AI-enhanced PET/CT image synthesis using CycleGAN for improved ovarian cancer imaging

Polish Journal of Radiology, Journal Year: 2025, Volume and Issue: 90, P. 26 - 35

Published: Jan. 17, 2025

Purpose Ovarian cancer is the fifth fatal among women. Positron emission tomography (PET), which offers detailed metabolic data, can be effectively used for early screening. However, proper attenuation correction essential interpreting data obtained by this imaging modality. Computed (CT) commonly performed alongside PET correction. This approach may introduce some issues in spatial alignment and registration of images two modalities. study aims to perform image using generative adversarial networks (GANs), without additional CT imaging. Material methods The PET/CT from 55 ovarian patients were study. Three GAN architectures: Conditional GAN, Wasserstein CycleGAN, evaluated statistical performance each model was assessed calculating mean squared error (MSE) absolute (MAE). radiological assessments models comparing standardised uptake value Hounsfield unit values whole body selected organs, synthetic real images. Results Based on results, CycleGAN demonstrated effective pseudo-CT generation, with high accuracy. MAE MSE all 2.15 ± 0.34 3.14 0.56, respectively. For reconstruction, such found 4.17 0.96 5.66 1.01, Conclusions results showed potential deep learning reducing radiation exposure improving quality Further refinement clinical validation are needed full applicability.

Language: Английский

---

Hybrid Vision Transformer and Xception Model for Reliable CT-Based Ovarian Neoplasms Diagnosis

Intelligence-Based Medicine, Journal Year: 2025, Volume and Issue: unknown, P. 100227 - 100227

Published: Feb. 1, 2025

Language: Английский

---

The YOLO Framework: A Comprehensive Review of Evolution, Applications, and Benchmarks in Object Detection

Computers, Journal Year: 2024, Volume and Issue: 13(12), P. 336 - 336

Published: Dec. 14, 2024

This paper provides a comprehensive review of the YOLO (You Only Look Once) framework up to its latest version, 11. As state-of-the-art model for object detection, has revolutionized field by achieving an optimal balance between speed and accuracy. The traces evolution variants, highlighting key architectural improvements, performance benchmarks, applications in domains such as healthcare, autonomous vehicles, robotics. It also evaluates framework’s strengths limitations practical scenarios, addressing challenges like small environmental variability, computational constraints. By synthesizing findings from recent research, this work identifies critical gaps literature outlines future directions enhance YOLO’s adaptability, robustness, integration into emerging technologies. researchers practitioners with valuable insights drive innovation detection related applications.

Language: Английский

---

OVision A raspberry Pi powered portable low cost medical device framework for cancer diagnosis

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Feb. 28, 2025

Cancer remains a major global health challenge, with significant disparities in access to advanced diagnostic and prognostic technologies, especially resource-constrained settings. Existing medical treatments devices for cancer diagnosis are often prohibitively expensive, limiting their reach impact. Pathologists' scarcity exacerbates accuracy, elevating mortality risks. To address these critical issues, this study presents OVision - low cost, deep learning-powered framework developed assist histopathological diagnosis. The key objective is leverage the portable, low-power computing Raspberry Pi. By designing standalone that eliminate need internet connectivity high-end infrastructure, we can dramatically reduce costs while maintaining accuracy. As proof of concept, demonstrated viability through compact, self-contained device capable accurately detecting ovarian subtypes 95% on par traditional methods, costing small fraction price. This off-grid solution has immense potential improve precision diagnostics, underserved regions world lack resources deploy infrastructure-heavy technologies. In addition, by classifying each tile, tool provide percentages histologic subtype detected within slide. capability enhances precision, offering detailed overview heterogeneity tissue sample, helps understanding complexity tailoring personalized treatment plans. conclusion, work proposes transformative model developing affordable, accessible bring healthcare benefits all, laying foundation more equitable, inclusive future medicine.

Language: Английский

---

Ovarian masses suggested for MRI examination: assessment of deep learning models based on non-contrast-enhanced MRI sequences for predicting malignancy

Abdominal Radiology, Journal Year: 2025, Volume and Issue: unknown

Published: March 21, 2025

Language: Английский

---

Improving Ovarian Cancer Subtyping with Computer Vision Models on Tiled Histopathological Images

Deleted Journal, Journal Year: 2025, Volume and Issue: unknown

Published: May 20, 2025

Ovarian cancer remains one of the most challenging cancers to diagnose due its non-specific symptoms, lack reliable screening tests, and complexity detecting abnormalities. Accurate subtype classification is crucial for personalised treatment improved patient outcomes. In this study, we developed a machine learning pipeline fine-tuning pre-trained computer vision models classify ovarian subtypes from whole slide images (WSI). Using targeted tissue masks necrosis, stroma, tumour regions as proof concept, demonstrated efficacy tiling masked transform complex detection-then-classification problem into simpler task. Our method achieved high accuracy in tile-level classification, with subsequent extension via majority voting on tiled images. Precision exceeds 90% across subtypes, which highlights potential scalable, automated systems assist diagnostics. These findings contribute broader field computational pathology, paving way enhanced diagnostic consistency accessibility clinical settings.

Language: Английский

---

Diagnosis of Medical Images Using Convolutional Neural Networks

Deleted Journal, Journal Year: 2024, Volume and Issue: 20(6s), P. 2371 - 2376

Published: May 2, 2024

Medical image diagnosis using Convolutional Neural Networks (CNNs) has emerged as a viable way to improve the accuracy and efficiency of disease identification categorization in clinical settings. In this study, they look at how CNNs can be used diagnose lung nodules from chest X-ray pictures, provide insights into technology's performance future applications. A dataset 10,000 tagged pictures showing both benign malignant was obtained preprocessed standard methods. The construct train proprietary CNN architecture, which then rigorously evaluated on distinct training, validation, test sets. model showed good (94.8%), sensitivity (92.1%), specificity (96.5%), precision, recall, F1 score, area under ROC curve (AUC), indicating its robustness generalization ability. These findings show that CNN-based diagnostic tools may help radiologists physicians discover cancer earlier, improving patient outcomes optimizing healthcare delivery. However, difficulties such interpretability, data privacy, regulatory approval must addressed before fully utilized medical imaging. This study emphasizes CNNs' transformative significance medicine necessity for additional research development realize their full potential practice.

Language: Английский

---

Decision-Level Fusion Classification of Ovarian CT Benign and Malignant Tumors Based on Radiomics and Deep Learning of Dual Views

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 102381 - 102395

Published: Jan. 1, 2024

Ovarian cancer is one of the most prevalent malignant tumors in female reproductive system, and its early diagnosis has always posed a challenge. Computed tomography (CT) widely utilized clinical management tool that can extract much detail through computer algorithms, playing vital role ovarian cancer. This research aims to develop an benign-malignant classification model based on radiomics deep learning dual views. A retrospective analysis CT images from 135 tumor patients was conducted using StratifiedKFold method (K =5) for cross-validation. Radiomics features were extracted data inputted into automated machine (A-ML) framework. Meanwhile, (DL) called Dual-View Global Representation Local Cross Transformer (D_GR_LCT) proposed global-local parallel approach end-to-end training. results indicate superiority 3D input over 2D, with average AUC-ROC 88.35% AUC-PR 88.73%. Comparative experiments demonstrate enhanced performance parameter settings. The DL achieves 88.15% 85.17%, respectively, validated by ablative comparative experiments. At decision-making level, fusion models demonstrates 91.35% 90.20%, utilizing stacking method. outperformed individual models. Thus, dual-view are recommended identification screening practice.

Language: Английский

---

Optimizing Deep Learning Models for Ovarian Cancer Subtype Classification: A Systematic Evaluation of Architectures and Data Augmentation Strategies

Research Square (Research Square), Journal Year: 2025, Volume and Issue: unknown

Published: May 7, 2025

Ovarian cancer is a leading cause of cancer-related mortality among women, and accurate classification its subtypes critical for effective treatment planning. This study systematically investigates the impact different network architectures data augmentation strategies on ovarian subtype classification. We evaluate two baseline models (VGG ViT) propose an efficient hybrid model that integrates convolutional self-attention mechanisms to balance local feature extraction global context modeling. Furthermore, we conduct comprehensive assessment various techniques, including geometric, color, spatial transformations, determine their effects generalization. Additionally, compare pre-trained non-pre-trained analyze benefits transfer learning in this domain. To enhance interpretability, utilize Grad-CAM visualizations examine decision-making processes models. Our findings reveal while ViT exhibits superior generalization capabilities with pre-training, VGG remains competitive even without pre-training due strong inductive biases. Among tested strategies, geometric transformations significantly improve performance, whereas color-based augmentations show limited or degrade performance. The proposed achieves comparable accuracy maintaining smaller parameter scale faster training efficiency. In conclusion, provides key insights into selection techniques pathological image design framework offers interpretable approach classification, potential applications broader medical imaging tasks.

Language: Английский

---

Bayesian and deep‐learning models applied to the early detection of ovarian cancer using multiple longitudinal biomarkers

Cancer Medicine, Journal Year: 2024, Volume and Issue: 13(7)

Published: April 1, 2024

Abstract Background Ovarian cancer is the most lethal of all gynecological cancers. Cancer Antigen 125 (CA125) best‐performing ovarian biomarker which however still not effective as a screening test in general population. Recent literature reports additional biomarkers with potential to improve on CA125 for early detection when using longitudinal multimarker models. Methods Our data comprised 180 controls and 44 cases serum samples sourced from multimodal arm UK Collaborative Trial Screening (UKCTOCS). models were based Bayesian change‐point recurrent neural networks. Results We obtained significantly higher performance CA125–HE4 model both methodologies (AUC 0.971, sensitivity 96.7% AUC 0.987, 96.7%) respect 0.949, 90.8% 0.953, 92.1%) (BCP) networks (RNN) approaches, respectively. One year before diagnosis, also ranked best, whereas at 2 years diagnosis no outperformed CA125. Conclusions study identified tested different combination multivariable that alone. showed candidate increase rate cancer.

Language: Английский

---