Small Region Of Interest Research Articles

Attention-driven visual emphasis for medical volumetric image visualization

AbstractDirect volume rendering (DVR) is a commonly utilized technique for three-dimensional visualization of volumetric medical images. A key goal of DVR is to enable users to visually emphasize regions of interest (ROIs) which may be occluded by other structures. Conventional methods for ROIs visual emphasis require extensive user involvement for the adjustment of rendering parameters to reduce the occlusion, dependent on the user’s viewing direction. Several works have been proposed to automatically preserve the view of the ROIs by eliminating the occluding structures of lower importance in a view-dependent manner. However, they require pre-segmentation labeling and manual importance assignment on the images. An alternative to ROIs segmentation is to use ‘saliency’ to identify important regions. This however lacks semantic information and thus leads to the inclusion of false positive regions. In this study, we propose an attention-driven visual emphasis method for volumetric medical image visualization. We developed a deep learning attention model, termed as focused-class attention map (F-CAM), trained with only image-wise labels for automated ROIs localization and importance estimation. Our F-CAM transfers the semantic information from the classification task for use in the localization of ROIs, with a focus on small ROIs that characterize medical images. Additionally, we propose an attention compositing module that integrates the generated attention map with transfer function within the DVR pipeline to automate the view-dependent visual emphasis of the ROIs. We demonstrate the superiority of our method compared to existing methods on a multi-modality PET-CT dataset and an MRI dataset.

Development of an initial training and evaluation programme for manual lower limb muscle MRI segmentation

BackgroundMagnetic resonance imaging (MRI) quantification of intramuscular fat accumulation is a responsive biomarker in neuromuscular diseases. Despite emergence of automated methods, manual muscle segmentation remains an essential foundation. We aimed to develop a training programme for new observers to demonstrate competence in lower limb muscle segmentation and establish reliability benchmarks for future human observers and machine learning segmentation packages.MethodsThe learning phase of the training programme comprised a training manual, direct instruction, and eight lower limb MRI scans with reference standard large and small regions of interest (ROIs). The assessment phase used test–retest scans from two patients and two healthy controls. Interscan and interobserver reliability metrics were calculated to identify underperforming outliers and to determine competency benchmarks.ResultsThree experienced observers undertook the assessment phase, whilst eight new observers completed the full training programme. Two of the new observers were identified as underperforming outliers, relating to variation in size or consistency of segmentations; six had interscan and interobserver reliability equivalent to those of experienced observers. The calculated benchmark for the Sørensen-Dice similarity coefficient between observers was greater than 0.87 and 0.92 for individual thigh and calf muscles, respectively. Interscan and interobserver reliability were significantly higher for large than small ROIs (all p < 0.001).ConclusionsWe developed, implemented, and analysed the first formal training programme for manual lower limb muscle segmentation. Large ROI showed superior reliability to small ROI for fat fraction assessment.Relevance statementObservers competent in lower limb muscle segmentation are critical to application of quantitative muscle MRI biomarkers in neuromuscular diseases. This study has established competency benchmarks for future human observers or automated segmentation methods.Key points• Observers competent in muscle segmentation are critical for quantitative muscle MRI biomarkers.• A training programme for muscle segmentation was undertaken by eight new observers.• We established competency benchmarks for future human observers or automated segmentation methods.Graphical

23Na MRI: inter-reader reproducibility of normal fibroglandular sodium concentration measurements at 3 T

BackgroundTo study the reproducibility of 23Na magnetic resonance imaging (MRI) measurements from breast tissue in healthy volunteers.MethodsUsing a dual-tuned bilateral 23Na/1H breast coil at 3-T MRI, high-resolution 23Na MRI three-dimensional cones sequences were used to quantify total sodium concentration (TSC) and fluid-attenuated sodium concentration (FASC). B1-corrected TSC and FASC maps were created. Two readers manually measured mean, minimum and maximum TSC and mean FASC values using two sampling methods: large regions of interest (LROIs) and small regions of interest (SROIs) encompassing fibroglandular tissue (FGT) and the highest signal area at the level of the nipple, respectively. The reproducibility of the measurements and correlations between density, age and FGT apparent diffusion coefficient (ADC) values were evaluatedss.ResultsNine healthy volunteers were included. The inter-reader reproducibility of TSC and FASC using SROIs and LROIs was excellent (intraclass coefficient range 0.945−0.979, p < 0.001), except for the minimum TSC LROI measurements (p = 0.369). The mean/minimum LROI TSC and mean LROI FASC values were lower than the respective SROI values (p < 0.001); the maximum LROI TSC values were higher than the SROI TSC values (p = 0.009). TSC correlated inversely with age but not with FGT ADCs. The mean and maximum FGT TSC and FASC values were higher in dense breasts in comparison to non-dense breasts (p < 0.020).ConclusionsThe chosen sampling method and the selected descriptive value affect the measured TSC and FASC values, although the inter-reader reproducibility of the measurements is in general excellent.Relevance statement23Na MRI at 3 T allows the quantification of TSC and FASC sodium concentrations. The sodium measurements should be obtained consistently in a uniform manner.Key points• 23Na MRI allows the quantification of total and fluid-attenuated sodium concentrations (TSC/FASC).• Sampling method (large/small region of interest) affects the TSC and FASC values.• Dense breasts have higher TSC and FASC values than non-dense breasts.• The inter-reader reproducibility of TSC and FASC measurements was, in general, excellent.• The results suggest the importance of stratifying the sodium measurements protocol.Graphical

LiverZap: a chemoptogenetic tool for global and locally restricted hepatocyte ablation to study cellular behaviours in liver regeneration.

The liver restores its mass and architecture after injury. Yet, investigating morphogenetic cell behaviours and signals that repair tissue architecture at high spatiotemporal resolution remains challenging. We developed LiverZap, a tuneable chemoptogenetic liver injury model in zebrafish. LiverZap employs the formation of a binary FAP-TAP photosensitiser followed by brief near-infrared illumination inducing hepatocyte-specific death and recapitulating mammalian liver injury types. The tool enables local hepatocyte ablation and extended live imaging capturing regenerative cell behaviours, which is crucial for studying cellular interactions at the interface of healthy and damaged tissue. Applying LiverZap, we show that targeted hepatocyte ablation in a small region of interest is sufficient to trigger local liver progenitor-like cell (LPC)-mediated regeneration, challenging the current understanding of liver regeneration. Surprisingly, the LPC response is also elicited in adjacent uninjured tissue, at up to 100 µm distance to the injury. Moreover, dynamic biliary network rearrangement suggests active cell movements from uninjured tissue in response to substantial hepatocyte loss as an integral step of LPC-mediated liver regeneration. This precisely targetable liver cell ablation tool will enable the discovery of key molecular and morphogenetic regeneration paradigms.

Unbiased centroiding of point targets close to the Cramer Rao limit.

Systematic errors affecting center-of-gravity (CoG) measurements may occur from coarse sampling of the point-spread-function (PSF) or from signal truncation at the boundaries of the region-of-interest (ROI). For small ROI and PSF widths, these effects are shown to become dominant, but this can be mitigated by introducing novel unbiased estimators that are largely free of systematic error and perform particularly well for low photon numbers. Analytical expressions for the estimator variances, comprising contributions from photon shot noise, random pixel noise, and residual systematic error, are derived and verified by Monte Carlo simulations. The accuracy and computational speed of the unbiased estimators are compared to those of other common estimators, including iteratively weighted CoG, thresholded CoG, iterative least squares fitting, and two-dimensional Gaussian regression. Each estimator is optimized with respect to ROI size and PSF radius and its error compared to the theoretical limit defined by the Cramer Rao lower bound (CRLB). The unbiased estimator with full systematic error correction operating on a small ROI [3×3] emerges as one of the most accurate estimators while requiring significantly less computing effort than alternative algorithms.

Compressive Raman imaging by combining scattering-projection interleaving with context-aware excitation.

Exciting an object with a laser-focus array and randomly interleaving its scattering projection has been proved to be an effective strategy for speeding up Raman imaging. The so-called scattering interleaved Raman imaging (SIRI) method allows Raman hyperspectral imaging with a single snapshot and exhibits excellent reconstruction fidelity and signal-to-noise ratios (SNRs). Here, we show that the performance of SIRI is significantly improved when combined with context-aware excitation. The experiments on micro-plastics demonstrate that the restriction of Raman excitation within a smaller region of interest as guided by bright-field microscopy improves the signal intensity and the SNR, and it is surprising that the spectral resolution is also significantly improved. The context-aware SIRI method is successfully used for imaging of lipid-producing yeast cells, suggesting that it is a promising analytical tool for studying live cells or tissues.

Deep-Interior: A new pathway to interior tomographic image reconstruction via a weighted backprojection and deep learning.

In recent years, deep learning strategies have been combined with either the filtered backprojection or iterative methods or the direct projection-to-image by deep learning only to reconstruct images. Some of these methods can be applied to address the interior reconstruction problems for centered regions of interest (ROIs) with fixed sizes. Developing a method to enable interior tomography with arbitrarily located ROIs with nearly arbitrary ROI sizes inside a scanning field of view (FOV) remains an openquestion. To develop a new pathway to enable interior tomographic reconstruction for arbitrarily located ROIs with arbitrary sizes using a single trained deep neural network model. The method consists of two steps. First, an analytical weighted backprojection reconstruction algorithm was developed to perform domain transform from divergent fan-beam projection data to an intermediate image feature space, , for an arbitrary size ROI at an arbitrary location inside the FOV. Second, a supervised learning technique was developed to train a deep neural network architecture to perform deconvolution to obtain the true image from the new feature space . This two-step method is referred to as Deep-Interior for convenience. Both numerical simulations and experimental studies were performed to validate the proposed Deep-Interior method. The results showed that ROIs as small as a diameter of 5 cm could be accurately reconstructed (similarity index 0.985 ± 0.018 on internal testing data and 0.940 ± 0.025 on external testing data) at arbitrary locations within an imaging object covering a wide variety of anatomical structures of different body parts. Besides, ROIs of arbitrary size can be reconstructed by stitching small ROIs without additionaltraining. The developed Deep-Interior framework can enable interior tomographic reconstruction from divergent fan-beam projections for short-scan and super-short-scan acquisitions for small ROIs (with a diameter larger than 5 cm) at an arbitrary location inside the scanning FOV with high quantitative reconstruction accuracy.

Diffusion-weighted imaging in pediatric extracranial germ cell tumors.

Germ cell tumors (GCTs) comprise a rare and heterogeneous group of neoplasms presenting different clinical and histological characteristics, leading to a challenging scenario in clinical practice. Diffusion-weighted imaging (DWI) has been suggested as an indirect marker of tumor density and cellularity and could be used to monitor therapeutic response. However, its role in pediatric GCTs needs to be clarified. Here, we evaluated the features of DWI in pediatric extracranial GCTs in a reference Brazilian institution. We included 43 pediatric patients with primary GCTs treated between 2008 and 2022 in Hospital de Amor de Barretos. The patients' MRI images included T1-weighted without contrast, T2-weighted, DWI and apparent diffusion coefficient (ADC) maps. DWI was evaluated in the section that exhibited the greatest restricted diffusion in the largest hypersignal area of the image. The lowest ADC value was determined to define the region of interest (ROI). We used a small ROI, avoiding necrotic, adipose tissue, noisy or nonenhancing lesion voxels as recommended. ROI determination was established by visual inspection by two radiologists in accordance. We used two values of b (b = 50 mm2/s or b = 800) for ADC values. The highest mean ADC (mADC) value was observed in pure teratomas (1,403.50 ± 161.76 x10-3 mm2/s; mean ± SD) compared to other histologies (yolk sac, mixed teratoma, dysgerminoma and mixed GCT) of GCT (p<0.001). Furthermore, ROC analysis determined a cutoff mADC value of 1,179.00 x 10-3 mm2/s that differentiated pure teratomas from the other GCT histologies with a sensitivity of 95.8% and a specificity of 92.9% (AUC = 0.979; p<0.01). A significant increase in mADC was observed for malignant GCTs in treatment (1,197.00 ± 372.00 mm2/s; p<0.001) compared to that exhibited at the time of diagnosis (780.00 ± 168.00 mm2/s; mean ± SD. Our findings suggest that mADC assessment could be used as a tool to distinguish pure teratomas from malignant CGT histologies at diagnosis. Additionally, we demonstrated reasonable evidence that it could be used as a complementary tool to monitor treatment response in patients with malignant GCT.

Impact of Region-of-Interest Size on the Diagnostic Performance of Shear Wave Elastography in Differentiating Thyroid Nodules.

This study investigated the impact of different region-of-interest (ROI) sizes (Max, 1 mm, and 2 mm) on shear wave elastography (SWE) in differentiating between malignant and benign thyroid nodules. The study cohort comprised 129 thyroid nodules (50 malignant, 79 benign) and 78 normal subjects. Diagnostic efficacy was assessed through pairwise comparisons of area under the curve (AUC) values in receiver operating characteristic analysis by using DeLong's test. Our results indicated significant differences in all SWE elasticity metrics between the groups, with malignant nodules exhibiting higher values than benign nodules (p < 0.05). Smaller ROIs (1 and 2 mm) were found to outperform the max ROI in terms of diagnostic accuracy, particularly for the Emax and Emin elasticity metrics. Emax(1mm) had the highest diagnostic accuracy, with an AUC of 0.883, sensitivity of 74.0%, and specificity of 86.1%. This study underscores the significant influence of ROI size selection on the diagnostic performance of SWE, offering valuable insights for future research and clinical applications in thyroid nodule assessment.

Accuracy of Electron Density Mapping of a Novel kVCBCT System Designed for Planning

A recently developed advanced kVCBCT system, designed for simulation and planning, includes improvements to the imaging panel and reconstruction technique, enabling more accurate electron density mapping than a typical CBCT. This study investigates the accuracy, limitations, and dosimetric impact of this new system. CBCT images were taken of both large (pelvis) and small (head) electron density phantoms with inserts ranging from lung (0.29) to cortical bone (1.69). Images were taken with several pre-set protocols with energies of 125 kVp and 140 kVp. The effects of longitudinal buildup (for scatter due to the cone-beam geometry) and blade position were also investigated. The HU values of each insert were measured over a small region of interest. Several electron density curves were generated - pelvis protocol on the large phantom, head protocol on the small phantom, and average - and imported into a treatment planning system. Dose calculation was performed using each curve and differences were observed. HU to electron density mapping was sensitive to the use of longitudinal buildup and blade position, with differences in the curves observed at both low and high electron densities, due to imager saturation effects not found in typical clinical scenarios. Subsequent studies used both longitudinal buildup and fully opened blades. Under these conditions, mapping was not sensitive to technique or phantom size at low electron densities. At higher values, the curves diverged, with the head protocol showing higher HU values and the pelvis protocol showing lower. The average curve matched very closely with the curve from our standard CT simulator. Dose calculation showed little dependence on the curve chosen, with max point dose differences of 1.2% between the pelvis and average scans, though most of the plan was <0.5% different. Results for the head protocol were similar. The system provides reliable HU values comparable to a CT simulator, though it is important to consider the effects of imager saturation by choosing appropriate scatter and collimation. The calibration is slightly sensitive to the choice of phantom size and beam energy, though these differences led to negligible (<0.5%) differences in dose calculation accuracy.

A novel process for H&E, immunofluorescence, and imaging mass cytometry on a single slide with a concise analytics pipeline.

Imaging mass cytometry (IMC) is a powerful spatial technology that utilizes cytometry time of flight to acquire multiplexed image datasets with up to 40 markers, via metal-tagged antibodies. Recent advances in IMC have led to the inclusion of RNAScope probes and multiple new analysis pipelines have led to faster analyses and better results. However, IMC still suffers from lower resolution (1 μm2 pixels) and relatively small regions of interest (ROIs) (<2 mm2 ) compared to other, light-based microscope technologies. Capturing higher-resolution images on serial sections causes great difficulty when attempting to align cells and structures across serial sections, especially when observing smaller cell types and structures. Therefore, we demonstrate the combination of H&E and multiplex immunofluorescence imaging, for much higher resolution of the structural and cellular compartments found throughout the entire tissue section, with the high-dimensionality of IMC for specific ROIs on a single slide. Additionally, we demonstrate a simple and effective open-source cell segmentation and IMC analysis pipeline with previously published and freely available software.

Initial experiences with Direct Imaging of Neuronal Activity (DIANA) in humans

Abstract Functional MRI (fMRI) has been widely used to study activity patterns in the human brain. It infers neuronal activity from the associated hemodynamic response, which fundamentally limits its spatiotemporal specificity. In mice, the Direct Imaging of Neuronal Activity (DIANA) method revealed MRI signals that correlated with extracellular electric activity, showing high spatiotemporal specificity. In this work, we attempted DIANA in humans. Five experimental paradigms were tested, exploring different stimulus types (flickering noise patterns, and naturalistic images), stimulus durations (50–200 ms), and imaging resolution (2 × 2 × 5 mm3 and 1 × 1 × 5 mm3). Regions of interest (ROI) were derived from Blood Oxygen Level Dependent (BOLD) fMRI acquisitions (both EPI and FLASH based) and T1-weighted anatomical scans. In Paradigm I (n = 1), using flickering noise patterns, signals were detected that resembled possible functional activity from a small ROI. However, changes in stimulus duration did not lead to corresponding signal changes (Paradigm II; n = 1). Therefore, care should be taken not to mistake artifacts for neuronal activity. In Paradigm III (n = 3), when averaged across multiple subjects, a ~200 ms long 0.02% signal increase was observed ~100 ms after the stimulus onset (10x smaller than the expected signal). However, white matter control ROIs showed similarly large signal fluctuations. In Paradigm IV (n = 3), naturalistic image stimuli were used, but did not reveal signs of a potential functional signal. To reduce partial voluming effects and improve ROI definition, in Paradigm V (n = 3), we acquired data with higher resolution (1 × 1 × 5 mm3) using naturalistic images. However, no sign of activation was found. It is important to note that repetitive experiments with short interstimulus intervals were found to be strenuous for the subjects, which likely impacted data quality. To obtain better data, improvements in sequence and stimulus designs are needed to maximize the DIANA signal and minimize confounds. However, without a clear understanding of DIANA’s biophysical underpinnings it is difficult to do so. Therefore, it may be more effective to first investigate DIANA signals with simultaneously recorded electrophysiological signals in more controlled settings, e.g., in anesthetized mice.

Simultaneously acquired rSUV and rCBF of 18F-FDG/MRI in peritumoral brain zone can help to differentiate the grade of gliomas

ObjectivesThe purpose of this study is to investigate the diagnostic performance of the peritumoral brain zone (PBZ) in differentiating glioma grades. This is accomplished by comparing the relative standardized uptake values (rSUV) and relative cerebral blood flow (rCBF) obtained from hybrid 18F-fluoro-2-deoxy-d-glucose positron emission tomography/magnetic resonance imaging (18F-FDG PET/MRI) within different regions of interest, including the solid portion (SP) and the PBZ. MethodsTwenty-four patients with gliomas who underwent preoperative 18F-PET/MRI were enrolled in this study. The maximum standardized uptake values (SUVmax) and relative maximum cerebral blood flow (rCBFmax) were obtained from the FDG-PET and ASL data, respectively. The relative SUVmax (rSUVmax) was calculated by standardizing against the contralateral normal-appearing brain cortex. Data from the solid portion (SP) of tumor and the peritumoral brain zone (PBZ) at distance of 5 ​mm, 10 ​mm, 15 ​mm, and 20 ​mm from the SP margin were recorded. Logistic regression was used to generate receiver-operating characteristic (ROC) curves. The areas under the ROC curves (AUCs) were calculated and compared to analyze the diagnostic utility of each parameter. ResultsIn comparison to low-grade glioma (LGG), high-grade glioma (HGG) exhibited significantly higher rSUVmax and rCBFmax values in both the SP and the proximal PBZ (P ​< ​0.05). Among the various distance parameters and their combinations, the single parameter rSUVmax-SP demonstrated the highest diagnostic efficacy with an AUC of 0.788 (P ​< ​0.05). However, the AUC of rSUVmax-SP did not show a significantly improvement when combined with PBZs (P ​> ​0.05). When combining PBZs and SP with rSUVmax and rCBFmax, the rSUVmax and rCBFmax values of SP to PBZ 20 ​mm exhibited superior performance compared to single parameters and smaller regions of interest, with an AUC of 0.848. The sensitivity and specificity were determined as 73.8% and 83.6%, respectively. ConclusionThe combination of rSUVmax and rCBFmax in the SP and PBZ, based on hybrid PET/MRI, proves to be superior to using parameters solely in the SP when it comes to differentiating between HGG and LGG. Expanding the study appropriately and incorporating the use of multiple parameters can offer more valuable diagnostic information, which holds potential for clinical applications.

MISM: A Medical Image Segmentation Metric for Evaluation of Weak Labeled Data.

Performance measures are an important tool for assessing and comparing different medical image segmentation algorithms. Unfortunately, the current measures have their weaknesses when it comes to assessing certain edge cases. These limitations arise when images with a very small region of interest or without a region of interest at all are assessed. As a solution to these limitations, we propose a new medical image segmentation metric: MISm. This metric is a composition of the Dice similarity coefficient and the weighted specificity. MISm was investigated for definition gaps, an appropriate scoring gradient, and different weighting coefficients used to propose a constant value. Furthermore, an evaluation was performed by comparing the popular metrics in the medical image segmentation and MISm using images of magnet resonance tomography from several fictitious prediction scenarios. Our analysis shows that MISm can be applied in a general way and thus also covers the mentioned edge cases, which are not covered by other metrics, in a reasonable way. In order to allow easy access to MISm and therefore widespread application in the community, as well as reproducibility of experimental results, we included MISm in the publicly available evaluation framework MISeval.

Optimal echo times for quantitative susceptibility mapping: A test-retest study on basal ganglia and subcortical brain nuclei

Quantitative Susceptibility Mapping (QSM) is a recent MRI-technique able to quantify the bulk magnetic susceptibility of myelin, iron, and calcium in the brain. Its variability across different acquisition parameters has prompted the need for standardisation across multiple centres and MRI vendors. However, a high level of agreement between repeated imaging acquisitions is equally important. With this study we aimed to assess the inter-scan repeatability of an optimised multi-echo GRE sequence in 28 healthy volunteers. We extracted and compared the susceptibility measures from the scan and rescan acquisitions across 7 bilateral brain regions (i.e., 14 regions of interest (ROIs)) relevant for neurodegeneration. Repeatability was first assessed while reconstructing QSM with a fixed number of echo times (i.e., 8). Excellent inter-scan repeatability was found for putamen, globus pallidus and caudate nucleus, while good performance characterised the remaining structures. An increased variability was instead noted for small ROIs like red nucleus and substantia nigra. Secondly, we assessed the impact exerted on repeatability by the number of echoes used to derive QSM maps. Results were impacted by this parameter, especially in smaller regions. Larger brain structures, on the other hand, showed more consistent performance. Nevertheless, with either 8 or 7 echoes we managed to obtain good inter-scan repeatability on almost all ROIs. These findings indicate that the designed acquisition/reconstruction protocol has wide applicability, particularly in clinical or research settings involving longitudinal acquisitions (e.g. rehabilitation studies).

Quantitative 3D assessment of the lung cancer microenvironment using multi-resolution open-top light-sheet microscopy.

2616 Background: Non-small cell lung cancer (NSCLC) tissue is a valuable resource for diagnosis, treatment planning, and drug development. Current 2D histopathological techniques introduce under-sampling error (i.e., a single 5 µm section represents 0.5% of a 1 mm thick biopsy), interobserver variability, and failure to capture the biology contained within the entire tissue sample. We have developed a suite of technologies to stain, chemically clarify, image, visualize, and analyze entire, intact lung tissue samples. Herein, we present analysis methods for comprehensive 3D assessment of intact NSCLC tissue. Methods: Human NSCLC tissue, stored frozen in optimal cutting temperature (OCT) compound, was fixed in 4% paraformaldehyde, stained with nuclear (TOPRO-3) and general protein (Eosin) fluorescent dyes, and optically cleared using a modified iDISCO protocol with ethyl cinnamate as the refractive index matching solution. Twenty tissue samples measuring 1-5 mm3 in volume were imaged at 2 µm/pixel resolution with a hybrid open-top light-sheet microscope. Smaller regions of interest (ROIs) with interesting pathologic features were re-imaged at higher resolution, 0.17 µm/pixel. Visualization and analysis was performed using Aivia software, custom spatial analysis software and python scripts. Results: All NSCLC tissue samples were successfully imaged in 3D at both low and high resolution. Computational approaches, such as machine learning, were used to segment and classify tumor cells and lymphocytes within the high resolution ROIs. In one example we quantified a 0.0045 mm3 ROI containing 1x109 voxels. We segmented 1,474 unique objects, and classified 683 (46%) of the objects as lymphocytes and 309 (20%) of the cells as tumor cells. We found that on average the lymphocytes were within 12 µm of the nearest tumor cell. Additionally the tissue was broken into separate lymphocyte dense (58% of volume) and tumor dense regions (42% of volume). Further analysis is in progress to quantify more features, such as tertiary lymphoid structures, stromal cells, and fibrosis over the full 20 sample dataset. Conclusions: We successfully implemented 3D machine learning analysis pipelines using images from intact NSCLC tissue samples measuring up to 5 mm3. The imaging was performed using our custom hybrid open-top light-sheet microscope. This novel method enables us to identify and quantify key histological features such as tumor cells and lymphocytes within the tumor microenvironment. In addition, machine learning driven algorithms clustered cells into distinct tissue regions in 3D which enables a more precise assessment of disease state and heterogeneity. Further investigation is ongoing to link these results to patient outcomes.

TRUSformer: improving prostate cancer detection from micro-ultrasound using attention and self-supervision.

A large body of previous machine learning methods for ultrasound-based prostate cancer detection classify small regions of interest (ROIs) of ultrasound signals that lie within a larger needle trace corresponding to a prostate tissue biopsy (called biopsy core). These ROI-scale models suffer from weak labeling as histopathology results available for biopsy cores only approximate the distribution of cancer in the ROIs. ROI-scale models do not take advantage of contextual information that are normally considered by pathologists, i.e., they do not consider information about surrounding tissue and larger-scale trends when identifying cancer. We aim to improve cancer detection by taking a multi-scale, i.e., ROI-scale and biopsy core-scale, approach. Our multi-scale approach combines (i) an "ROI-scale" model trained using self-supervised learning to extract features from small ROIs and (ii) a "core-scale" transformer model that processes a collection of extracted features from multiple ROIs in the needle trace region to predict the tissue type of the corresponding core. Attention maps, as a by-product, allow us to localize cancer at the ROI scale. We analyze this method using a dataset of micro-ultrasound acquired from 578 patients who underwent prostate biopsy, and compare our model to baseline models and other large-scale studies in the literature. Our model shows consistent and substantial performance improvements compared to ROI-scale-only models. It achieves [Formula: see text] AUROC, a statistically significant improvement over ROI-scale classification. We also compare our method to large studies on prostate cancer detection, using other imaging modalities. Taking a multi-scale approach that leverages contextual information improves prostate cancer detection compared to ROI-scale-only models. The proposed model achieves a statistically significant improvement in performance and outperforms other large-scale studies in the literature. Our code is publicly available at www.github.com/med-i-lab/TRUSFormer .

CT Imaging of Eustachian Tube Balloon Dilation: Method Development on Cadaver Heads.

Objective: To develop a methodology for the measurement of balloon dilation (BD) effects on Eustachian Tube (ET) structure using Computerized Tomography (CT) images. Methods: The BD of the ET was performed on three cadaver heads (five ears) through the nasopharyngeal orifice. The axial CT images of the temporal bones were obtained before dilation, while an inflated balloon was in the lumen of ET, and after balloon removal in each ear. Utilizing Dicom images captured by the ImageJ software 3D volume viewer function, the anatomical landmark coordinates of the ET were matched with their pre- and post-dilation counterparts, and the longitudinal axis of the ET was captured with serial images. The histograms of the regions of interest (ROI) and three different lumen width and length measurements were obtained from captured images. The densities of air, tissue, and bone were determined with histograms as a baseline to determine the BD rate as a function of increased air in the lumen. Results: The small ROI box included the area of prominently dilated ET lumen after BD and best represented the visually obvious changes in the lumen, compared to the ROIs that extended the wider areas (longest and longer). Air density was the outcome measure for comparison with each corresponding baseline value. The average increase in air density in the small ROI was 64%, while the longest and long ROI boxes showed 44 and 56% increases, respectively. Conclusion: This study describes a method to image the ET and quantify the outcomes of BD of the ET using anatomical landmarks.

Transfer function analysis assesses resting cerebral perfusion metrics using hypoxia-induced deoxyhemoglobin as a contrast agent.

Introduction: Use of contrast in determining hemodynamic measures requires the deconvolution of an arterial input function (AIF) selected over a voxel in the middle cerebral artery to calculate voxel wise perfusion metrics. Transfer function analysis (TFA) offers an alternative analytic approach that does not require identifying an AIF. We hypothesised that TFA metrics Gain, Lag, and their ratio, Gain/Lag, correspond to conventional AIF resting perfusion metrics relative cerebral blood volume (rCBV), mean transit time (MTT) and relative cerebral blood flow (rCBF), respectively. Methods: 24 healthy participants (17 M) and 1 patient with steno-occlusive disease were recruited. We used non-invasive transient hypoxia-induced deoxyhemoglobin as an MRI contrast. TFA and conventional AIF analyses were used to calculate averages of whole brain and smaller regions of interest. Results: Maps of these average metrics had colour scales adjusted to enhance contrast and identify areas of high congruence. Regional gray matter/white matter (GM/WM) ratios for MTT and Lag, rCBF and Gain/Lag, and rCBV and Gain were compared. The GM/WM ratios were greater for TFA metrics compared to those from AIF analysis indicating an improved regional discrimination. Discussion: Resting perfusion measures generated by The BOLD analysis resulting from a transient hypoxia induced variations in deoxyhemoglobin analyzed by TFA are congruent with those analyzed by conventional AIF analysis.

Eye-Lens Dose Reduction using Region of Interest (ROI) Attenuators in Neuroimaging.

Lens dose can be high during neuro-interventional procedures, increasing the risk of cataractogenesis. Although beam collimation can be effective in reducing lens dose, it also restricts the FOV. ROI imaging with a reduced-dose peripheral field permits full-field information with reduced lens dose. This work investigates the magnitude of lens-dose reduction possible with ROI imaging. EGSnrc Monte-Carlo calculations of lens dose were made for the Zubal head phantom as a function of gantry angulation and head shift from isocenter for both large and small FOV's. The lens dose for ROI attenuators of varying transmission was simulated as the weighted sum of the lens dose from the small ROI FOV and that from the attenuated larger FOV. Image intensity and quantum mottle differences between ROI and periphery can be equalized by image processing. The lens dose varies considerably with beam angle, head shift, and field size. For both eyes, the lens-dose reduction with an ROI attenuator increases with LAO angulation, being highest for lateral projections and lowest for PA. For an attenuator with small ROI field (5 × 5 cm) and 20% transmission, the lens dose for lateral projections is reduced by about 75% compared to a full dose 10 ×10 cm FOV, while the reduction ranges between 30 and 40% for PA projections. Use of ROI attenuators can substantially reduce the dose to the lens of the eye for all gantry angles and head shifts, while allowing peripheral information to be seen in a larger FOV.