Search

Abstract Purpose Volumetric assessment of the liver regularly yields discrepant results between pre- and intraoperatively determined volumes. Nevertheless, the main factor responsible for this discrepancy remains still unclear. The aim of this study was to systematically determine the difference between in vivo CT-volumetry and ex vivo volumetry in a pig animal model. Material and Methods Eleven pigs were studied. Liver density assessment, CT-volumetry and water displacement volumetry was performed after surgical removal of the complete liver. Known possible errors of volume determination like resection or segmentation borders were eliminated in this model. Regression analysis was performed and differences between CT-volumetry and water displacement determined. Results Median liver density was 1.07 g/ml. Regression analysis showed a high correlation of r2 = 0.985 between CT-volumetry and water displacement. CTvolumetry was found to be 13% higher than water displacement volumetry (p < 0.0001). Conclusion In this study the only relevant factor leading to the difference between in vivo CT-volumetry and ex vivo water displacement volumetry seems to be blood perfusion of the liver. The systematic difference of 13 percent has to be taken in account when dealing with those measures.

Purpose To develop and evaluate a publicly available deep learning model for segmenting and classifying cardiac implantable electronic devices (CIEDs) on Digital Imaging and Communications in Medicine (DICOM) and smartphone-based chest radiographs. Materials and Methods This institutional review board-approved retrospective study included patients with implantable pacemakers, cardioverter defibrillators, cardiac resynchronization therapy devices, and cardiac monitors who underwent chest radiography between January 2012 and January 2022. A U-Net model with a ResNet-50 backbone was created to classify CIEDs on DICOM and smartphone images. Using 2321 chest radiographs in 897 patients (median age, 76 years [range, 18-96 years]; 625 male, 272 female), CIEDs were categorized into four manufacturers, 27 models, and one "other" category. Five smartphones were used to acquire 11 072 images. Performance was reported using the Dice coefficient on the validation set for segmentation or balanced accuracy on the test set for manufacturer and model classification, respectively. Results The segmentation tool achieved a mean Dice coefficient of 0.936 (IQR: 0.890-0.958). The model had an accuracy of 94.36% (95% CI: 90.93%, 96.84%; 251 of 266) for CIED manufacturer classification and 84.21% (95% CI: 79.31%, 88.30%; 224 of 266) for CIED model classification. Conclusion The proposed deep learning model, trained on both traditional DICOM and smartphone images, showed high accuracy for segmentation and classification of CIEDs on chest radiographs. Keywords: Conventional Radiography, Segmentation Supplemental material is available for this article . © RSNA, 2024 See also the commentary by Júdice de Mattos Farina and Celi in this issue.

Competing Interests: Disclosures Dr Barbieri declares research grant support from Abbott Laboratories and Boston Scientific, speaker honoraria from Edwards Lifesciences, and consulting fees from Boston Scientific. Dr Dreger reports research support from Abbott Laboratories and speaker honoraria from Abbott Laboratories and Edwards Lifesciences. The other authors report no conflicts.

Objectives: The aim of the study was to investigate computed tomography-based thermography (CTT) for ablation zone prediction in microwave ablation (MWA)., Methods: CTT was investigated during MWA in an in vivo porcine liver. For CTT, serial volume scans were acquired every 30 s during ablations and every 60 s immediately after MWA. After the procedure, contrast-enhanced computed tomography (CECT) was performed. After euthanasia, the liver was removed for sampling and further examination. Color-coded CTT maps were created for visualization of ablation zones, which were compared with both CECT and macroscopy. Average CT attenuation values in Hounsfield units (HU) were statistically correlated with temperatures using Spearman's correlation coefficient. CTT was retrospectively evaluated in one patient who underwent radiofrequency ablation (RFA) treatment of renal cell carcinoma., Results: A significant correlation between HU and temperature was found with r =  - 0.77 (95% confidence interval (CI), - 0.89 to - 0.57) and p < 0.001. Linear regression yielded a slope of - 1.96 HU/°C (95% CI, - 2.66 to - 1.26). Color-coded CTT maps provided superior visualization of ablation zones., Conclusion: Our results show that CTT allows visualization of the ablation area and measurement of its size and is feasible in patients, encouraging further exploration in a clinical setting., Critical Relevance Statement: CT-based thermography research software allows visualization of the ablation zone and is feasible in patients, encouraging further exploration in a clinical setting to assess risk reduction of local recurrence., (© 2023. The Author(s).)

Computed tomography (CT)-based Thermography (CTT) is currently being investigated as a non-invasive temperature monitoring method during ablation procedures. Since multiple CT scans with defined time intervals were acquired during this procedure, interscan motion artifacts can occur between the images, so registration is required. The aim of this study was to investigate different registration algorithms and their combinations for minimizing inter-scan motion artifacts during thermal ablation. Four CTT datasets were acquired using microwave ablation (MWA) of normal liver tissue performed in an in vivo porcine model. During each ablation, spectral CT volume scans were sequentially acquired. Based on initial reconstructions, rigid or elastic registration, or a combination of these, were carried out and rated by 15 radiologists. Friedman's test was used to compare rating results in reader assessments and revealed significant differences for the ablation probe movement rating only ( p = 0.006; range, 5.3-6.6 points). Regarding this parameter, readers assessed rigid registration as inferior to other registrations. Quantitative analysis of ablation probe movement yielded a significantly decreased distance for combined registration as compared with unregistered data. In this study, registration was found to have the greatest influence on ablation probe movement, with connected registration being superior to only one registration process.

Background and Objectives: Bedside chest radiographs (CXRs) are challenging to interpret but important for monitoring cardiothoracic disease and invasive therapy devices in critical care and emergency medicine. Taking surrounding anatomy into account is likely to improve the diagnostic accuracy of artificial intelligence and bring its performance closer to that of a radiologist. Therefore, we aimed to develop a deep convolutional neural network for efficient automatic anatomy segmentation of bedside CXRs., Methods: To improve the efficiency of the segmentation process, we introduced a "human-in-the-loop" segmentation workflow with an active learning approach, looking at five major anatomical structures in the chest (heart, lungs, mediastinum, trachea, and clavicles). This allowed us to decrease the time needed for segmentation by 32% and select the most complex cases to utilize human expert annotators efficiently. After annotation of 2,000 CXRs from different Level 1 medical centers at Charité - University Hospital Berlin, there was no relevant improvement in model performance, and the annotation process was stopped. A 5-layer U-ResNet was trained for 150 epochs using a combined soft Dice similarity coefficient (DSC) and cross-entropy as a loss function. DSC, Jaccard index (JI), Hausdorff distance (HD) in mm, and average symmetric surface distance (ASSD) in mm were used to assess model performance. External validation was performed using an independent external test dataset from Aachen University Hospital (n = 20)., Results: The final training, validation, and testing dataset consisted of 1900/50/50 segmentation masks for each anatomical structure. Our model achieved a mean DSC/JI/HD/ASSD of 0.93/0.88/32.1/5.8 for the lung, 0.92/0.86/21.65/4.85 for the mediastinum, 0.91/0.84/11.83/1.35 for the clavicles, 0.9/0.85/9.6/2.19 for the trachea, and 0.88/0.8/31.74/8.73 for the heart. Validation using the external dataset showed an overall robust performance of our algorithm., Conclusions: Using an efficient computer-aided segmentation method with active learning, our anatomy-based model achieves comparable performance to state-of-the-art approaches. Instead of only segmenting the non-overlapping portions of the organs, as previous studies did, a closer approximation to actual anatomy is achieved by segmenting along the natural anatomical borders. This novel anatomy approach could be useful for developing pathology models for accurate and quantifiable diagnosis., Competing Interests: Declaration of Competing Interest SMN declares a relationship (honorary speaker) with the following companies: Canon Medical Systems, Bracco Imaging, and Teleflex. All other authors declare no conflict of interest., (Copyright © 2023 Elsevier B.V. All rights reserved.)

Purpose: Preprocedural computed tomography (CT) imaging before transcatheter aortic valve implantation/replacement (TAVI/TAVR) requires high diagnostic accuracy without motion artifacts. The aim of this retrospective study is to compare the image quality of a high-pitch non-electrocardiography (ECG)-gated CT protocol used in patients with atrial tachyarrhythmias with a prospectively ECG-gated CT protocol used in patients with sinus rhythm., Materials and Methods: We retrospectively included 108 patients who underwent preprocedural CT imaging before TAVI/TAVR. 52 patients with sinus rhythm were imaged using a prospectively ECG-gated protocol (Group A), and 56 patients with atrial tachyarrhythmias were imaged using the high-pitch non-ECG-gated protocol (Group B). Image quality was rated subjectively by two experienced radiologists and assessed by objective parameters including radiation dose, image noise, contrast-to-noise ratio (CNR), and signal-to-noise ratio (SNR) at the levels of the aortic root and abdominal aorta., Results: Subjective image quality was equally good with both CT protocols, and interrater agreement was substantial in both groups but tended to be higher in Group B at the level of the aortic root (Group A: κw = 0.644, Group B: κw = 0.741). With the high-pitch non-ECG-gated CT protocol, image noise was significantly increased (p = 0.001), whereas the SNR, CNR, and radiation dose were significantly decreased (p = 0.002, p = 0.003, and p < 0.001, respectively) at the level of the aortic root compared to the prospectively ECG-gated CT protocol., Conclusion: The high-pitch non-ECG-gated protocol yields images with similar subjective image quality compared with the prospectively ECG-gated CT protocol and allows motion-free assessment of the aortic root for accurate TAVI/TAVR planning. The high-pitch non-ECG-gated protocol may be used as an alternative for preprocedural CT imaging in patients with atrial tachyarrhythmias., Key Points: · In patients with atrial tachyarrhythmias, a high-pitch non-ECG-gated CT protocol achieves similar subjective image quality compared to a prospective ECG-gated CT protocol.. · At the level of the aortic root, image noise is significantly increased, whereas SNR and CNR are significantly decreased using the high-pitch non-ECG-gated protocol.. · Radiation dose is reduced by 55 % using the high-pitch non-ECG-gated protocol.., Citation Format: · Shnayien S, Beetz N, Bressem KK et al. Comparison of a High-Pitch Non-ECG-Gated and a Prospective ECG-Gated Protocol for Preprocedural Computed Tomography Imaging Before TAVI/TAVR. Fortschr Röntgenstr 2023; 195: 139 - 147., Competing Interests: S. Shnayien: nothing to disclose.N.L. Beetz: nothing to disclose.K.K. Bressem nothing to disclose.B. Hamm: grant money from Abbott, AbbVie, Ablative Solutions, Accovion, Achaogen, Actelion Pharmaceuticals, ADIR, Advanced Sleep Research, Aesculap, AGO, Alexion Pharmaceuticals, Amgen, AO Foundation, Arbeitsgemeinschaft Internistische Onkologie, Arena Pharmaceuticals, ARMO BioSciences, art photonics, Astellas, AstraZeneca, B. Braun, BARD, Bayer, Berlin-Brandenburger Centrum für Regenerative Therapien, Berliner Krebsgesellschaft, BIOTRONIK, Bioven, Boehringer Ingelheim, Boston Biomedical, Bracco, BrainsGate, Bristol-Myers Squibb, Bundesministerium für Bildung und Forschung, Canon Medical Systems, Cascadian Therapeutics, Celgene, CellAct Pharma, Celldex Therapeutics, CeloNova Bio-Sciences, Charité Research Organisation, Chiltern, Clovis Oncology, Corvia Medical, Covance, Cubist Pharmaceuticals, Curevac, Curis, Daiichi Sankyō, DC Devices, Delcath Systems, Demira, DFG, Deutsche Krebshilfe, Deutsche Rheuma-Liga, Deutsche Stiftung für Herzforschung, DSM Nutritional Products, Dynavax Technologies, Eli Lilly and Company, Eisai, EORTC, Epizyme, Essex Pharma, EU Funding Programmes, Euroscreen, FIBREX Medical, Focused Ultrasound Surgery Foundation, Fraunhofer-Gesellschaft, Galena Biopharma, Galmed Pharmaceuticals, Ganymed Pharmaceuticals, GETNE, Genentech, Gilead Sciences, GlaxoSmithKline, Glycotope, Goethe-Universität Frankfurt am Main, Guerbet, Guidant Europe, Halozyme, Hewlett Packard, ICON, Idera Pharmaceuticals, Ignyta, Immunocore, Immunomedics, INC Research, Incyte, Innate Pharma, INSIGHTEC, inVentiv Health, iOMEDICO, Ionis Pharmaceuticals, Ipsen Pharma, IQVIA, ISA Pharmaceutical, ITM Solucin, Janssen Pharmaceutical, Kantar Health, Kartos Therapeutics, Karyopharm Therapeutics, Klinische Forschung Berlin-Mitte, Kite Pharma, lnspireMD, Land Berlin, Lion Biotechnology, LMU München, Lombard Medical, Loxo Oncology, LSK BioPartners, Lundbeck, Lux Biosciences, LYSARC, MacroGenics, MagForce, MedImmune, Medpace, MedPass International, Medtronic, Merck, Merrimack Pharmaceuticals, MeVis Medical Solutions, Millennium Pharmaceuticals, Mologen, MorphoSys, Monika Kutzner Stiftung Berlin, MSD, Neovacs, NewLink Genetics, Nexus Oncology, NIH, Novartis, Novocure, Nuvaira, Nuvisan, Ockham Oncology, OHIRC, Orion Corporation, Parexel, Perceptive, Pfizer, Pharmacyclics, PharmaCept, Pharmacyclics, PharmaMar, Philips, PIQUR Therapeutics, Pluristem Therapeutics, PPD, PneumRX, Portola Pharmaceuticals, PRA Health Sciences, Premier-Research, Provectus Biopharmaceuticals, PSI CRO, Pulmonx International, Quintiles, Regeneron Pharmaceuticals, Respicardia, Roche, Samsung, Sanofis, Seattle Genetics, Servier, SGS Life Science Services, Shire Human Genetic Therapies, Siemens, Silence Therapeutics, Spectranetics, Spectrum Pharmaceuticals, St. Jude Medical, Symphogen, Taiho Pharmaceutical, TauRx, Terumo Medical Corporation, Tesaro, TETEC, Teva, Theorem Clinical Research, Theradex, Threshold Pharmaceuticals, TNS Healthcare, UCB, Vertex Pharmaceuticals, Winicker Norimed, Wyeth Pharma, Xcovery and Zukunftsfond Berlin. The funding had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. S.M. Niehues: grant money from Bayer, Canon Medical Systems, Guerbet, and Teleflex Medical. The funding had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript., (Thieme. All rights reserved.)

Gleamer BoneView © is a commercially available AI algorithm for fracture detection in radiographs. We aim to test if the algorithm can assist in better sensitivity and specificity for fracture detection by residents with prospective integration into clinical workflow. Radiographs with inquiry for fracture initially reviewed by two residents were randomly assigned and included. A preliminary diagnosis of a possible fracture was made. Thereafter, the AI decision on presence and location of possible fractures was shown and changes to diagnosis could be made. Final diagnosis of fracture was made by a board-certified radiologist with over eight years of experience, or if available, cross-sectional imaging. Sensitivity and specificity of the human report, AI diagnosis, and assisted report were calculated in comparison to the final expert diagnosis. 1163 exams in 735 patients were included, with a total of 367 fractures (31.56%). Pure human sensitivity was 84.74%, and AI sensitivity was 86.92%. Thirty-five changes were made after showing AI results, 33 of which resulted in the correct diagnosis, resulting in 25 additionally found fractures. This resulted in a sensitivity of 91.28% for the assisted report. Specificity was 97.11, 84.67, and 97.36%, respectively. AI assistance showed an increase in sensitivity for both residents, without a loss of specificity.

Background: Computed tomography is a standard imaging procedure for the detection of liver lesions, such as metastases, which can often be small and poorly contrasted, and therefore hard to detect. Advances in image reconstruction have shown promise in reducing image noise and improving low-contrast detectability., Purpose: To examine a novel, specialized, model-based iterative reconstruction (MBIR) technique for improved low-contrast liver lesion detection., Material and Methods: Patient images with reported poorly contrasted focal liver lesions were retrospectively reconstructed with the low-contrast attenuating algorithm (FIRST-LCD) from primary raw data. Liver-to-lesion contrast, signal-to-noise, and contrast-to-noise ratios for background and liver noise for each lesion were compared for all three FIRST-LCD presets with the established hybrid iterative reconstruction method (AIDR-3D). An additional visual conspicuity score was given by two experienced radiologists for each lesion., Results: A total of 82 lesions in 57 examinations were included in the analysis. All three FIRST-LCD algorithms provided statistically significant increases in liver-to-lesion contrast, with FIRST MILD showing the largest increase (40.47 HU in AIDR-3D; 45.84 HU in FIRST MILD ; P  < 0.001). Substantial improvement was shown in contrast-to-noise metrics. Visual analysis of the lesions shows decreased lesion visibility with all FIRST methods in comparison to AIDR-3D, with FIRST STR showing the closest results ( P  < 0.001)., Conclusion: Objective image metrics show promise for MBIR methods in improving the detectability of low-contrast liver lesions; however, subjective image quality may be perceived as inferior. Further improvements are necessary to enhance image quality and lesion detection.

Background. The impact of vascular cooling effects in hepatic microwave ablation (MWA) is controversially discussed. The objective of this study was a systematic assessment of vascular cooling effects in hepatic MWA ex vivo. Methods. Microwave ablations were performed in fresh porcine liver ex vivo with a temperature-controlled MWA generator (902-928 MHz) and a non-cooled 14-G-antenna. Energy input was set to 9.0 kJ. Hepatic vessels were simulated by glass tubes. Three different vessel diameters (3.0, 5.0, 8.0 mm) and vessel to antenna distances (5, 10, 20 mm) were examined. Vessels were perfused with saline solution at nine different flow rates (0-500 mL/min). Vascular cooling effects were assessed at the largest cross-sectional ablation area. A quantitative and semi-quantitative/morphologic analysis was carried out. Results. 228 ablations were performed. Vascular cooling effects were observed at close (5 mm) and medium (10 mm) antenna to vessel distances ( P < .05). Vascular cooling effects occurred around vessels with flow rates ≥1.0 mL/min ( P < .05) and a vessel diameter ≥3 mm ( P < .05). Higher flow rates did not result in more distinct cooling effects ( P > .05). No cooling effects were measured at large (20 mm) antenna to vessel distances ( P > .05). Conclusion. Vascular cooling effects occur in hepatic MWA and should be considered in treatment planning. The vascular cooling effect was mainly affected by antenna to vessel distance. Vessel diameter and vascular flow rate played a minor role in vascular cooling effects.

Background MRI is frequently used for early diagnosis of axial spondyloarthritis (axSpA). However, evaluation is time-consuming and requires profound expertise because noninflammatory degenerative changes can mimic axSpA, and early signs may therefore be missed. Deep neural networks could function as assistance for axSpA detection. Purpose To create a deep neural network to detect MRI changes in sacroiliac joints indicative of axSpA. Materials and Methods This retrospective multicenter study included MRI examinations of five cohorts of patients with clinical suspicion of axSpA collected at university and community hospitals between January 2006 and September 2020. Data from four cohorts were used as the training set, and data from one cohort as the external test set. Each MRI examination in the training and test sets was scored by six and seven raters, respectively, for inflammatory changes (bone marrow edema, enthesitis) and structural changes (erosions, sclerosis). A deep learning tool to detect changes indicative of axSpA was developed. First, a neural network to homogenize the images, then a classification network were trained. Performance was evaluated with use of area under the receiver operating characteristic curve (AUC), sensitivity, and specificity. P < .05 was considered indicative of statistically significant difference. Results Overall, 593 patients (mean age, 37 years ± 11 [SD]; 302 women) were studied. Inflammatory and structural changes were found in 197 of 477 patients (41%) and 244 of 477 (51%), respectively, in the training set and 25 of 116 patients (22%) and 26 of 116 (22%) in the test set. The AUCs were 0.94 (95% CI: 0.84, 0.97) for all inflammatory changes, 0.88 (95% CI: 0.80, 0.95) for inflammatory changes fulfilling the Assessment of SpondyloArthritis international Society definition, and 0.89 (95% CI: 0.81, 0.96) for structural changes indicative of axSpA. Sensitivity and specificity on the external test set were 22 of 25 patients (88%) and 65 of 91 patients (71%), respectively, for inflammatory changes and 22 of 26 patients (85%) and 70 of 90 patients (78%) for structural changes. Conclusion Deep neural networks can detect inflammatory or structural changes to the sacroiliac joint indicative of axial spondyloarthritis at MRI. © RSNA, 2022 Online supplemental material is available for this article.

In the present work, we present a publicly available, expert-segmented representative dataset of 158 3.0 Tesla biparametric MRIs [1]. There is an increasing number of studies investigating prostate and prostate carcinoma segmentation using deep learning (DL) with 3D architectures [2], [3], [4], [5], [6], [7]. The development of robust and data-driven DL models for prostate segmentation and assessment is currently limited by the availability of openly available expert-annotated datasets [8], [9], [10]. The dataset contains 3.0 Tesla MRI images of the prostate of patients with suspected prostate cancer. Patients over 50 years of age who had a 3.0 Tesla MRI scan of the prostate that met PI-RADS version 2.1 technical standards were included. All patients received a subsequent biopsy or surgery so that the MRI diagnosis could be verified/matched with the histopathologic diagnosis. For patients who had undergone multiple MRIs, the last MRI, which was less than six months before biopsy/surgery, was included. All patients were examined at a German university hospital (Charité Universitätsmedizin Berlin) between 02/2016 and 01/2020. All MRI were acquired with two 3.0 Tesla MRI scanners (Siemens VIDA and Skyra, Siemens Healthineers, Erlangen, Germany). Axial T2W sequences and axial diffusion-weighted sequences (DWI) with apparent diffusion coefficient maps (ADC) were included in the data set. T2W sequences and ADC maps were annotated by two board-certified radiologists with 6 and 8 years of experience, respectively. For T2W sequences, the central gland (central zone and transitional zone) and peripheral zone were segmented. If areas of suspected prostate cancer (PIRADS score of ≥ 4) were identified on examination, they were segmented in both the T2W sequences and ADC maps. Because restricted diffusion is best seen in DWI images with high b-values, only these images were selected and all images with low b-values were discarded. Data were then anonymized and converted to NIfTI (Neuroimaging Informatics Technology Initiative) format., Competing Interests: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (© 2022 The Author(s).)

For computed tomography (CT), representing the diagnostic standard for trauma patients, image quality is essential. The positioning of the patient’s arms next to the abdomen causes artifacts and is also considered to increase radiation exposure. The aim of this study was to evaluate the effect of various positionings during different CT examination steps on the extent of artifacts as well as radiation dose using iterative reconstruction (IR). 354 trauma-CTs were analyzed retrospectively. All datasets were reconstructed using IR and three different examination protocols were applied. Arm elevation led to a significant improvement of the image quality across all examination protocols (p < 0.001). Variation in arm positioning during image acquisition did not lead to a reduction of radiation dose (p = 0.123). Only elevation during scout acquisition resulted in the reduction of radiation exposure (p < 0.001). To receive high-quality CT images, patients should be placed with elevated arms for the trunk scan, as artifacts remain even with the IR. Arm repositioning during the examination itself had no effect on the applied radiation dose because its modulation refers to the initial scout obtained. In order to achieve a dose effect by different positioning, a two-scout protocol (dual scout) should be used.

The aim of this study was a three-dimensional analysis of vascular cooling effects on microwave ablation (MWA) in an ex vivo porcine model. A glass tube, placed in parallel to the microwave antenna at distances of 2.5, 5.0 and 10.0 mm (A-V distance), simulated a natural liver vessel. Seven flow rates (0, 1, 2, 5, 10, 100, 500 ml/min) were evaluated. Ablations were segmented into 2 mm slices for a 3D-reconstruction. A qualitative and quantitative analysis was performed. 126 experiments were carried out. Cooling effects occurred in all test series with flow rates ≥ 2 ml/min in the ablation periphery. These cooling effects had no impact on the total ablation volume (p > 0.05) but led to changes in ablation shape at A-V distances of 5.0 mm and 10.0 mm. Contrary, at a A-V distance of 2.5 mm only flow rates of ≥ 10 ml/min led to relevant cooling effects in the ablation centre. These cooling effects influenced the ablation shape, whereas the total ablation volume was reduced only at a maximal flow rate of 500 ml/min (p = 0.002). Relevant cooling effects exist in MWA. They mainly depend on the distance of the vessel to the ablation centre., (© 2022. The Author(s).)

A wide range of deep learning (DL) architectures with varying depths are available, with developers usually choosing one or a few of them for their specific task in a nonsystematic way. Benchmarking (i.e., the systematic comparison of state-of-the art architectures on a specific task) may provide guidance in the model development process and may allow developers to make better decisions. However, comprehensive benchmarking has not been performed in dentistry yet. We aimed to benchmark a range of architecture designs for 1 specific, exemplary case: tooth structure segmentation on dental bitewing radiographs. We built 72 models for tooth structure (enamel, dentin, pulp, fillings, crowns) segmentation by combining 6 different DL network architectures (U-Net, U-Net++, Feature Pyramid Networks, LinkNet, Pyramid Scene Parsing Network, Mask Attention Network) with 12 encoders from 3 different encoder families (ResNet, VGG, DenseNet) of varying depth (e.g., VGG13, VGG16, VGG19). On each model design, 3 initialization strategies (ImageNet, CheXpert, random initialization) were applied, resulting overall into 216 trained models, which were trained up to 200 epochs with the Adam optimizer (learning rate = 0.0001) and a batch size of 32. Our data set consisted of 1,625 human-annotated dental bitewing radiographs. We used a 5-fold cross-validation scheme and quantified model performances primarily by the F1-score. Initialization with ImageNet or CheXpert weights significantly outperformed random initialization ( P < 0.05). Deeper and more complex models did not necessarily perform better than less complex alternatives. VGG-based models were more robust across model configurations, while more complex models (e.g., from the ResNet family) achieved peak performances. In conclusion, initializing models with pretrained weights may be recommended when training models for dental radiographic analysis. Less complex model architectures may be competitive alternatives if computational resources and training time are restricting factors. Models developed and found superior on nondental data sets may not show this behavior for dental domain-specific tasks.