprehensive evaluation cancer detection offering insights guide future improvements

Language: Английский

---