Deep-learning-based radiomics of intratumoral and peritumoral MRI images to predict the pathological features of adjuvant radiotherapy in early-stage cervical squamous cell carcinoma

BMC Women s Health, Journal Year: 2024, Volume and Issue: 24(1)

Published: March 19, 2024

Abstract Background Surgery combined with radiotherapy substantially escalates the likelihood of encountering complications in early-stage cervical squamous cell carcinoma(ESCSCC). We aimed to investigate feasibility Deep-learning-based radiomics intratumoral and peritumoral MRI images predict pathological features adjuvant ESCSCC minimize occurrence adverse events associated treatment. Methods A dataset comprising MR was obtained from 289 patients who underwent radical hysterectomy pelvic lymph node dissection between January 2019 April 2022. The randomly divided into two cohorts a 4:1 ratio.The postoperative options were evaluated according Peter/Sedlis standard. extracted clinical features, as well radiomic using least absolute shrinkage selection operator (LASSO) regression. constructed Clinical Signature (Clinic_Sig), Radiomics (Rad_Sig) Deep Transformer Learning (DTL_Sig). Additionally, we fused Rad_Sig DTL_Sig create Radiomic (DLR_Sig). prediction performance models Area Under Curve (AUC), calibration curve, Decision Analysis (DCA). Results DLR_Sig showed high level accuracy predictive capability, demonstrated by area under curve (AUC) 0.98(95% CI: 0.97–0.99) for training cohort 0.79(95% 0.67–0.90) test cohort. In addition, Hosmer-Lemeshow test, which provided p -values 0.87 0.15 cohort, respectively, indicated good fit. DeLong that effectiveness significantly better than Clinic_Sig( P < 0.05 both cohorts). plot excellent consistency actual predicted probabilities, while DCA demonstrating greater utility predicting radiotherapy. Conclusion based on has potential preoperatively carcinoma (ESCSCC).

Language: Английский

---

Fully automated 3D body composition analysis and its association with overall survival in head and neck squamous cell carcinoma patients

Frontiers in Oncology, Journal Year: 2023, Volume and Issue: 13

Published: Oct. 19, 2023

We developed a method for fully automated deep-learning segmentation of tissues to investigate if 3D body composition measurements are significant survival Head and Neck Squamous Cell Carcinoma (HNSCC) patients.3D including spine, spine muscles, abdominal subcutaneous adipose tissue (SAT), visceral (VAT), internal organs within volumetric region limited by L1 L5 levels was accomplished using deep convolutional architecture - U-net implemented in nnUnet framework. It trained on separate dataset 560 single-channel CT slices used pre-radiotherapy (Pre-RT) post-radiotherapy (Post-RT) whole PET/CT or scans 215 HNSCC patients. Percentages were overall analysis Cox proportional hazard (PH) model.Our learning model successfully segmented all mentioned with Dice's coefficient exceeding 0.95. The difference between Pre-RT post-RT abdomen muscles percentage, VAT percentage sum together BMI Cancer Site selected at the level 5% survival. Aside from Site, lowest ratio (HR) value (HR, 0.7527; 95% CI, 0.6487-0.8735; p = 0.000183) observed percentage.Fully quantitative

Language: Английский

---

RGGC-UNet: Accurate Deep Learning Framework for Signet Ring Cell Semantic Segmentation in Pathological Images

Bioengineering, Journal Year: 2023, Volume and Issue: 11(1), P. 16 - 16

Published: Dec. 23, 2023

Semantic segmentation of Signet Ring Cells (SRC) plays a pivotal role in the diagnosis SRC carcinoma based on pathological images. Deep learning-based methods have demonstrated significant promise computer-aided over past decade. However, many existing approaches rely heavily stacking layers, leading to repetitive computational tasks and unnecessarily large neural networks. Moreover, lack available ground truth data for SRCs hampers advancement techniques these cells. In response, this paper introduces an efficient accurate deep learning framework (RGGC-UNet), which is UNet including our proposed residual ghost block with coordinate attention, featuring encoder-decoder structure tailored semantic SRCs. We designed novel encoder using attention. Benefiting from utilization attention encoder, overhead model effectively minimized. For practical application diagnosis, we enriched DigestPath 2019 dataset fully annotated mask labels Experimental outcomes underscore that significantly surpasses other leading-edge models accuracy while ensuring efficiency.

Language: Английский

---

Assessing Machine Learning Models for Predicting Age with Intracranial Vessel Tortuosity and Thickness Information

Brain Sciences, Journal Year: 2023, Volume and Issue: 13(11), P. 1512 - 1512

Published: Oct. 26, 2023

This study aimed to develop and validate machine learning (ML) models that predict age using intracranial vessels’ tortuosity diameter features derived from magnetic resonance angiography (MRA) data. A total of 171 subjects’ three-dimensional (3D) time-of-flight MRA image data were considered for analysis. After annotations two endpoints in each arterial segment, such as the sum angle metrics, triangular index, relative length, product distance, well features, extracted used train ML prediction. Features right left internal carotid arteries (ICA) basilar inputs six regression with a four-fold cross validation. The random forest model resulted lowest root mean square error 14.9 years highest average coefficient determination 0.186. linear showed absolute percentage (MAPE) Pearson correlation (0.532). ICA vessel segment was most important feature contributing prediction out four considered. An descriptors modest between real ML-predicted age. Further studies are warranted assessment model’s predictions patients diseases.

Language: Английский

---

Light on Alzheimer’s disease: from basic insights to preclinical studies

Frontiers in Aging Neuroscience, Journal Year: 2024, Volume and Issue: 16

Published: March 18, 2024

Alzheimer’s disease (AD), referring to a gradual deterioration in cognitive function, including memory loss and impaired thinking skills, has emerged as substantial worldwide challenge with profound social economic implications. As the prevalence of AD continues rise population ages, there is an imperative demand for innovative imaging techniques help improve our understanding these complex conditions. Photoacoustic (PA) forms hybrid modality by integrating high-contrast optical deep-penetration ultrasound imaging. PA enables visualization characterization tissue structures multifunctional information at high resolution and, demonstrated promising preliminary results study diagnosis AD. This review endeavors offer thorough overview current applications potential on treatment. Firstly, structural, functional, molecular parameter changes associated AD-related brain captured will be summarized, shaping diagnostic standpoint this review. Then, therapeutic methods aimed discussed further. Lastly, solutions clinical expand extent into deeper scenarios proposed. While certain aspects might not fully covered, mini-review provides valuable insights treatment through utilization photothermal effects. We hope that it spark further exploration field, fostering improved earlier theranostics

Language: Английский

---