Acetylsalicylic acid use and development of cardiac allograft vasculopathy: A national prospective study using highly automated 3-D optical coherence tomography analysis.

BACKGROUND: There is conflicting evidence on the role of acetylsalicylic acid (ASA) use in the development of cardiac allograft vasculopathy (CAV). METHODS: A nationwide prospective two-center study investigated changes in the coronary artery vasculature by highly automated 3-D optical coherence tomography (OCT) analysis at 1 month and 12 months after heart transplant (HTx). The influence of ASA use on coronary artery microvascular changes was analyzed in the overall study cohort and after propensity score matching for selected clinical CAV risk factors. RESULTS: In total, 175 patients (mean age 52 ± 12 years, 79% male) were recruited. During the 1-year follow-up, both intimal and media thickness progressed, with ASA having no effect on its progression. However, detailed OCT analysis revealed that ASA use was associated with a lower increase in lipid plaque (LP) burden (p = .013), while it did not affect the other observed pathologies. Propensity score matching of 120 patients (60 patient pairs) showed similar results, with ASA use associated with lower progression of LPs (p = .002), while having no impact on layered fibrotic plaque (p = .224), calcification (p = .231), macrophage infiltration (p = .197), or the absolute coronary artery risk score (p = .277). According to Kaplan-Meier analysis, ASA use was not associated with a significant difference in survival (p = .699) CONCLUSION: This study showed a benefit of early ASA use after HTx on LP progression. However, ASA use did not have any impact on the progression of other OCT-observed pathologies or long-term survival.

Sensitivity of CNN image analysis to multifaceted measurements of neurite growth.

Quantitative analysis of neurite growth and morphology is essential for understanding the determinants of neural development and regeneration, however, it is complicated by the labor-intensive process of measuring diverse parameters of neurite outgrowth. Consequently, automated approaches have been developed to study neurite morphology in a high-throughput and comprehensive manner. These approaches include computer-automated algorithms known as 'convolutional neural networks' (CNNs)-powerful models capable of learning complex tasks without the biases of hand-crafted models. Nevertheless, their complexity often relegates them to functioning as 'black boxes.' Therefore, research in the field of explainable AI is imperative to comprehend the relationship between CNN image analysis output and predefined morphological parameters of neurite growth in order to assess the applicability of these machine learning approaches. In this study, drawing inspiration from the field of automated feature selection, we investigate the correlation between quantified metrics of neurite morphology and the image analysis results from NeuriteNet-a CNN developed to analyze neurite growth. NeuriteNet accurately distinguishes images of neurite growth based on different treatment groups within two separate experimental systems. These systems differentiate between neurons cultured on different substrate conditions and neurons subjected to drug treatment inhibiting neurite outgrowth. By examining the model's function and patterns of activation underlying its classification decisions, we discover that NeuriteNet focuses on aspects of neuron morphology that represent quantifiable metrics distinguishing these groups. Additionally, it incorporates factors that are not encompassed by neuron morphology tracing analyses. NeuriteNet presents a novel tool ideally suited for screening morphological differences in heterogeneous neuron groups while also providing impetus for targeted follow-up studies.

AI-based classification of three common malignant tumors in neuro-oncology: A multi-institutional comparison of machine learning and deep learning methods.

PURPOSE: To determine if machine learning (ML) or deep learning (DL) pipelines perform better in AI-based three-class classification of glioblastoma (GBM), intracranial metastatic disease (IMD) and primary CNS lymphoma (PCNSL). METHODOLOGY: Retrospective analysis included 502 cases for training (208 GBM, 67 PCNSL and 227 IMD), with external validation on 86 cases (27:27:32). Multiparametric MRI images (T1W, T2W, FLAIR, DWI and T1-CE) were co-registered, resampled, denoised and intensity normalized, followed by semiautomatic 3D segmentation of the enhancing tumor (ET) and peritumoral region (PTR). Model performance was assessed using several ML pipelines and 3D-convolutional neural networks (3D-CNN) using sequence specific masks, as well as combination of masks. All pipelines were trained and evaluated with 5-fold nested cross-validation on internal data followed by external validation using multi-class AUC. RESULTS: Two ML models achieved similar performance on test set, one using T2-ET and T2-PTR masks (AUC: 0.885, 95% CI: [0.816, 0.935] and another using T1-CE-ET and FLAIR-PTR mask (AUC: 0.878, CI: [0.804, 0.930]). The best performing DL models achieved an AUC of 0.854, (CI [0.774, 0.914]) on external data using T1-CE-ET and T2-PTR masks, followed by model derived from T1-CE-ET, ADC-ET and FLAIR-PTR masks (AUC: 0.851, CI [0.772, 0.909]). CONCLUSION: Both ML and DL derived pipelines achieved similar performance. T1-CE mask was used in three of the top four overall models. Additionally, all four models had some mask derived from PTR, either T2WI or FLAIR.

Role of genetics in the development of cardiac allograft vasculopathy.

BACKGROUND: The association between genetic polymorphisms and early cardiac allograft vasculopathy (CAV) development is relatively unexplored. Identification of genes involved in the CAV process may offer new insights into pathophysiology and lead to a wider range of therapeutic options. METHODS: This prospective study of 109 patients investigated 44 single nucleotide polymorphisms (SNPs) within the susceptibility loci potentially related to coronary artery disease, carotid artery intima-media thickness (cIMT), and in nitric oxide synthase gene. Genotyping was done by the Fluidigm SNP Type assays and Fluidigm 48.48 Dynamic Array IFC. The intima thickness progression (IT) was evaluated by coronary optical coherence tomography performed 1 month and 12 months after heart transplantation (HTx). RESULTS: During the first post-HTx year, the mean intima thickness (IT) increased by 24.0 ± 34.2 µm (p < 0.001) and lumen area decreased by ‒0.9 ± 1.8 mm2 (p < 0.001). The rs1570360 (A/G) SNP of the vascular endothelial growth factor A (VEGFA) gene showed the strongest association with intima thickness progression, even in the presence of the traditional CAV risk factors. SNPs previously related to carotid artery intima-media thickness rs11785239 (PRAG1), rs6584389 (PAX2), rs13225723 (LINC02577) and rs17477177 (CCDC71L), were among the five most significantly associated with IT progression but lost their significance once traditional CAV risk factors had been added. CONCLUSION: Results of this study suggest that genetic variability may play an important role in CAV development. The vascular endothelial growth factor A gene SNP rs1570360 showed the strongest association with intima thickness (IT) progression measured by OCT, even in the presence of the traditional CAV risk factors (Tab. 3, Fig. 3, Ref. 36). Text in PDF www.elis.sk Keywords: cardiac allograft vasculopathy, optical coherence tomography, vascular endothelial growth factor A, intimal thickening, genetic polymorphism.

Computed Tomography Perfusion-Based Prediction of Core Infarct and Tissue at Risk: Can Artificial Intelligence Help Reduce Radiation Exposure?

BACKGROUND AND PURPOSE: We explored the feasibility of automated, arterial input function independent, vendor neutral prediction of core infarct, and penumbral tissue using complete and partial computed tomographic perfusion data sets through neural networks. METHODS: Using retrospective computed tomographic perfusion data from 57 patients, split as training/validation (60%/40%), we developed and validated separate 2-dimensional U-net models for cerebral blood flow (CBF) and time to maximum (Tmax) maps calculation to predict core infarct and tissue at risk, respectively. Once trained, the full sets of 28 input images were sequentially reduced to equitemporal 14, 10, and 7 time points. The averaged structural similarity index measure between the model-derived images and ground truth perfusion maps was compared. Volumes for core infarct and Tmax were compared using the Pearson correlation coefficient. RESULTS: Both CBF and Tmax maps derived using 28 and 14 time points had similar structural similarity index measure (0.80-0.81; P>0.05) when compared with ground truth images. The Pearson correlation for the CBF and Tmax volumes derived from the model using 28-tp with ground truth volumes derived from the RAPID software was 0.69 for CBF and 0.74 for Tmax. The predicted maps were fully concordant in terms of laterality to the commercial perfusion maps. The mean Dice scores were 0.54 for the core infarct and 0.63 for the hypoperfusion maps. CONCLUSIONS: Artificial intelligence model-derived volumes show good correlation with RAPID-derived volumes for CBF and Tmax. Within the constraints of a small sample size, the perfusion map quality is similar when using 14-tp instead of 28-tp. Our findings provide proof of concept that vendor neutral artificial intelligence models for computed tomographic perfusion processing using complete or partial image data sets appear feasible. The model accuracy could be further optimized using larger data sets.

Quantitative muscle MRI as a sensitive marker of early muscle pathology in myotonic dystrophy type 1.

BACKGROUND: Quantitative muscle MRI as a sensitive marker of early muscle pathology and disease progression in adult-onset myotonic dystrophy type 1. The utility of muscle MRI as a marker of muscle pathology and disease progression in adult-onset myotonic dystrophy type 1 (DM1) was evaluated. METHODS: This prospective, longitudinal study included 67 observations from 36 DM1 patients (50% female), and 92 observations from 49 healthy adults (49% female). Lower-leg 3T magnetic resonance imaging (MRI) scans were acquired. Volume and fat fraction (FF) were estimated using a three-point Dixon method, and T2-relaxometry was determined using a multi-echo spin-echo sequence. Muscles were segmented automatically. Mixed linear models were conducted to determine group differences across muscles and image modality, accounting for age, sex, and repeated observations. Differences in rate of change in volume, T2-relaxometry, and FF were also determined with mixed linear regression that included a group by elapsed time interaction. RESULTS: Compared with healthy adults, DM1 patients exhibited reduced volume of the tibialis anterior, soleus, and gastrocnemius (GAS) (all, P < .05). T2-relaxometry and FF were increased across all calf muscles in DM1 compared to controls. (all, P < .01). Signs of muscle pathology, including reduced volume, and increased T2-relaxometry and FF were already noted in DM1 patients who did not exhibit clinical motor symptoms of DM1. As a group, DM1 patients exhibited a more rapid change than did controls in tibialis posterior volume (P = .05) and GAS T2-relaxometry (P = .03) and FF (P = .06). CONCLUSIONS: Muscle MRI renders sensitive, early markers of muscle pathology and disease progression in DM1. T2 relaxometry may be particularly sensitive to early muscle changes related to DM1.

Radiomics-based differentiation between glioblastoma and primary central nervous system lymphoma: a comparison of diagnostic performance across different MRI sequences and machine learning techniques.

OBJECTIVES: Despite the robust diagnostic performance of MRI-based radiomic features for differentiating between glioblastoma (GBM) and primary central nervous system lymphoma (PCNSL) reported on prior studies, the best sequence or a combination of sequences and model performance across various machine learning pipelines remain undefined. Herein, we compare the diagnostic performance of multiple radiomics-based models to differentiate GBM from PCNSL. METHODS: Our retrospective study included 94 patients (34 with PCNSL and 60 with GBM). Model performance was assessed using various MRI sequences across 45 possible model and feature selection combinations for nine different sequence permutations. Predictive performance was assessed using fivefold repeated cross-validation with five repeats. The best and worst performing models were compared to assess differences in performance. RESULTS: The predictive performance, both using individual and a combination of sequences, was fairly robust across multiple top performing models (AUC: 0.961-0.977) but did show considerable variation between the best and worst performing models. The top performing individual sequences had comparable performance to multiparametric models. The best prediction model in our study used a combination of ADC, FLAIR, and T1-CE achieving the highest AUC of 0.977, while the second ranked model used T1-CE and ADC, achieving a cross-validated AUC of 0.975. CONCLUSION: Radiomics-based predictive accuracy can vary considerably, based on the model and feature selection methods as well as the combination of sequences used. Also, models derived from limited sequences show performance comparable to those derived from all five sequences. KEY POINTS: • Radiomics-based diagnostic performance of various machine learning models for differentiating glioblastoma and PCNSL varies considerably. • ML models using limited or multiple MRI sequences can provide comparable performance, based on the chosen model. • Embedded feature selection models perform better than models using a priori feature reduction.

Heart rate and early progression of cardiac allograft vasculopathy: A prospective study using highly automated 3-D optical coherence tomography analysis.

INTRODUCTION: Heart rate slowing agents are frequently prescribed to manage heart transplant (HTx) patients with the assumption that higher heart rate is a risk factor in cardiovascular disease. PATIENTS AND METHODS: This prospective two-center study investigated early progression of cardiac allograft vasculopathy (CAV) in 116 HTx patients. Examinations by coronary optical coherence tomography and 24-hour ambulatory ECG monitoring were performed both at baseline (1 month after HTx) and during follow-up (12 months after HTx). RESULTS: During the first post-HTx year, we observed a significant reduction in the mean coronary luminal area from 9.0 ± 2.5 to 8.0 ± 2.4 mm2 (P < .001), and progression in mean intimal thickness (IT) from 106.5 ± 40.4 to 130.1 ± 53.0 µm (P < .001). No significant relationship was observed between baseline and follow-up mean heart rates and IT progression (R = .02, P = .83; R = -.13, P = .18). We found a mild inverse association between beta-blocker dosage at 12 months and IT progression (R = -.20, P = .035). CONCLUSION: Our study did not confirm a direct association between mean heart rate and progression of CAV. The role of beta blockers warrants further investigation, with our results indicating that they may play a protective role in early CAV development.

Retinal neurodegeneration may precede microvascular changes characteristic of diabetic retinopathy in diabetes mellitus.

Diabetic retinopathy (DR) has long been recognized as a microvasculopathy, but retinal diabetic neuropathy (RDN), characterized by inner retinal neurodegeneration, also occurs in people with diabetes mellitus (DM). We report that in 45 people with DM and no to minimal DR there was significant, progressive loss of the nerve fiber layer (NFL) (0.25 µm/y) and the ganglion cell (GC)/inner plexiform layer (0.29 µm/y) on optical coherence tomography analysis (OCT) over a 4-y period, independent of glycated hemoglobin, age, and sex. The NFL was significantly thinner (17.3 µm) in the eyes of six donors with DM than in the eyes of six similarly aged control donors (30.4 µm), although retinal capillary density did not differ in the two groups. We confirmed significant, progressive inner retinal thinning in streptozotocin-induced "type 1" and B6.BKS(D)-Lepr(db)/J "type 2" diabetic mouse models on OCT; immunohistochemistry in type 1 mice showed GC loss but no difference in pericyte density or acellular capillaries. The results suggest that RDN may precede the established clinical and morphometric vascular changes caused by DM and represent a paradigm shift in our understanding of ocular diabetic complications.

Plaque volume and plaque risk profile in diabetic vs. non-diabetic patients undergoing lipid-lowering therapy: a study based on 3D intravascular ultrasound and virtual histology.

BACKGROUND: Coronary atherosclerosis progresses faster in patients with diabetes mellitus (DM) and causes higher morbidity and mortality in such patients compared to non-diabetics ones (non-DM). We quantify changes in plaque volume and plaque phenotype during lipid-lowering therapy in DM versus non-DM patients using advanced intracoronary imaging. METHODS: We analyzed data from 61 patients with stable angina pectoris included to the PREDICT trial searching for prediction of plaque changes during intensive lipid-lowering therapy (40 mg rosuvastatin daily). Geometrically correct, fully 3-D representation of the vascular wall surfaces and intravascular ultrasound virtual histology (IVUS-VH) defined tissue characterization was obtained via fusion of two-plane angiography and IVUS-VH. Frame-based indices of plaque morphology and virtual histology analyses were computed and averaged in 5 mm long baseline/follow-up registered vessel segments covering the entire length of the two sequential pullbacks (baseline, 1-year). We analyzed 698 5-mm-long segments and calculated the Liverpool active plaque score (LAPS). RESULTS: Despite reaching similar levels of LDL cholesterol (DM 2.12 ± 0.91 mmol/l, non-DM 1.8 ± 0.66 mmol/l, p = 0.21), DM patients experienced, compared to non-DM ones, higher progression of mean plaque area (0.47 ± 1.15 mm2 vs. 0.21 ± 0.97, p = 0.001), percent atheroma volume (0.7 ± 2.8% vs. - 1.4 ± 2.5%, p = 0.007), increase of LAPS (0.23 ± 1.66 vs. 0.13 ± 1.79, p = 0.018), and exhibited more locations with TCFA (Thin-Cap Fibro-Atheroma) plaque phenotype in 5 mm vessel segments (20.3% vs. 12.5%, p = 0.01). However, only non-DM patients reached significant decrease of LDL cholesterol. Plaque changes were more pronounced in PIT (pathologic intimal thickening) compared to TCFA with increased plaque area in both phenotypes in DM patients. CONCLUSION: Based on detailed 3D analysis, we found advanced plaque phenotype and further atherosclerosis progression in DM patients despite the same reached levels of LDLc as in non-DM patients. Trial registration ClinicalTrials.gov identifier: NCT01773512.

The accuracy of detailed analysis of optical coherence tomography in detection of plaque lipid content: dual-imaging study with optical coherence tomography and near-infrared spectroscopy.

BACKGROUND: Lipid-rich plaque covered by a thin fibrous cap (FC) has been identified as a frequent morphological substrate for the development of acute coronary syndrome. Optical coherence tomography (OCT) permits the identification and measurement of the FC. Near-infrared spectroscopy (NIRS) has been approved for detection of coronary lipids. AIMS: We aimed to assess the ability of detailed OCT analysis to identify coronary lipids, using NIRS as the reference method. METHODS: In total, 40 patients with acute coronary syndrome underwent imaging of a non-culprit lesion by both NIRS and OCT. For each segment, the NIRS-derived 4 mm segment with maximal lipid core burden index (maxLCBI4mm) was assessed. OCT analysis was performed using a semi-automated method including measurement of the fibrous cap thickness (FCT) of all detected fibroatheromas. Subsequent quantitative volumetric evaluation furnished FCT, FC surface area (FC SA), lipid arc, and FC (fibrous cap) volume data. OCT features of lipid plaques were compared with maxLCBI4mm. Predictors of maxLCBI4mm >400 was assessed by using univariable and multivariable analysis. RESULTS: OCT features (mean FCT, total FC SA, FC volume, maximal, mean, and total lipid arcs) strongly correlated with the maxLCBI4mm (p = 0.012 for the mean FCT, respectively p < 0.001 for all other aforementioned features). The strongest predictors of maxLCBI4mm >400 were the maximal (p = 0.002) and mean (p = 0.002) lipid arc, and total FC SA (p = 0.012). CONCLUSIONS: We found a strong correlation between the OCT-derived features and NIRS findings. Detailed OCT analysis may be reliably used for detection of the presence of coronary lipids.

Radiomics Based Differentiation of Glioblastoma and Metastatic Disease: Impact of Different T1-Contrast Enhanced Sequences on Radiomic Features and Model Performance.

BACKGROUND AND PURPOSE: To evaluate the radiomics-based model performance for differentiation between glioblastoma (GB) and brain metastases (BM) using magnetization prepared rapid gradient echo (MPRAGE) and volumetric interpolated breath-hold examination (VIBE) T1-contrast enhanced sequences. MATERIALS AND METHODS: T1-CE MPRAGE and VIBE sequences acquired in 108 patients (31 GBs and 77 BM) during the same MRI session were retrospectively evaluated. Post standardized image pre-processing and segmentation, radiomics features were extracted from necrotic and enhancing tumor components. Pearson correlation analysis of radiomics features from tumor subcomponents was also performed. A total of 90 machine learning (ML) pipelines were evaluated using a five-fold cross validation. Performance was measured by mean AUC-ROC, Log-loss and Brier scores. RESULTS: A feature-wise comparison showed that the radiomic features between sequences were strongly correlated, with the highest correlation for shape-based features. The mean AUC across the top-ten pipelines ranged between 0.851-0.890 with T1-CE MPRAGE and between 0.869-0.907 with T1-CE VIBE sequence. Top performing models for the MPRAGE sequence commonly used support vector machines, while those for VIBE sequence used either support vector machines or random forest. Common feature reduction methods for top-performing models included linear combination filter and least absolute shrinkage and selection operator (LASSO) for both sequences. For the same ML-feature reduction pipeline, model performances were comparable (AUC-ROC difference range: [-0.078, 0.046]). CONCLUSIONS: Radiomic features derived from T1-CE MPRAGE and VIBE sequences are strongly correlated and may have similar overall classification performance for differentiating GB from BM. ABBREVIATIONS: BM: Brain metastases, GB: glioblastoma, T1-CE: T1 contrast enhanced sequence, MPRAGE: magnetization prepared rapid gradient echo, ML: machine learning, RF: random forest, VIBE: volumetric interpolated breath-hold examination.

Hybrid deep learning and optimal graph search method for optical coherence tomography layer segmentation in diseases affecting the optic nerve.

Accurate segmentation of retinal layers in optical coherence tomography (OCT) images is critical for assessing diseases that affect the optic nerve, but existing automated algorithms often fail when pathology causes irregular layer topology, such as extreme thinning of the ganglion cell-inner plexiform layer (GCIPL). Deep LOGISMOS, a hybrid approach that combines the strengths of deep learning and 3D graph search to overcome their limitations, was developed to improve the accuracy, robustness and generalizability of retinal layer segmentation. The method was trained on 124 OCT volumes from both eyes of 31 non-arteritic anterior ischemic optic neuropathy (NAION) patients and tested on three cross-sectional datasets with available reference tracings: Test-NAION (40 volumes from both eyes of 20 NAION subjects), Test-G (29 volumes from 29 glaucoma subjects/eyes), and Test-JHU (35 volumes from 21 multiple sclerosis and 14 control subjects/eyes) and one longitudinal dataset without reference tracings: Test-G-L (155 volumes from 15 glaucoma patients/eyes). In the three test datasets with reference tracings (Test-NAION, Test-G, and Test-JHU), Deep LOGISMOS achieved very high Dice similarity coefficients (%) on GCIPL: 89.97±3.59, 90.63±2.56, and 94.06±1.76, respectively. In the same context, Deep LOGISMOS outperformed the Iowa reference algorithms by improving the Dice score by 17.5, 5.4, and 7.5, and also surpassed the deep learning framework nnU-Net with improvements of 4.4, 3.7, and 1.0. For the 15 severe glaucoma eyes with marked GCIPL thinning (Test-G-L), it demonstrated reliable regional GCIPL thickness measurement over five years. The proposed Deep LOGISMOS approach has potential to enhance precise quantification of retinal structures, aiding diagnosis and treatment management of optic nerve diseases.

A Technique to Enable Efficient Adaptive Radiation Therapy: Automated Contouring of Prostate and Adjacent Organs.

Purpose: The purpose of this work was to investigate the use of a segmentation approach that could potentially improve the speed and reproducibility of contouring during magnetic resonance-guided adaptive radiation therapy. Methods and Materials: The segmentation algorithm was based on a hybrid deep neural network and graph optimization approach that also allows rapid user intervention (Deep layered optimal graph image segmentation of multiple objects and surfaces [LOGISMOS] + just enough interaction [JEI]). A total of 115 magnetic resonance-data sets were used for training and quantitative assessment. Expert segmentations were used as the independent standard for the prostate, seminal vesicles, bladder, rectum, and femoral heads for all 115 data sets. In addition, 3 independent radiation oncologists contoured the prostate, seminal vesicles, and rectum for a subset of patients such that the interobserver variability could be quantified. Consensus contours were then generated from these independent contours using a simultaneous truth and performance level estimation approach, and the deviation of Deep LOGISMOS + JEI contours to the consensus contours was evaluated and compared with the interobserver variability. Results: The absolute accuracy of Deep LOGISMOS + JEI generated contours was evaluated using median absolute surface-to-surface distance which ranged from a minimum of 0.20 mm for the bladder to a maximum of 0.93 mm for the prostate compared with the independent standard across all data sets. The median relative surface-to-surface distance was less than 0.17 mm for all organs, indicating that the Deep LOGISMOS + JEI algorithm did not exhibit a systematic under- or oversegmentation. Interobserver variability testing yielded a mean absolute surface-to-surface distance of 0.93, 1.04, and 0.81 mm for the prostate, seminal vesicles, and rectum, respectively. In comparison, the deviation of Deep LOGISMOS + JEI from consensus simultaneous truth and performance level estimation contours was 0.57, 0.64, and 0.55 mm for the same organs. On average, the Deep LOGISMOS algorithm took less than 26 seconds for contour segmentation. Conclusions: Deep LOGISMOS + JEI segmentation efficiently generated clinically acceptable prostate and normal tissue contours, potentially limiting the need for time intensive manual contouring with each fraction.

Kernel-weighted contribution: a method of visual attribution for 3D deep learning segmentation in medical imaging.

Purpose: Explaining deep learning model decisions, especially those for medical image segmentation, is a critical step toward the understanding and validation that will enable these powerful tools to see more widespread adoption in healthcare. We introduce kernel-weighted contribution, a visual explanation method for three-dimensional medical image segmentation models that produces accurate and interpretable explanations. Unlike previous attribution methods, kernel-weighted contribution is explicitly designed for medical image segmentation models and assesses feature importance using the relative contribution of each considered activation map to the predicted segmentation. Approach: We evaluate our method on a synthetic dataset that provides complete knowledge over input features and a comprehensive explanation quality metric using this ground truth. Our method and three other prevalent attribution methods were applied to five different model layer combinations to explain segmentation predictions for 100 test samples and compared using this metric. Results: Kernel-weighted contribution produced superior explanations of obtained image segmentations when applied to both encoder and decoder sections of a trained model as compared to other layer combinations (p<0.0005). In between-method comparisons, kernel-weighted contribution produced superior explanations compared with other methods using the same model layers in four of five experiments (p<0.0005) and showed equivalently superior performance to GradCAM++ when only using non-transpose convolution layers of the model decoder (p=0.008). Conclusion: The reported method produced explanations of superior quality uniquely suited to fully utilize the specific architectural considerations present in image and especially medical image segmentation models. Both the synthetic dataset and implementation of our method are available to the research community.

Segmentation quality assessment by automated detection of erroneous surface regions in medical images.

Despite the advancement in deep learning-based semantic segmentation methods, which have achieved accuracy levels of field experts in many computer vision applications, the same general approaches may frequently fail in 3D medical image segmentation due to complex tissue structures, noisy acquisition, disease-related pathologies, as well as the lack of sufficiently large datasets with associated annotations. For expeditious diagnosis and quantitative image analysis in large-scale clinical trials, there is a compelling need to predict segmentation quality without ground truth. In this paper, we propose a deep learning framework to locate erroneous regions on the boundary surfaces of segmented objects for quality control and assessment of segmentation. A Convolutional Neural Network (CNN) is explored to learn the boundary related image features of multi-objects that can be used to identify location-specific inaccurate segmentation. The predicted error locations can facilitate efficient user interaction for interactive image segmentation (IIS). We evaluated the proposed method on two data sets: Osteoarthritis Initiative (OAI) 3D knee MRI and 3D calf muscle MRI. The average sensitivity scores of 0.95 and 0.92, and the average positive predictive values of 0.78 and 0.91 were achieved, respectively, for erroneous surface region detection of knee cartilage segmentation and calf muscle segmentation. Our experiment demonstrated promising performance of the proposed method for segmentation quality assessment by automated detection of erroneous surface regions in medical images.

Patch-wise 3D segmentation quality assessment combining reconstruction and regression networks.

Purpose: General deep-learning (DL)-based semantic segmentation methods with expert level accuracy may fail in 3D medical image segmentation due to complex tissue structures, lack of large datasets with ground truth, etc. For expeditious diagnosis, there is a compelling need to predict segmentation quality without ground truth. In some medical imaging applications, maintaining the quality of segmentation is crucial to the localized regions where disease is prevalent rather than just globally maintaining high-average segmentation quality. We propose a DL framework to identify regions of segmentation inaccuracies by combining a 3D generative adversarial network (GAN) and a convolutional regression network. Approach: Our approach is methodologically based on the learned ability to reconstruct the original images identifying the regions of location-specific segmentation failures, in which the reconstruction does not match the underlying original image. We use conditional GAN to reconstruct input images masked by the segmentation results. The regression network is trained to predict the patch-wise Dice similarity coefficient (DSC), conditioned on the segmentation results. The method relies directly on the extracted segmentation related features and does not need to use ground truth during the inference phase to identify erroneous regions in the computed segmentation. Results: We evaluated the proposed method on two public datasets: osteoarthritis initiative 4D (3D + time) knee MRI (knee-MR) and 3D non-small cell lung cancer CT (lung-CT). For the patch-wise DSC prediction, we observed the mean absolute errors of 0.01 and 0.04 with the independent standard for the knee-MR and lung-CT data, respectively. Conclusions: This method shows promising results in localizing the erroneous segmentation regions that may aid the downstream analysis of disease diagnosis and prognosis prediction.

Assisted annotation in Deep LOGISMOS: Simultaneous multi-compartment 3D MRI segmentation of calf muscles.

BACKGROUND: Automated segmentation of individual calf muscle compartments in 3D MR images is gaining importance in diagnosing muscle disease, monitoring its progression, and prediction of the disease course. Although deep convolutional neural networks have ushered in a revolution in medical image segmentation, achieving clinically acceptable results is a challenging task and the availability of sufficiently large annotated datasets still limits their applicability. PURPOSE: In this paper, we present a novel approach combing deep learning and graph optimization in the paradigm of assisted annotation for solving general segmentation problems in 3D, 4D, and generally n-D with limited annotation cost. METHODS: Deep LOGISMOS combines deep-learning-based pre-segmentation of objects of interest provided by our convolutional neural network, FilterNet+, and our 3D multi-objects LOGISMOS framework (layered optimal graph image segmentation of multiple objects and surfaces) that uses newly designed trainable machine-learned cost functions. In the paradigm of assisted annotation, multi-object JEI for efficient editing of automated Deep LOGISMOS segmentation was employed to form a new larger training set with significant decrease of manual tracing effort. RESULTS: We have evaluated our method on 350 lower leg (left/right) T1-weighted MR images from 93 subjects (47 healthy, 46 patients with muscular morbidity) by fourfold cross-validation. Compared with the fully manual annotation approach, the annotation cost with assisted annotation is reduced by 95%, from 8 h to 25 min in this study. The experimental results showed average Dice similarity coefficient (DSC) of 96.56 ± 0.26 % $96.56\pm 0.26 \%$ and average absolute surface positioning error of 0.63 pixels (0.44 mm) for the five 3D muscle compartments for each leg. These results significantly improve our previously reported method and outperform the state-of-the-art nnUNet method. CONCLUSIONS: Our proposed approach can not only dramatically reduce the expert's annotation efforts but also significantly improve the segmentation performance compared to the state-of-the-art nnUNet method. The notable performance improvements suggest the clinical-use potential of our new fully automated simultaneous segmentation of calf muscle compartments.

Using computed tomography to recover hidden medieval fragments beneath early modern leather bindings, first results.

Medieval bindings fragments have become increasingly interesting to Humanities researchers as sources for the textual and material history of medieval Europeans. Later book binders used these discarded and repurposed pieces of earlier medieval manuscripts to reinforce the structures of other manuscripts and printed books. That many of these fragments are contained within and obscured by decorative bindings that cannot be dismantled ethically has limited their discovery and description. Although previous attempts to recover these texts using IRT and MA-XRF scanning have been successful, the extensive time required to scan a single book, and the need to modify or create specialized IRT or MA-XRF equipment for this method are drawbacks. Our research proposes and tests the capabilities of medical CT scanning technologies (commonly available at research university medical schools) for making visible and legible these fragments hidden under leather bindings. Our research team identified three sixteenth-century printed codices in our university libraries that were evidently bound in tawed leather by one workshop. The damaged cover of one of these three had revealed medieval manuscript fragments on the book spine; this codex served as a control for testing the other two volumes to see if they, too, contain fragments. The use of a medical CT scanner proved successful in visualizing interior book-spine structures and some letterforms, but not all of the text was made visible. The partial success of CT-scanning points to the value of further experimentation, given the relatively wide availability of medical imaging technologies, with their potential for short, non-destructive, 3D imaging times.

Caldendrin represses neurite regeneration and growth in dorsal root ganglion neurons.

Caldendrin is a Ca2+ binding protein that interacts with multiple effectors, such as the Cav1 L-type Ca2+ channel, which play a prominent role in regulating the outgrowth of dendrites and axons (i.e., neurites) during development and in response to injury. Here, we investigated the role of caldendrin in Cav1-dependent pathways that impinge upon neurite growth in dorsal root ganglion neurons (DRGNs). By immunofluorescence, caldendrin was localized in medium- and large- diameter DRGNs. Compared to DRGNs cultured from WT mice, DRGNs of caldendrin knockout (KO) mice exhibited enhanced neurite regeneration and outgrowth. Strong depolarization, which normally represses neurite growth through activation of Cav1 channels, had no effect on neurite growth in DRGN cultures from female caldendrin KO mice. Remarkably, DRGNs from caldendrin KO males were no different from those of WT males in terms of depolarization-dependent neurite growth repression. We conclude that caldendrin opposes neurite regeneration and growth, and this involves coupling of Cav1 channels to growth-inhibitory pathways in DRGNs of females but not males.