Comparison of quantitative Krenning Scores with visual assessment in 99mTc-EDDA/HYNIC-TOC SPECT-CT

Nuclear Medicine Communications, Год журнала: 2025, Номер unknown

Опубликована: Фев. 27, 2025

Purpose The aim of this study is to assess inter-observer variability the Krenning Score for 99m Tc-EDDA/HYNIC-TOC single photon emission computed tomography (SPECT)-computed (CT) images and compare against quantitative metrics obtained from tumour physiological uptake measurements. Methods Thirty-two patients with 117 lesions visible on SPECT-CT were scored by two expert observers using Score. Five less extensive experience also visual assessment. Inter-observer agreement comparison consensus was tested. Three made measurements uptake, intra-observer variation investigated. Assessment between made. Results assessment 44.3% proportions 0.576 Fleiss’ Kappa, whilst best-performing metric Kappa equal 1. observer 89.8% percentage 0.914 Cohen’s similar (a derived Score) at 86.4% κ = 0.877. Standardised value maximum (SUV max ) showed levels 85.1% 0.871. Conclusion A Score, or alternatively SUV , can provide an as assessment, reduced variability. Quantification deliver greater consistency in scoring over important factor when imaging used determine patient eligibility peptide receptor radiotherapy.

Язык: Английский

---

Deep learning image enhancement algorithms in PET/CT imaging: a phantom and sarcoma patient radiomic evaluation

European Journal of Nuclear Medicine and Molecular Imaging, Год журнала: 2025, Номер unknown

Опубликована: Фев. 27, 2025

PET/CT imaging data contains a wealth of quantitative information that can provide valuable contributions to characterising tumours. A growing body work focuses on the use deep-learning (DL) techniques for denoising PET data. These models are clinically evaluated prior use, however, image assessment provides potential further evaluation. This uses radiomic features compare two manufacturer enhancement algorithms, one which has been commercialised, against 'gold-standard' reconstruction in phantom and sarcoma patient set (N=20). All studies retrospective clinical [ 18 F]FDG dataset were acquired either GE Discovery 690 or 710 scanner with volumes segmented by an experienced nuclear medicine radiologist. The modular heterogeneous used this was filled F]FDG, five repeat acquisitions scanner. DL-enhanced images compared algorithms trained emulate input images. difference between sets tested significance 93 international biomarker standardisation initiative (IBSI) standardised features. Comparing 'gold-standard', 4.0% 9.7% measured significantly different (pcritical < 0.0005) respectively (averaged over DL algorithms). Larger differences observed comparing algorithm 29.8% 43.0% measuring found be similar generated using target method more than 80% not all comparisons across unseen result offers insight into performance demonstrate applications harmonisation radiomics evaluation algorithms.

Язык: Английский

---

Robust vs. Non-robust radiomic features: the quest for optimal machine learning models using phantom and clinical studies

Cancer Imaging, Год журнала: 2025, Номер 25(1)

Опубликована: Март 12, 2025

Abstract Purpose This study aimed to select robust features against lung motion in a phantom and use them as input feature selection algorithms machine learning classifiers clinical predict the lymphovascular invasion (LVI) of non-small cell cancer (NSCLC). The results were also compared with conventional techniques without considering robustness radiomic features. Methods An in-house developed was two 22mm lesion sizes based on study. A specific motor built simulate orthogonal directions. Lesions both studies segmented using Fuzzy C-means-based segmentation algorithm. After inducing extracting 105 4 sets, including shape, first-, second-, higher-order statistics from each region interest (ROI) image, statistical analyses performed motion. Subsequently, these total extracted 126 data. Various (FS) multiple (ML) implemented LVI NSCLC, followed by comparing predicting common not Results Our demonstrated that selecting FS ML surges sensitivity, which has gentle negative effect accuracy area under curve (AUC) predictions commonly used methods 12 15 outcomes. top performance prediction achieved NB classifier RFE 95% AUC, 67% accuracy, 100% sensitivity. Moreover, belonged Boruta 92% 86% Conclusion Robustness over various influential factors is critical should be considered Selecting solution overcome low reproducibility Although setting minor impact AUC prediction, it boosts sensitivity large extent.

Язык: Английский

---

Robustness of textural analysis features in quantitative 99 mTc and 177Lu SPECT-CT phantom acquisitions

EJNMMI Physics, Год журнала: 2025, Номер 12(1)

Опубликована: Апрель 17, 2025

Abstract Background Textural Analysis features in molecular imaging require to be robust under repeat measurement and independent of volume for optimum use clinical studies. Recent EANM SNMMI guidelines radiomics provide advice on the potential phantoms identify (Hatt EJNMMI, 2022). This study applies suggested SPECT quantification two radionuclides, 99 m Tc 177 Lu. Methods Acquisitions were made with a uniform phantom test dependency customised ‘Revolver’ phantom, based PET described Hatt (EJNMMI, 2022) but local adaptations SPECT. Each was filled separately Sixty-seven extracted tested robustness dependency. Results Features showing high or Coefficient Variation (indicating poor repeatability) removed from list that may suitable After feature reduction, there 39 33 Lu remaining. Conclusion The Revolver repeatable is possible quantitative using Selection such likely centre-dependent due differences camera performance as well acquisition reconstruction protocols.

Язык: Английский

---

Deep Learning-Based CT-Less Cardiac Segmentation of PET Images: A Robust Methodology for Multi-Tracer Nuclear Cardiovascular Imaging

Deleted Journal, Год журнала: 2025, Номер unknown

Опубликована: Май 6, 2025

Abstract Quantitative cardiovascular PET/CT imaging is useful in the diagnosis of multiple cardiac perfusion and motion pathologies. The common approach for segmentation consists using co-registered CT images, exploiting publicly available deep learning (DL)-based models. However, mismatch between structural images PET uptake limits usefulness these approaches. Besides, performance DL models not consistent over low-dose or ultra-low-dose commonly used clinical imaging. In this work, we developed a DL-based methodology to tackle issue by segmenting directly images. This study included 406 from 146 patients (43 18 F-FDG, 329 13 N-NH 3 , 37 82 Rb images). Using previously trained nnU-Net our group, segmented whole heart three main components, namely left myocardium (LM), ventricle cavity (LV), right (RV) on was resampled resolution edited through combination automated image processing manual correction. corrected masks SUV were fed V2 pipeline be fivefold data split strategy defining two tasks: task #1 #2 components. Fifteen as external validation set. delineated compared with standard reference Dice coefficient, Jaccard distance, mean surface segment volume relative error (%). Task average coefficient internal 0.932 ± 0.033. 15 cases comparable reaching an 0.941 0.018. 0.88 0.063, 0.828 0.091, 0.876 0.062 LM, LV, RV, respectively. There no statistically significant difference among coefficients, neither acquired radiotracers nor different folds ( P -values > 0.05). overall prediction components less than 2%. We acceptable accuracy robust test set nuclear proposed can overcome unreliable segmentations performed

Язык: Английский

---