Medical physics最新文献

{"title":"Lightweight cross-resolution coarse-to-fine network for efficient deformable medical image registration.","authors":"Jun Liu, Nuo Shen, Wenyi Wang, Xiangyu Li, Wei Wang, Yongfeng Yuan, Ye Tian, Gongning Luo, Kuanquan Wang","doi":"10.1002/mp.17827","DOIUrl":"https://doi.org/10.1002/mp.17827","url":null,"abstract":"<p><strong>Background: </strong>Accurate and efficient deformable medical image registration is crucial in medical image analysis. While recent deep learning-based registration methods have achieved state-of-the-art accuracy, they often suffer from extensive network parameters and slow inference times, leading to inefficiency. Efforts to reduce model size and input resolution can improve computational efficiency but frequently result in suboptimal accuracy.</p><p><strong>Purpose: </strong>To address the trade-off between high accuracy and efficiency, we propose a Lightweight Cross-Resolution Coarse-to-Fine registration framework, termed LightCRCF.</p><p><strong>Methods: </strong>Our method is built on an ultra-lightweight U-Net architecture with only 0.1 million parameters, offering remarkable efficiency. To mitigate accuracy degradation resulting from fewer parameters while preserving the lightweight nature of the networks, LightCRCF introduces three key innovations as follows: (1) selecting an efficient cross-resolution coarse-to-fine (C2F) registration strategy and integrating it into the lightweight network to progressively decompose the deformation fields into multiresolution subfields to capture fine-grained deformations; (2) a Texture-aware Reparameterization (TaRep) module that integrates Sobel and Laplacian operators to extract rich textural information; (3) a Group-flow Reparameterization (GfRep) module that captures diverse deformation modes by decomposing the deformation field into multiple groups. Furthermore, we introduce a structural reparameterization technique that enhances training accuracy through multibranch structures of the TaRep and GfRep modules, while maintaining efficient inference by equivalently transforming these multibranch structures into single-branch standard convolutions.</p><p><strong>Results: </strong>We evaluate LightCRCF against various methods on the three public MRI datasets (LPBA, OASIS, and ACDC) and one CT dataset (abdomen CT). Following the previous data division methods, the LPBA dataset comprises 30 training image pairs and nine testing image pairs. For the OASIS dataset, the training, validation, and testing data consist of 1275, 110, and 660 image pairs, respectively. Similarly, for the ACDC dataset, the training, validation, and testing data include 180, 20, and 100 image pairs, respectively. For intersubject registration of the abdomen CT dataset, there are 380 training pairs, six validation pairs, and 42 testing pairs. Compared to state-of-the-art C2F methods, LightCRCF achieves comparable accuracy scores (DSC, HD95, and MSE), while demonstrating significantly superior performance across all efficiency metrics (Params, VRAM, FLOPs, and inference time). Relative to efficiency-first approaches, LightCRCF significantly outperforms these methods in accuracy metrics.</p><p><strong>Conclusions: </strong>Our LightCRCF method offers a favorable trade-off between accuracy and efficiency, maintaining high ","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144048855","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Evaluation of a motion correction algorithm in lung cancer PET/CT: Phantom validation and patient studies.","authors":"Ziyang Wang, Jianjing Liu, Di Lu, Guoqing Sui, Yaya Wang, Lina Tong, Xueyao Liu, Yan Zhang, Jie Fu, Wengui Xu, Dong Dai","doi":"10.1002/mp.17846","DOIUrl":"https://doi.org/10.1002/mp.17846","url":null,"abstract":"<p><strong>Background: </strong>Data-driven gating (DDG) is an emerging technology that can reduce the respiratory motion artifacts in positron emission tomography (PET) images.</p><p><strong>Purpose: </strong>The aim of this study is to use phantom and patient data to validate the performance of DDG with a motion correction algorithm based on the reconstruct, register, and average (RRA) method.</p><p><strong>Methods: </strong>A customized motion platform drove the phantom (five spheres with diameters of 10-28 mm) using a periodic motion that had a duration of 3-5 s and amplitudes of 2-4 cm. Normalized ratio of ungated and RRA PET relative to the ground-truth static PET was calculated for RSUVmax, RSUVmean, RSUVpeak, RVolume, and relative contrast-to-noise ratio (RCNR). Additionally, 30 lung cancer patients with 76 lung lesions less than 3 cm in diameter were prospectively studied. The overall image quality of patient examination was scored using a 5-point scale by two radiologists. SUVmax, SUVmean, SUVpeak, volume, and CNR of lesions measured in ungated and RRA PET were compared, and subgroup analysis was conducted.</p><p><strong>Results: </strong>In RRA PET images, motion artifacts of the spheres in the phantom were effectively mitigated, regardless of changes in movement amplitudes or duration. For all spheres with different ranges of motion and cycles, RSUVmax, RSUVmean, RSUVpeak, and RCNR increased significantly (p ≤ 0.001) and RVolume decreased significantly (p < 0.001) in RRA PET images. The average radiologist scores of image quality were 3.90 ± 0.86 with RRA PET, and 3.03 ± 1.19 with ungated PET. In RRA PET images, the SUVmax (p < 0.001), SUVmean (p < 0.001), SUVpeak (p < 0.001), and CNR (p < 0.001) of the lesions increased, while the volume (p < 0.001) of the lesions decreased. Δ%SUVmax, Δ%SUVmean, Δ%SUVpeak, and Δ%CNR of the lesions increased by 3.9%, 6.5%, 5.6%, and 4.3%, respectively, while Δ%Volume of the lesions decreased by 18.4%. Subgroup analysis showed that in lesions in the upper and middle lobes, only SUVpeak (p < 0.001) significantly increased by 5.6% in RRA PET, while their volume (p < 0.001) notably decreased by 12.4% (p < 0.001).</p><p><strong>Conclusion: </strong>DDG integrated with RRA motion correction algorithm can effectively mitigate motion artifacts, thus enhancing the quantification accuracy and visual quality of images in lung cancer PET/CT.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144040235","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Geometrically focused training and evaluation of organs-at-risk segmentation via deep learning.","authors":"Ruiyan Ni, Elizabeth Chuk, Kathy Han, Jennifer Croke, Anthony Fyles, Jelena Lukovic, Michael Milosevic, Benjamin Haibe-Kains, Alexandra Rink","doi":"10.1002/mp.17840","DOIUrl":"https://doi.org/10.1002/mp.17840","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Deep learning methods are promising in automating segmentation of organs at risk (OARs) in radiotherapy. However, the lack of a geometric indicator for dosimetry accuracy remains to be a problem. This issue is particularly pronounced in specific radiotherapy treatments where only the proximity of structures to the radiotherapy target affects the dose planning. In cervical cancer high dose-rate (HDR) brachytherapy, treatment planning is motivated by limiting dose to the hottest 2 cubic centimeters (D2cm&lt;sup&gt;3&lt;/sup&gt;) of the OARs. Similarly, Ethos online adaptive radiotherapy system prioritizes only the closest target structures for adaptive plan generation.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Purpose: &lt;/strong&gt;We propose a novel geometrically focused deep learning training method and evaluation metric, using cervical brachytherapy as a case study. A distance-penalized (DP) loss function was developed to focus attention on the near-to-target OAR regions. We also introduced and evaluated a novel geometric metric, weighted dice similarity coefficient (wDSC), correlated with OARs D2cm&lt;sup&gt;3&lt;/sup&gt;.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;A model was trained using a 3D U-Net architecture and 170 T2-weighted magnetic resonance (MR) images (56 patients) with clinical contours. The dataset was split into subsets at the patient level: 45 patients (150 scans) as the training set for five-fold cross-validation and 11 patients (20 scans) as the testing set. Another dataset from our institution, consisting of 35 MR scans from 22 cervical cancer patients, was used as an independent internal testing set. A distance map, emphasizing errors near high-risk clinical target volume (CTV&lt;sub&gt;HR&lt;/sub&gt;), was used to penalize two commonly used loss functions, cross-entropy (CE) loss and DiceCE loss. The wDSC emphasizes the accuracy of OAR regions proximal to CTV&lt;sub&gt;HR&lt;/sub&gt; by incorporating a weighted factor in the original vDSC. The Pearson correlation coefficient (r) was used to quantify the strength of the relationship between D2cm&lt;sup&gt;3&lt;/sup&gt; accuracy and six evaluation metrics (wDSC and five standard metrics). A physician rated and revised the auto-contours for the clinical acceptability tests.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;The wDSC moderately correlated (r = -0.55) with D2cm&lt;sup&gt;3&lt;/sup&gt; accuracy, outperforming standard geometric metrics. Models using DP loss functions consistently yielded higher wDSCs compared to their respective non-DP counterparts. DP loss models also improved D2cm&lt;sup&gt;3&lt;/sup&gt; accuracy, indicating an enhanced accuracy in dosimetry. The clinical acceptability tests revealed that more than 94% of bladder and rectum contours and approximately half of the sigmoid and small bowel contours were clinically accepted.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusion: &lt;/strong&gt;We developed and evaluated a new geometric metric, wDSC, as a better indicator of D2cm&lt;sup&gt;3&lt;/sup&gt; accuracy, which has the potential to become a surrogate for dosimetric accuracy in cervical brachytherapy","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144056867","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Enhancing CT image segmentation accuracy through ensemble loss function optimization.","authors":"Chengyin Li, Rafi Ibn Sultan, Hassan Bagher-Ebadian, Yao Qiang, Kundan Thind, Dongxiao Zhu, Indrin J Chetty","doi":"10.1002/mp.17848","DOIUrl":"https://doi.org/10.1002/mp.17848","url":null,"abstract":"<p><strong>Background: </strong>In CT-based medical image segmentation, the choice of loss function profoundly impacts the training efficacy of deep neural networks. Traditional loss functions like cross entropy (CE), Dice, Boundary, and TopK each have unique strengths and limitations, often introducing biases when used individually.</p><p><strong>Purpose: </strong>This study aims to enhance segmentation accuracy by optimizing ensemble loss functions, thereby addressing the biases and limitations of single loss functions and their linear combinations.</p><p><strong>Methods: </strong>We implemented a comprehensive evaluation of loss function combinations by integrating CE, Dice, Boundary, and TopK loss functions through both loss-level linear combination and model-level ensemble methods. Our approach utilized two state-of-the-art 3D segmentation architectures, Attention U-Net (AttUNet) and SwinUNETR, to test the impact of these methods. The study was conducted on two large CT dataset cohorts: an institutional dataset containing pelvic organ segmentations, and a public dataset consisting of multiple organ segmentations. All the models were trained from scratch with different loss settings, and performance was evaluated using Dice similarity coefficient (DSC), Hausdorff distance (HD), and average surface distance (ASD). In the ensemble approach, both static averaging and learnable dynamic weighting strategies were employed to combine the outputs of models trained with different loss functions.</p><p><strong>Results: </strong>Extensive experiments revealed the following: (1) the linear combination of loss functions achieved results comparable to those of single loss-driven methods; (2) compared to the best non-ensemble methods, ensemble-based approaches resulted in a 2%-7% increase in DSC scores, along with notable reductions in HD (e.g., a 19.1% reduction for rectum segmentation using SwinUNETR) and ASD (e.g., a 49.0% reduction for prostate segmentation using AttUNet); (3) the learnable ensemble approach with optimized weights produced finer details in predicted masks, as confirmed by qualitative analyses; and (4) the learnable ensemble consistently outperforms the static ensemble across most metrics (DSC, HD, ASD) for both AttUNet and SwinUNETR architectures.</p><p><strong>Conclusions: </strong>Our findings support the efficacy of using ensemble models with optimized weights to improve segmentation accuracy, highlighting the potential for broader applications in automated medical image analysis.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144061785","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Dual-way magnetic resonance image translation with transformer-based adversarial network.","authors":"Wenxin Li, Jun Xia, Weilin Gao, Zaiqi Hu, Shengdong Nie, Yafen Li","doi":"10.1002/mp.17837","DOIUrl":"https://doi.org/10.1002/mp.17837","url":null,"abstract":"<p><strong>Background: </strong>The magnetic resonance (MR) image translation model is designed to generate MR images of required sequence from the images of existing sequence. However, the generalization performance of MR image generation models on external datasets tends to be unsatisfactory due to the inconsistency in the data distribution of MR images across different centers or scanners.</p><p><strong>Purpose: </strong>The aim of this study is to propose a cross-sequence MR image synthesis model that could generate high-quality MR synthetic images with high transferability for small-sized external datasets.</p><p><strong>Methods: </strong>We proposed a dual-way magnetic resonance image translation model using transformer-based adversarial network (DMTrans) for MR image synthesis across sequences. It integrates a transformer-based generative architecture with an innovative discriminator design. The shifted window-based multi-head self-attention mechanism in DMTrans enables efficient capture of global and local features from MR images. The sequential dual-scale discriminator is designed to distinguish features of the generated images at multi-scale.</p><p><strong>Results: </strong>We pre-trained DMTrans model for bi-directional image synthesis on a T1/T2-weighted MR image dataset comprising 4229 slices. It demonstrates superior performance to baseline methods on both qualitative and quantitative measurements. The SSIM, PSNR, and MAE metrics for synthetic T1 images generation based on T2 images are 0.91 ± 0.04, 25.30 ± 2.40, and 24.65 ± 10.46, while the metric values are 0.90 ± 0.04, 24.72 ± 1.62, and 23.28 ± 7.40 for the opposite direction. Fine-tuning is then utilized to adapt the model to another public dataset with T1/T2/proton-weighted (PD) images, so that only 6 patients of 500 slices are required for model adaptation to achieve high-quality T1/T2, T1/PD, and T2/PD image translation results.</p><p><strong>Conclusions: </strong>The proposed DMTrans achieves the state-of-the-art performance for cross-sequence MR image conversion, which could provide more information assisting clinical diagnosis and treatment. It also offered a versatile and efficient solution to the needs of high-quality MR image synthesis in data-scarce conditions at different centers.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144061782","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Timing-synchronized passive ultrasound imaging of cavitation using eigenspace-based minimum variance beamforming and principal component analysis.","authors":"Shukuan Lu, Ruibo Su, Yingping Ma, Mingxi Wan","doi":"10.1002/mp.17853","DOIUrl":"https://doi.org/10.1002/mp.17853","url":null,"abstract":"<p><strong>Background: </strong>Passive ultrasound imaging (PUI) allows to spatially resolve cavitation triggered during ultrasound irradiation, its application in therapeutic ultrasound has been gaining attention in recent years. The diffraction mode of the imaging transducer greatly limits the PUI axial resolution, which can be improved by transmit-receive synchronization and employment of delay sum beamforming (DSB) when transmitting short pulses, however, DSB yields poor performance in resolution and anti-interference.</p><p><strong>Purpose: </strong>Inspired by adaptive beamforming and its low-complexity algorithm in active imaging field, this paper aims to develop an improved timing-synchronized PUI (TSPUI) algorithm for detection of short-pulse transmission-induced cavitation.</p><p><strong>Methods: </strong>The passive array data collected by timing synchronization is processed by minimum variance beamforming (MVB), whose weights are optimized by projection on the eigendecomposed signal subspace, that is, eigenspace-based MVB (EMVB), with the sum of the flight times on the transmitting and receiving paths as the delay. Applying principal component analysis (PCA) on the pre-collected MVB weight samples, a conversion matrix is constructed to allow the matrix inversion and eigendecomposition involved in weight calculation to be performed in a low dimension. The algorithm performance is confirmed by experiments, where a high-intensity focused ultrasound transducer and a linear-array transducer configured in a common parallel or vertical manner are employed for cavitation induction and cavitation imaging, and evaluated with the established indicators.</p><p><strong>Results: </strong>Reducing the eigenvalue threshold coefficient allows more sidelobes to be removed, and choosing an appropriate principal component number can reduce the time cost while guaranteeing the reconstruction quality. EMVB-PCA provides high resolution and anti-interference performance relative to DSB, with a reduction of over 60% in the point spread area and over 14 dB in the sidelobe and noise level, meanwhile, its time cost is considerably lower than EMVB, with a reduction of over 80%. Additionally, constructing the conversion matrix by simulation is feasible and valid, providing convenience for real imaging.</p><p><strong>Conclusions: </strong>EMVB-PCA allows for high-quality TSPUI reconstruction of cavitation at a fast rate, providing an effective tool for detecting short-duration cavitation and further benefiting short-pulse therapeutic ultrasound applications.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144065561","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Hepatocellular carcinoma ¹⁸F-FDG PET/CT kinetic parameter estimation based on the advantage actor-critic algorithm.","authors":"Jianfeng He, Siming Li, Yiwei Xiong, Yu Yao, Siyu Wang, Sidan Wang, Shaobo Wang","doi":"10.1002/mp.17851","DOIUrl":"https://doi.org/10.1002/mp.17851","url":null,"abstract":"<p><strong>Background: </strong>Kinetic parameters estimated with dynamic ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) positron emission tomography (PET)/computed tomography (CT) help characterize hepatocellular carcinoma (HCC), and deep reinforcement learning (DRL) can improve kinetic parameter estimation.</p><p><strong>Purpose: </strong>The advantage actor-critic (A2C) algorithm is a DRL algorithm with neural networks that seek the optimal parameters. The aim of this study was to preliminarily assess the role of the A2C algorithm in estimating the kinetic parameters of ¹⁸F-FDG PET/CT in patients with HCC.</p><p><strong>Materials and methods: </strong>¹⁸F-FDG PET data from 14 liver tissues and 17 HCC tumors obtained via a previously developed, abbreviated acquisition protocol (5-min dynamic PET/CT imaging supplemented with 1-min static imaging at 60 min) were prospectively collected. The A2C algorithm was used to estimate kinetic parameters with a reversible double-input, three-compartment model, and the results were compared with those of the conventional nonlinear least squares (NLLS) algorithm. Fitting errors were compared via the root-mean-square errors (RMSEs) of the time activity curves (TACs).</p><p><strong>Results: </strong>Significant differences in K₁, k₂, k₃, k₄, f_a, and v_b according to the A2C algorithm and k₃, f_a, and v_b according to the NLLS algorithm were detected between HCC and normal liver tissues (all p < 0.05). Furthermore, A2C demonstrated superior diagnostic performance over NLLS in terms of k₃ and v_b (both p < 0.05 in the Delong test). Notably, A2C yielded a smaller fitting error for normal liver tissue (0.62 ± 0.24 vs. 1.04 ± 1.00) and HCC tissue (1.40 ± 0.42 vs. 1.51 ± 0.97) than did NLLS.</p><p><strong>Conclusions: </strong>Compared with the conventional postreconstruction NLLS method, the A2C algorithm can more precisely estimate ¹⁸F-FDG kinetic parameters with a reversible double-input, three-compartment model for HCC tumors, attaining better TAC fitting with a lower RMSE.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144056242","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Predicting [177Lu]Lu-DOTA-TATE dosimetry by using pre-therapy [68Ga]Ga-DOTA-TATE PET/CT and biomarkers in patient with neuroendocrine tumors.","authors":"Hongxing Yang, Ming Qi, Zhihao Chen, Fei Liu, Junyan Xu, Xiaoping Xu, Qing Kong, Jianping Zhang, Shaoli Song","doi":"10.1002/mp.17852","DOIUrl":"https://doi.org/10.1002/mp.17852","url":null,"abstract":"<p><strong>Background: </strong>Lutetium-177 DOTA-TATE peptide receptor radionuclide therapy (PRRT) is an established and effective treatment modality for patients with metastatic neuroendocrine tumors (NETs).</p><p><strong>Purpose: </strong>This study aims to predict patient-absorbed doses from [177Lu]Lu-DOTA-TATE PRRT in the liver, kidney and lesion by utilizing patient-specific absorbed doses from pre-therapeutic [68Ga]Ga-DOTA-TATE PET/CT.</p><p><strong>Methods: </strong>Before the treatment of cycle 1, 11 patients with NETs underwent PET/CT scans at 0.5, 1.0, 2.0 and 4.0 h after the injection of [68Ga]Ga-DOTA-TATE. Patients then received [177Lu]Lu-DOTA-TATE PRRT and underwent SPECT/CT scans at 4, 24, 96, and 168 h post-administration. The segmentations and dosimetry were performed by using a professional software. The linear regression model used the absorbed doses from [68Ga]Ga-DOTA-TATE alone as the predictor variable. The multiple linear regression model used the absorbed doses from [68Ga]Ga-DOTA-TATE and the relevant clinical biomarkers as the predictor variables.</p><p><strong>Results: </strong>The mean absorbed doses from [177Lu]Lu-DOTA-TATE PRRT in kidney and liver were 4.1 and 2.1 Gy, respectively. In comparison, the mean absorbed doses from [68Ga]Ga-DOTA-TATE were significantly lower: 18.0 mGy and 11.0 mGy, respectively. For lesions, the maximum absorbed dose from [68Ga]Ga-DOTA-TATE ranged from 24.1 to 170.4 mGy, while the maximum absorbed dose from [177Lu]Lu-DOTA-TATE PRRT was significantly higher, ranging from 9.6 to 77.9 Gy. The linear regression model yielded moderate R-squared values of 0.50, 0.59, and 0.36 for kidney, liver and lesion, respectively. The performance of multiple linear regression model was better, with R-squared values increasing to 0.81, 0.77, and 0.84.</p><p><strong>Conclusion: </strong>Absorbed doses from [177Lu]Lu-DOTA-TATE PRRT can be accurately predicted. Moreover, our models are formalized into simple equations.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144003902","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multi-modality multiorgan image segmentation using continual learning with enhanced hard attention to the task.","authors":"Ming-Long Wu, Yi-Fan Peng","doi":"10.1002/mp.17842","DOIUrl":"https://doi.org/10.1002/mp.17842","url":null,"abstract":"<p><strong>Background: </strong>Enabling a deep neural network (DNN) to learn multiple tasks using the concept of continual learning potentially better mimics human brain functions. However, current continual learning studies for medical image segmentation are mostly limited to single-modality images at identical anatomical locations.</p><p><strong>Purpose: </strong>To propose and evaluate a continual learning method termed eHAT (enhanced hard attention to the task) for performing multi-modality, multiorgan segmentation tasks using a DNN.</p><p><strong>Methods: </strong>Four public datasets covering the lumbar spine, heart, and brain acquired by magnetic resonance imaging (MRI) and computed tomography (CT) were included to segment the vertebral bodies, the right ventricle, and brain tumors, respectively. Three-task (spine CT, heart MRI, and brain MRI) and four-task (spine CT, heart MRI, brain MRI, and spine MRI) models were tested for eHAT, with the three-task results compared with state-of-the-art continual learning methods. The effectiveness of multitask performance was measured using the forgetting rate, defined as the average difference in Dice coefficients and Hausdorff distances between multiple-task and single-task models. The ability to transfer knowledge to different tasks was evaluated using backward transfer (BWT).</p><p><strong>Results: </strong>The forgetting rates were -2.51% to -0.60% for the three-task eHAT models with varying task orders, substantially better than the -18.13% to -3.59% using original hard attention to the task (HAT), while those in four-task models were -2.54% to -1.59%. In addition, four-task U-net models with eHAT using only half the number of channels (1/4 parameters) yielded nearly equal performance with or without regularization. A retrospective model comparison showed that eHAT with fixed or automatic regularization had significantly superior BWT (-3% to 0%) compared to HAT (-22% to -4%).</p><p><strong>Conclusion: </strong>We demonstrate for the first time that eHAT effectively achieves continual learning of multi-modality, multiorgan segmentation tasks using a single DNN, with improved forgetting rates compared with HAT.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144048767","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Leveraging prior knowledge in machine intelligence to improve lesion diagnosis for early cancer detection.","authors":"Zhengrong J Liang, Shaojie Chang, Yongfeng Gao, Weiguo Cao, Licheng R Kuo, Marc J Pomeroy, Lihong C Li, Almas F Abbasi, Jela Bandovic, Michael J Reiter, Perry J Pickhardt","doi":"10.1002/mp.17841","DOIUrl":"https://doi.org/10.1002/mp.17841","url":null,"abstract":"<p><strong>Background: </strong>Experts' interpretations of medical images for lesion diagnosis may not always align with the underlying in vivo tissue pathology and, therefore, cannot be considered the definitive truth regarding malignancy or benignity. While current machine learning (ML) models in medical imaging can replicate expert interpretations, their results may also diverge from the actual ground truth.</p><p><strong>Purpose: </strong>This study investigates various factors contributing to these discrepancies and proposes solutions.</p><p><strong>Methods: </strong>The central idea of the proposed solution is to integrate prior knowledge into ML models to enhance the characterization of in vivo tissues. The incorporation of prior knowledge into decision-making is task-specific, tailored to the data acquired for that task. This central idea was tested on the diagnosis of lesions using low dose computed tomography (LdCT) for early cancer detection, particularly focusing on more challenging, ambiguous or indeterminate lesions (IDLs) as classified by experts. One key piece of prior knowledge involves CT x-ray energy spectrum, where different energies interact with in vivo tissues within a lesion, producing variable but reproducible image contrasts that encapsulate biological information. Typically, CT imaging devices use only the high-energy portion of this spectrum for data acquisition; however, this study considers the full spectrum for lesion diagnostics. Another critical aspect of prior knowledge includes the functional or dynamic properties of in vivo tissues, such as elasticity, which can indicate pathological conditions. Instead of relying solely on abstract image features as current ML models do, this study extracts these tissue pathological characteristics from the image contrast variations.</p><p><strong>Results: </strong>The method was tested on LdCT images of four sets of IDLs, including pulmonary nodules and colorectal polyps, with pathological reports serving as the ground truth for malignancy or benignity. The method achieved an area under the receiver operating characteristic curve (AUC) of 0.98 ± 0.03, demonstrating a significant improvement over existing state-of-the-art ML models, which typically have AUCs in the 0.70 range.</p><p><strong>Conclusion: </strong>Leveraging prior knowledge in machine intelligence can enhance lesion diagnosis, resolve the ambiguity of IDLs interpreted by experts, and improve the effectiveness of LdCT screening for early-stage cancers.</p>","PeriodicalId":94136,"journal":{"name":"Medical physics","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144056006","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}