Figure - available from: European Journal of Hybrid Imaging
This content is subject to copyright. Terms and conditions apply.
Representative slices through the image quality phantom produced for a PET scanner using (a) Anger positioning logic and (b) deep learning-based algorithm. Reprinted with permission from MDPI (Sanaat and Zaidi 2020)
Source publication
This brief review summarizes the major applications of artificial intelligence (AI), in particular deep learning approaches, in molecular imaging and radiation therapy research. To this end, the applications of artificial intelligence in five generic fields of molecular imaging and radiation therapy, including PET instrumentation design, PET image...
Citations
... While AI holds transformative potential, challenges exist in its clinical implementation, including technical, ethical, and legal domains, such as patient data privacy. The limited transparency of the entire system raises ethical questions and warrants careful consideration from various perspectives [5,6]. ...
... The roles and responsibilities of professionals, including a core team led by radiation oncology and medical physics experts, need to be clearly defined. New training for professionals should cover organizing services with integrated AI tools, selecting and implementing AI applications, defining input for AI, and evaluating output competently [6]. ...
... Common nuclear image acquisition technologies are Single Photon Emission Computed Tomography (SPECT) and Positron Emission Tomography (PET) [7]. AI applications in the medical imaging sector can be categorized into several primary fields including image pre-and post-processing, image reconstruction, image denoising, estimation of full-count SPECT, and SPECT Attenuation Correction (AC) [6,8]. ...
In the realm of resource allocation within the healthcare sector, numerous challenges exist that necessitate the automation of these processes for effective management [1].The increasing demand for automation in healthcare is driven by a surge in patients, a need for enhancements in quality, such as early identification and tailored treatments, and the increasing burden on healthcare professionals like doctors and nurses. Artificial Intelligence (AI) is already demonstrating its potential in screening routines, rivaling human performance in tasks such as breast cancer screening. The integration of AI within the radiology departments has yielded notable advancements. This integration not only enhances resource utilization but also offers opportunities for further optimization, particularly in the realm of nuclear medical applications. The synergy between AI and radiology stands to significantly streamline operations related to radioactive isotopes, thereby enhancing overall efficiency and efficacy in clinical settings. The synergy between AI and radiology stands to significantly streamline operations related to radioactive isotopes, thereby enhancing overall efficiency and efficacy in clinical settings [2, 3]. The utilization of Artificial intelligence especially machine/deep learning into different aspects of nuclear medical applications advances exponentially, showcasing their potential roles in physics and clinical tasks including radiotherapy, medical imaging, radiopharmacy, and disease theranostics. These four domains are outlined briefly as follows. A) Radiotherapy, also called radiation oncology or therapeutic radiology, employs nuclear or ionizing radiation to damage and eradicate cancerous tumor cells. AI systems hold the capability to streamline various aspects of the intricate radiotherapy process including segmentation, planning, prediction of dose delivery, integration with radiomics, measuring radiation dosage, utilizing computer-aided detection, and forecasting outcomes. Nevertheless, viewing artificial and swarm intelligence as a "black box" raises concerns, as human operators might comprehend only the input, output predictions, and applying it to clinical practice poses challenges [4].
... Deep learning (DL) models are increasingly used for semi-automated and automated segmentation in PSMA PET/CT [19][20][21][22][23][24]. However, accurate segmentation remains a significant challenge due to various factors. ...
Background
Prostate-specific membrane antigen (PSMA) PET/CT imaging is widely used for quantitative image analysis, especially in radioligand therapy (RLT) for metastatic castration-resistant prostate cancer (mCRPC). Unknown features influencing PSMA biodistribution can be explored by analyzing segmented organs at risk (OAR) and lesions. Manual segmentation is time-consuming and labor-intensive, so automated segmentation methods are desirable. Training deep-learning segmentation models is challenging due to the scarcity of high-quality annotated images. Addressing this, we developed shifted windows UNEt TRansformers (Swin UNETR) for fully automated segmentation. Within a self-supervised framework, the model’s encoder was pre-trained on unlabeled data. The entire model was fine-tuned, including its decoder, using labeled data.
Methods
In this work, 752 whole-body [ ⁶⁸ Ga]Ga-PSMA-11 PET/CT images were collected from two centers. For self-supervised model pre-training, 652 unlabeled images were employed. The remaining 100 images were manually labeled for supervised training. In the supervised training phase, 5-fold cross-validation was used with 64 images for model training and 16 for validation, from one center. For testing, 20 hold-out images, evenly distributed between two centers, were used. Image segmentation and quantification metrics were evaluated on the test set compared to the ground-truth segmentation conducted by a nuclear medicine physician.
Results
The model generates high-quality OARs and lesion segmentation in lesion-positive cases, including mCRPC. The results show that self-supervised pre-training significantly improved the average dice similarity coefficient (DSC) for all classes by about 3%. Compared to nnU-Net, a well-established model in medical image segmentation, our approach outperformed with a 5% higher DSC. This improvement was attributed to our model’s combined use of self-supervised pre-training and supervised fine-tuning, specifically when applied to PET/CT input. Our best model had the lowest DSC for lesions at 0.68 and the highest for liver at 0.95.
Conclusions
We developed a state-of-the-art neural network using self-supervised pre-training on whole-body [ ⁶⁸ Ga]Ga-PSMA-11 PET/CT images, followed by fine-tuning on a limited set of annotated images. The model generates high-quality OARs and lesion segmentation for PSMA image analysis. The generalizable model holds potential for various clinical applications, including enhanced RLT and patient-specific internal dosimetry.
... (Yörük Kurtaran/Support Foundation for Civil Society, Çağdaş Özbakan/Genç Foundation, Özgür Mehmet Kütküt) AI is recognized as a tool for organizations to reduce costs, improve service quality, coordination, productivity and implementation efficiency (Vijayakumar, 2023). Such uses can make it possible to redeploy human resources to higher value tasks (Fukumura et al., 2021;Arabi and Zaidi, 2020). Again, the integration of AI in businesses fosters collaboration between human-machine teams and can contribute to the learning process in the workplace by assisting employees with work-related tasks (Mirbabaie et al., 2021;Wilkens, 2020). ...
This report is based on the "Roundtable on Artificial Intelligence for Social Good" that took
place on December 5, 2023 at Istanbul Bilgi University and was prepared as part of the
Artificial Intelligence Literacy Project supported by Google Turkey. The meeting focused on
how AI can be used for social good and discussed the potential and needs in Turkey.
Participants included digital communication and technology experts, civil society
representatives and academics. The aim was to understand the societal benefits of AI, explore
how we can better utilize this technology and identify opportunities for collaboration.
... Qualitative and quantitative evaluations of this AC method demonstrated a reliable performance and thus, the deep-learning AC methods today, have to be considered as the technical state-of-the-art in AC in PET/MR in specific body regions like head and pelvis. [36][37][38][39][40][41] Nevertheless, the deep-learning-based AC suffers from some limitations with regard to accuracy and speed. 14 Especially in the context of whole-body PET/MRI there is a limited number of deep-learning methods for the generation of AC-maps available. ...
Background
Attenuation correction (AC) is an important methodical step in positron emission tomography/magnetic resonance imaging (PET/MRI) to correct for attenuated and scattered PET photons.
Purpose
The overall quality of magnetic resonance (MR)‐based AC in whole‐body PET/MRI was evaluated in direct comparison to computed tomography (CT)‐based AC serving as reference. The quantitative impact of isolated tissue classes in the MR‐AC was systematically investigated to identify potential optimization needs and strategies.
Methods
Data of n = 60 whole‐body PET/CT patients with normal lung tissue and without metal implants/prostheses were used to generate six different AC‐models based on the CT data for each patient, simulating variations of MR‐AC. The original continuous CT‐AC (CT‐org) is referred to as reference. A pseudo MR‐AC (CT‐mrac), generated from CT data, with four tissue classes and a bone atlas represents the MR‐AC. Relative difference in linear attenuation coefficients (LAC) and standardized uptake values were calculated. From the results two improvements regarding soft tissue AC and lung AC were proposed and evaluated.
Results
The overall performance of MR‐AC is in good agreement compared to CT‐AC. Lungs, heart, and bone tissue were identified as the regions with most deviation to the CT‐AC (myocardium −15%, bone tissue −14%, and lungs ±20%). Using single‐valued LACs for AC in the lung only provides limited accuracy. For improved soft tissue AC, splitting the combined soft tissue class into muscles and organs each with adapted LAC could reduce the deviations to the CT‐AC to < ±1%. For improved lung AC, applying a gradient LAC in the lungs could remarkably reduce over‐ or undercorrections in PET signal compared to CT‐AC (±5%).
Conclusions
The AC is important to ensure best PET image quality and accurate PET quantification for diagnostics and radiotherapy planning. The optimized segment‐based AC proposed in this study, which was evaluated on PET/CT data, inherently reduces quantification bias in normal lung tissue and soft tissue compared to the CT‐AC reference.
... Radiomic features can identify patterns and provide additional information not perceptible by the human eyes [38]. In this regard, therapy response assessment, restaging, segmentation, and dose prediction were studied, as presented in the following section. ...
... Therefore, manual segmentation is not a practical solution in clinical practice because it takes time and effort. In this regard, several authors developed semi-automatic or automatic segmentation methods [38]. The semi-automatic segmentation strategy uses [ 68 Ga]Ga-PSMA-SSTR and PSMA PET/CT imaging biomarkers. ...
... Automated tumor segmentation has emerged in most research areas to address all the shortcomings of manual or semiautomatic segmentation procedures [38]. A review by Brosch-Lenz et al. [124] discussed prominent and emerging segmentation methods, and their possible applications in the RPT dosimetry. ...
Radiotheranostics refers to pairing of radioactive imaging biomarkers with radioactive therapeutic compounds that deliver ionizing radiation. Given the introduction of very promising radio-pharmaceuticals, the radiotheranostics approach is creating a novel paradigm in personalized, targeted radionuclide therapies (TRTs), also known as radiopharmaceuticals (RPTs). Radiother-apeutic pairs targeting somatostatin receptors (SSTR) and prostate-specific membrane antigens (PSMA) are increasingly used to diagnose and treat patients with metastatic neuroendocrine tumors (NETs) and prostate cancer. In parallel, radiomics and artificial intelligence (AI) as important areas in quantitative image analysis are paving the way for significantly enhanced workflows in diagnostic and theranostic fields, from data and image processing to clinical decision support, improving patient selection, personalized treatment strategies, response prediction, and prognostication. Furthermore, AI has the potential for tremendous effectiveness in patient dosimetry which copes with complex and time-consuming tasks in the RPT workflow. The present work provides a comprehensive overview of radiomics and AI application in radio-theranostics, focusing on pairs of SSTR- or PSMA‑targeting radioligands, describing the funda-mental concepts and specific imaging/treatment features. Our review includes ligands radio-labeled by 68Ga, 18F, 177Lu, 64Cu, 90Y, and 225Ac. Specifically, contributions by radiomics and AI towards improved image acquisition, reconstruction, treatment response, segmentation, restag-ing, lesion classification, dose prediction, and estimation as well as ongoing developments and future directions are discussed.
... It should be noted that owing to the astonishing performance of artificial intelligence-based algorithms, in particular deep learning methods, novel approaches for performing attenuation and scatter correction (ASC) on PET Fig. 3 Outcome of a survey from 15 experts in the field of PET instrumentation who provided feedback on six main questions about dedicated PET scanners and future developments data without using anatomical images have been developed [101][102][103]. These include ASC in the image domain [104], attenuation correction factor estimation in the sinogram domain [104], hybrid MLAA and deep learning methods [105], and attenuation map estimation from non-ASC PET images [106]. ...
We focus on reviewing state-of-the-art developments of dedicated PET scanners with irregular geometries and the potential of different aspects of multifunctional PET imaging. First, we discuss advances in non-conventional PET detector geometries. Then, we present innovative designs of organ-specific dedicated PET scanners for breast, brain, prostate, and cardiac imaging. We will also review challenges and possible artifacts by image reconstruction algorithms for PET scanners with irregular geometries, such as non-cylindrical and partial angular coverage geometries and how they can be addressed. Then, we attempt to address some open issues about cost/benefits analysis of dedicated PET scanners, how far are the theoretical conceptual designs from the market/clinic, and strategies to reduce fabrication cost without compromising performance.
... Effortstoenhance 68 Ga-PSMimages focus on re ning reconstruction algorithms, particular lyaddressing scattered radiation [14,16]. Standard evaluation methods typically involve expert-guided image reconstruction as a reference, often involving manual parameter adjustments [17,18].On occasion, Monte Carlo simulation is employed to generate accurate data, although its time-intensive nature prevents widespread use [19].A burgeoning eld in the context is the use of arti cial intelligence, particularly deep learning techniques, in PET imaging and the correction of PET data for attenuation and scatter [7,15,[20][21][22]. In the case of image domain attenuation and scatter correction, deep learning models are fed PET non-AC images to predict PET AC images [23,24]. ...
... The quantitative accuracy of the developed deep learning model was assessed using the clinical data of 92 patients regarding the CT-based PET AC as a reference. The standard quanti cation metrics such as mean error (ME), mean absolute error (MAE), root mean square error (RMSE), and relative error (RE%) in terms of standard uptake value (SUV) were calculated to quantify the accuracy of the model [18,21].In addition, the peak signal-to-noise ratio (PSNR) and structural similarity index (SSI) were calculated between PET-DLAC and PET-CT AC images [15,27]. ...
Objective: This study aims to demonstrate the feasibility and benefits of using a deep learning-based approach for attenuation correction in 68-Ga-PSMA whole-body PET scans.
Materials & Methods: A dataset comprising 700 patients (a mean age: 67.6±5.9 years old, range: 45-85 years) with prostate cancer who underwent 68-Ga-PSMA PET/CT examinations was collected. A deep learning model was trained on 700 whole-body68-Ga-PSMA clinical images to perform attenuation correction (AC) in the image domain. To assess the quantitative accuracy of the developed deep learning model, clinical data from 92 patients were used as a reference for CT-based PET AC (PET-CTAC). Standard quantification metrics, including mean error (ME), mean absolute error (MAE), and root mean square error (RMSE) were calculated in terms of standard uptake value (SUV) to gauge the accuracy of the model. For clinical evaluation, three specialists conducted a blinded assessment of synthesized PET images’ quality in terms of lesion detectability across 50 clinical subjects, comparing them with PET-CTAC images.
Results: Quantitative analysis of the deep learning AC (DLAC) model revealed ME, MAE, and RMSE values of -0.007±0.032, 0.08±0.033, and 0.252±125 (SUV), respectively. Additionally, regarding lesion detection analysis, the deep learning model demonstrated superior image quality for 16 subjects out of 50 compared to the PET-CT AC images. In 56% of cases, PET-DLAC and PET-CTAC images exhibited closely comparable image quality and lesion delectability.
Conclusion: This study emphasizes the significant improvement in image quality and lesion detection capabilities achieved through the integration of deep learning-based attenuation correction in 68-Ga-PSMA PET imaging. This innovation not only provides a compelling solution to the challenges posed by bladder radioactivity but also a promising way to minimize patient radiation exposure through the coordinated integration of low-dose CT and deep learning-based AC, while simultaneously increasing the image quality.
... The broad adoption of artificial intelligence (AI) in medicine in recent years also opens up the possibility of advancing the integration of routine patient-specific dosimetry into clinical workflows in multiple ways, e. g., by assisting in accelerating image acquisition, by offering a high level of automatization (e. g., regarding image segmentation and registration), or by improving accuracy and speed of the absorbed dose calculation process itself [14,15,16]. Therefore, the purpose of this review is to highlight recent advances in the application of AI to internal dosimetry for RPT. ...
Routine clinical dosimetry along with radiopharmaceutical therapies is key for future treatment personalization. However, dosimetry is considered complex and time-consuming with various challenges amongst the required steps within the dosimetry workflow. The general workflow for image-based dosimetry consists of quantitative imaging, the segmentation of organs and tumors, fitting of the time-activity-curves, and the conversion to absorbed dose. This work reviews the potential and advantages of the use of artificial intelligence to improve speed and accuracy of every single step of the dosimetry workflow.
... [29][30][31] Accurate dose prediction is important for optimizing clinical plans, and AI has proven to be an effective tool for predicting therapy response and guiding tailored therapy planning in the future. [32][33][34] ML and deep learning (DL) models have started being used to address the challenge of personalized pretreatment planning in nuclear medicine, with promising results. 21,35,36 Furthermore, researchers have investigated the predictive usefulness of SSTR-expressing tumor agonists labeled with several radioisotopes, including 177 Lu-DOTATATE. ...
Background
Standardized patient‐specific pretreatment dosimetry planning is mandatory in the modern era of nuclear molecular radiotherapy, which may eventually lead to improvements in the final therapeutic outcome. Only a comprehensive definition of a dosage therapeutic window encompassing the range of absorbed doses, that is, helpful without being detrimental can lead to therapy individualization and improved outcomes. As a result, setting absorbed dose safety limits for organs at risk (OARs) requires knowledge of the absorbed dose–effect relationship. Data sets of consistent and reliable inter‐center dosimetry findings are required to characterize this relationship.
Purpose
We developed and standardized a new pretreatment planning model consisting of a predictive dosimetry procedure for OARs in patients with neuroendocrine tumors (NETs) treated with ¹⁷⁷Lu‐DOTATATE (Lutathera). In the retrospective study described herein, we used machine learning (ML) regression algorithms to predict absorbed doses in OARs by exploiting a combination of radiomic and dosiomic features extracted from patients’ imaging data.
Methods
Pretreatment and posttreatment data for 20 patients with NETs treated with ¹⁷⁷Lu‐DOTATATE were collected from two clinical centers. A total of 3412 radiomic and dosiomic features were extracted from the patients’ computed tomography (CT) scans and dose maps, respectively. All dose maps were generated using Monte Carlo simulations. An ML regression model was designed based on ML algorithms for predicting the absorbed dose in every OAR (liver, left kidney, right kidney, and spleen) before and after the therapy and between each therapy session, thus predicting any possible radiotoxic effects.
Results
We evaluated nine ML regression algorithms. Our predictive model achieved a mean absolute dose error (MAE, in Gy) of 0.61 for the liver, 1.58 for the spleen, 1.30 for the left kidney, and 1.35 for the right kidney between pretherapy ⁶⁸Ga‐DOTATOC positron emission tomography (PET)/CT and posttherapy ¹⁷⁷Lu‐DOTATATE single photon emission (SPECT)/CT scans. Τhe best predictive performance observed was based on the gradient boost for the liver, the left kidney and the right kidney, and on the extra tree regressor for the spleen. Evaluation of the model's performance according to its ability to predict the absorbed dose in each OAR in every possible combination of pretherapy ⁶⁸Ga‐DOTATOC PET/CT and any posttherapy ¹⁷⁷Lu‐DOTATATE treatment cycle SPECT/CT scans as well as any ¹⁷⁷Lu‐DOTATATE SPECT/CT treatment cycle and the consequent ¹⁷⁷Lu‐DOTATATE SPECT/CT treatment cycle revealed mean absorbed dose differences ranges from −0.55 to 0.68 Gy. Incorporating radiodosiomics features from the ⁶⁸Ga‐DOTATOC PET/CT and first ¹⁷⁷Lu‐DOTATATE SPECT/CT treatment cycle scans further improved the precision and minimized the standard deviation of the predictions in nine out of 12 instances. An average improvement of 57.34% was observed (range: 17.53%–96.12%). However, it's important to note that in three instances (i.e., Ga,C.1 → C3 in spleen and left kidney, and Ga,C.1 → C2 in right kidney) we did not observe an improvement (absolute differences of 0.17, 0.08, and 0.05 Gy, respectively). Wavelet‐based features proved to have high correlated predictive value, whereas non‐linear‐based ML regression algorithms proved to be more capable than the linear‐based of producing precise prediction in our case.
Conclusions
The combination of radiomics and dosiomics has potential utility for personalized molecular radiotherapy (PMR) response evaluation and OAR dose prediction. These radiodosiomic features can potentially provide information on any possible disease recurrence and may be highly useful in clinical decision‐making, especially regarding dose escalation issues.
... In recent years, deep learning has outperformed traditional methods in various medical image analyses [11][12][13][14]. Traditional image denoising methods such as Gaussian, nonlocal-mean filtering [15,16], and block-matching and 3D filtering (BM3D) [17] approaches cause significant signal loss. ...
Purpose
The axial field of view (AFOV) of a positron emission tomography (PET) scanner greatly affects the quality of PET images. Although a total-body PET scanner (uEXPLORER) with a large AFOV is more sensitive, it is more expensive and difficult to widely use. Therefore, we attempt to utilize high-quality images generated by uEXPLORER to optimize the quality of images from short-axis PET scanners through deep learning technology while controlling costs.
Methods
The experiments were conducted using PET images of three anatomical locations (brain, lung, and abdomen) from 335 patients. To simulate PET images from different axes, two protocols were used to obtain PET image pairs (each patient was scanned once). For low-quality PET (LQ-PET) images with a 320-mm AFOV, we applied a 300-mm FOV for brain reconstruction and a 500-mm FOV for lung and abdomen reconstruction. For high-quality PET (HQ-PET) images, we applied a 1940-mm AFOV during the reconstruction process. A 3D Unet was utilized to learn the mapping relationship between LQ-PET and HQ-PET images. In addition, the peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM) were employed to evaluate the model performance. Furthermore, two nuclear medicine doctors evaluated the image quality based on clinical readings.
Results
The generated PET images of the brain, lung, and abdomen were quantitatively and qualitatively compatible with the HQ-PET images. In particular, our method achieved PSNR values of 35.41 ± 5.45 dB (p < 0.05), 33.77 ± 6.18 dB (p < 0.05), and 38.58 ± 7.28 dB (p < 0.05) for the three beds. The overall mean SSIM was greater than 0.94 for all patients who underwent testing. Moreover, the total subjective quality levels of the generated PET images for three beds were 3.74 ± 0.74, 3.69 ± 0.81, and 3.42 ± 0.99 (the highest possible score was 5, and the minimum score was 1) from two experienced nuclear medicine experts. Additionally, we evaluated the distribution of quantitative standard uptake values (SUV) in the region of interest (ROI). Both the SUV distribution and the peaks of the profile show that our results are consistent with the HQ-PET images, proving the superiority of our approach.
Conclusion
The findings demonstrate the potential of the proposed technique for improving the image quality of a PET scanner with a 320 mm or even shorter AFOV. Furthermore, this study explored the potential of utilizing uEXPLORER to achieve improved short-axis PET image quality at a limited economic cost, and computer-aided diagnosis systems that are related can help patients and radiologists.
