Region Of Interest Segmentation Research Articles

Three-dimensional semi-supervised lumbar vertebrae region of interest segmentation based on MAE pre-training

Background: The annotation of the regions of interest (ROI) of lumbar vertebrae by radiologists for bone density assessment is a tedious and time-intensive task. However, deep learning (DL) methods for image segmentation has the potential to substitute manual annotations which can significantly improve the efficiency of clinical diagnostics. Objective: The paper proposes a semi-supervised three-dimensional (3D) segmentation method for the ROI of lumbar vertebrae by integrating the tube masking masked autoencoder (MAE) pre-training. Methods: The paper proposes a method that modifies the masking strategy of the original MAE pre-training network. And the pre-training network is only trained by images without segmentation labels, when the training is finished, the weights will be saved for segmentation tasks. In downstream tasks, a semi-supervised approach utilizing pseudo-label generation is employed for training. This method leverages a small amount of labeled data to achieve the segmentation of ROI of the lumbar vertebrae. Results: The experimental results demonstrate that under the condition of limited annotated data, the proposed network improves the dice coefficient by 5–7% and reduces the hausdorff distance by 0.2∼0.6 mm compared to using the UNetr network alone for segmentation. When compared to the conventional MAE, the tube masking MAE presented in this paper assists effectively in segmentation, resulting in a 2% increase in the dice coefficient and a 0.24 mm reduction in the hausdorff distance. Conclusion: Automatic segmentation of the ROI of the lumbar vertebrae helps to shorten the time for doctors to annotate vertebrae during clinical bone density examinations. The paper employs the tube masking MAE pre-trained model to effectively extract contextual information of the 3D lumbar vertebrae, combining it with a semi-supervised network leveraging pseudo-label generation for fine-tuning, which leads to effective 3D segmentation of the lumbar vertebrae.

Methodology for pediatric head computed tomography image segmentation and volumetric calculation using a tablet computer and stylus pen.

This study presents a MATrix LABoratory (MATLAB)-based methodology for calculating intracranial volumes from head computed tomography (CT) data and compares it with established methods. Regions of interest (ROI) were manually segmented on CT images using a stylus pen, facilitated by mirroring a computer desktop onto a tablet. The volumetric process involved three main steps: (1) calculating the volume of a single voxel, (2) counting the total number of voxels within the segmented ROI, and (3) multiplying this voxel count by the single-voxel volume. This method was applied to 83 pediatric head CT scans from patients with minor head trauma, and the volumetric results were compared with those obtained from OsiriX. A paired t-test revealed a statistically significant difference (p < 0.001) between volumes obtained with our MATLAB-based method and those from OsiriX, with our method measuring 0.32% higher. However, an unpaired t-test found no statistically significant differences between the volumetric population groups (p = 0.84). The significant difference identified by the paired t-test likely reflects statistical distinctions arising from differences in the calculation methods of the two approaches. Conversely, the unpaired t-test suggests no statistically detectable differences between the volumetric populations. Although this does not imply that the two methods produce identical results, the volumetric populations derived from our method may originate from the same underlying population as those obtained using OsiriX. By taking these points into account, our method has the potential to serve as a valuable tool for volumetric measurements.

Detecting Conjunctival Hyperemia Using an Effective Machine Learning based Method

Among the many complex sensory organs, the human eye stands out. Prevention of eye illnesses is of the utmost importance since irreversible vision loss might result from postponed treatment. Therefore, to manage the continuous course of eye illnesses, early identification and monitoring are vital. Hyperemia, or redness, from increased blood flow, is one sign of conjunctivitis, an eye disorder defined by inflammation of the conjunctiva. Many illnesses and problems are often curable or significantly reducible with the help of the greatest medicines, sophisticated procedures, and early, accurate diagnosis by medical experts. Because there is a severe lack of diagnostic specialists, patients with vision difficulties are often not given the care they need because their conditions are not properly diagnosed. Segmentation approaches are crucial for detecting and quantifying hyperemic areas, which are necessary for diagnosing and evaluating conjunctivitis. This paper presents a very efficient machine learning framework for the detection of eye problems. The system does this by integrating segmentation approaches with feature extraction techniques. The use of the discrete cosine transform (DCT) generates feature vectors from the segmented regions of interest. Random forests and neural networks are two machine learning classifiers that are learned using feature vectors. This approach has a 96% success rate and has potential to assist ophthalmologists in assessing the severity of illnesses in objective and accurate manner. This strategy might be advantageous for adopting ICT-based solutions in public health systems in remote regions.

Radiomics Analysis for the Identification of Invasive Pulmonary Subsolid Nodules From Longitudinal Presurgical CT Scans.

Effective identification of malignant part-solid lung nodules is crucial to eliminate risks due to therapeutic intervention or lack thereof. We aimed to develop delta radiomics and volumetric signatures, characterize changes in nodule properties over three presurgical time points, and assess the accuracy of nodule invasiveness identification when combined with immediate presurgical time point radiomics signature and clinical biomarkers. Cohort included 156 part-solid lung nodules with immediate presurgical CT scans and a subset of 122 nodules with scans at 3 presurgical time points. Region of interest segmentation was performed using ITK-SNAP, and feature extraction using CaPTk. Image parameter heterogeneity was mitigated at each time point using nested ComBat harmonization. For 122 nodules, delta radiomics features (ΔR AB = (R B -R A )/R A ) and delta volumes (ΔV AB = (V B -V A )/V A ) were computed between the time points. Principal Component Analysis was performed to construct immediate presurgical radiomics (Rs 1 ) and delta radiomics signatures (ΔRs 31 + ΔRs 21 + ΔRs 32 ). Identification of nodule pathology was performed using logistic regression on delta radiomics and immediate presurgical time point signatures, delta volumes (ΔV 31 + ΔV 21 + ΔV 32 ), and clinical variable (smoking status, BMI) models (train test split (2:1)). In delta radiomics analysis (n= 122 nodules), the best-performing model combined immediate pre-surgical time point and delta radiomics signatures, delta volumes, and clinical factors (classification accuracy [AUC]): (77.5% [0.73]) (train); (71.6% [0.69]) (test). Delta radiomics and volumes can detect changes in nodule properties over time, which are predictive of nodule invasiveness. These tools could improve conventional radiologic assessment, allow for earlier intervention for aggressive nodules, and decrease unnecessary intervention-related morbidity.

INNV-02. BRAINMETS: A CLINICIAN-FRIENDLY SOFTWARE TO MANUALLY SEGMENT BRAIN METASTASES AND DETECT PRIMARY TUMOR ORIGIN USING DEEP LEARNING

Abstract Estimates suggest that around 20%-40% of cancer patients are affected by brain metastases. A subset of these patients first present with central nervous system manifestations with an unknown primary. The issue then arises regarding which region of the body should be imaged, if not the entire body, exposing the patient to more radiation. To address this issue, streamline workflows, and save precious time, we developed a software clinicians can use to identify primary tumor locus. “BrainMets” is a user-friendly software that handles a DICOM input, allows the user to segment the region(s) of interest manually, and uses a deep learning model to predict the most likely tumor origin. The model is a mixed-effects neural network pre-trained on radiomic features from open-source data by Ocaña-Tienda et al. (2023) using 637 high-resolution imaging studies from 10 different medical centers with an accuracy of 90.5%, precision of 92.7%, recall of 87.8%, and F1 score of 90.2%. The software leverages the open-source software 3D-Slicer to help segment regions of interest, extract radiomic features, and make predictions. It is built using Flask, a Python package for backend development. The frontend user interface is created using HTML/CSS and Bootstrap. The “BrainMets” software thus strives to enhance diagnostic workflow significantly, reduce appointment times for scans, and improve patient experience. Future enhancements include integrating more AI models, improving healthcare IT system compatibility, and updating datasets to boost predictive accuracy and generalizability. This tool sets a new standard in personalized medicine, optimizing treatment strategies for cancer patients with brain metastases. As adoption increases, it could substantially improve patient outcomes and reduce healthcare costs associated with misdiagnosis and overtreatment. Further research and development will focus on expanding its applicability across various types of cancers.

Spatial Profiling of Ovarian Clear Cell Carcinoma Reveals Immune-Hot Features

Ovarian clear cell carcinoma (OCCC) has a high incidence in Asia, with a frequent occurrence at an early stage, but without sufficient data on molecular stratification for high-risk patients. Recently, immune-hot features have been proposed as indicators of poor prognosis in early stage OCCC. Specific patterns of intratumoral heterogeneity associated with immune-hot features must be defined. NanoString Digital Spatial Profiling technology (Cold spring biotech corp.) was used to decipher the spatial distribution of the 18-plex protein panel. Regions of interest (ROIs) were collected based on the reference hematoxylin and eosin–stained morphology. Areas of illumination (AOIs) were defined according to the ROI segmentation using the fluorescence signals of the visualization markers pan-cytokeratin (PanCK), CD45, or DNA. Unsupervised hierarchical clustering of 595 AOIs from 407 ROIs showed that the PanCK segments expressed different combinations of immune markers, suggestive of immune mimicry. The following 3 immune-hot clusters were identified: granzyme B–high, immune signal-high , and immune-like cells; the following 2 immune-cold clusters were identified: fibronectin-high and immune checkpoint–high cells. In tumor samples at the International Federation of Gynecology and Obstetrics stage IC1/2 experiencing recurrence, there was an increased occurrence of PanCK+ AOIs with immune signal-high and immune-like cell groups in the papillary morphology surrounded by macrophage lineage tumor-infiltrating immune cells (TIIs). In contrast, for tumor samples at the International Federation of Gynecology and Obstetrics stage IC3/II with recurrence, PanCK+ AOIs were prevalent in the fibronectin-high group, particularly in those with a tubulocystic morphology surrounded by lymphoid lineage non-TIIs. Our study on the spatial profiling of early stage OCCC tumors revealed that the immune mimicry of tumor cells, presence of TIIs, and morphologic patterns were associated with recurrence, which switched during tumor progression.

A Two-Phase Methodology leveraging X-DenseNet for Multi-Organ Segmentation in Abdominal CT Scans

Introduction: Multi organ segmentation (MOS) is crucial for various medical applications such as disease diagnosis, treatment planning, and surgery. Manual segmentation of CT organs is inefficient, laborious and time-consuming, necessitating the development of automated techniques. To address these challenges, automatic techniques based on deep learning models are explored in this field. These techniques require improvements to deal with patient-to-patient variability in organ size, location, and shape. Despite the transformative potential of deep learning techniques, research studies often inadequately address low contrast and overlapping structures within CT scans. This paper proposes a novel two phase methodology to integrate deep learning models and image enhancement strategies. Methods: Our framework contributes in two stages named (a) Optimized Contrast Limited Adaptive Histogram Equalization-Weighted Grey Wolf Optimization (optiCLAHE-Weighted GWO), where the organs contrast is enhanced by employing weighted GWO for CLAHE clip limit selection (b) X-DenseNet architecture, where segmented regions of interest are produced with X-DenseNet model from the enhanced CT. Results: For validation of proposed multi organ segmentation (PMOS) technique, experimentation and comparative analysis with existing models has been conducted on FLARE 22 challenge dataset. The Dice Score (DSC) of liver, aorta and spleen for proposed Multi organ segmentation (PMOS) approach is 90.4%, 99.38% and 98.33% respectively. The mean of Precision (Pre), Accuracy (Acc), F-score and DSC of the proposed approach are 95.05%, 95.55%, 95.29% and 96.06% respectively.

Dual-Tree Complex Wavelet Pooling and Attention-Based Modified U-Net Architecture for Automated Breast Thermogram Segmentation and Classification.

Thermography is a non-invasive and non-contact method for detecting cancer in its initial stages by examining the temperature variation between both breasts. Preprocessing methods such as resizing, ROI (region of interest) segmentation, and augmentation are frequently used to enhance the accuracy of breast thermogram analysis. In this study, a modified U-Net architecture (DTCWAU-Net) that uses dual-tree complex wavelet transform (DTCWT) and attention gate for breast thermal image segmentation for frontal and lateral view thermograms, aiming to outline ROI for potential tumor detection, was proposed. The proposed approach achieved an average Dice coefficient of 93.03% and a sensitivity of 94.82%, showcasing its potential for accurate breast thermogram segmentation. Classification of breast thermograms into healthy or cancerous categories was carried out by extracting texture- and histogram-based features and deep features from segmented thermograms. Feature selection was performed using Neighborhood Component Analysis (NCA), followed by the application of machine learning classifiers. When compared to other state-of-the-art approaches for detecting breast cancer using a thermogram, the proposed methodology showed a higher accuracy of 99.90% for VGG16 deep features with NCA and Random Forest classifier. Simulation results expound that the proposed method can be used in breast cancer screening, facilitating early detection, and enhancing treatment outcomes.

Reimagining otitis media diagnosis: A fusion of nested U-Net segmentation with graph theory-inspired feature set

Otitis media (OM) is a common infection or inflammation of the middle ear causing conductive hearing loss that primarily affects children and may delay speech, language, and cognitive development. OM can manifest itself in different forms, and can be diagnosed using (video) otoscopy (visualizing the tympanic membrane) or (video) pneumatic otoscopy and tympanometry. Accurate diagnosis of OM is challenging due to subtle differences in otoscopic features. This research aims to develop an automated computer-aided design (CAD) system to assist clinicians in diagnosing OM using otoscopy images. The ground truths, generated manually and validated by otolaryngologists, are utilized to train the proposed nested U-Net++ model. Ten clinically relevant gray level co-occurrence matrix (GLCM) and morphological features were extracted from the segmented Region of Interest (ROI) and validated for OM classification based on a statistical significance test. These features serve as input for a Graph Neural Network (GNN) model, the base model in our research. An optimized GNN model is proposed after ablation study of the base model. Three datasets, one private dataset, and two public ones have been used, where the private dataset is utilized for both training and testing, and the public datasets are used to test the robustness of the proposed GNN model only. The proposed GNN model obtained the highest accuracy in diagnosing OM: 99.38 %, 93.51 %, and 91.38 % for the private dataset, public dataset1, and public dataset2, respectively. The proposed methodology and results of this research might enhance clinicians' effectiveness in diagnosing OM.

Reliability of quantitative magnetic susceptibility imaging metrics for cerebral cortex and major subcortical structures.

Susceptibility estimates derived from quantitative susceptibility mapping (QSM) images for the cerebral cortex and major subcortical structures are variably reported in brain magnetic resonance imaging (MRI) studies, as average of all ( ), absolute ( ), or positive- ( ) and negative-only ( ) susceptibility values using a region of interest (ROI) approach. This pilot study presents a reliability analysis of currently used ROI-QSM metrics and an alternative ROI-based approach to obtain voxel-weighted ROI-QSM metrics ( and ). Ten healthy subjects underwent repeated (test-retest) 3-dimensional multi-echo gradient-echo (3DMEGE) 3 Tesla MRI measurements. Complex-valued 3DMEGE images were acquired and reconstructed with slice thicknesses of 1 and 2mm (3DMEGE1, 3DMEGE2) along with 3DT1-weighted isometric (voxel 1 mm3) images for independent registration and ROI segmentation. Agreement, consistency, and reproducibility of ROI-QSM metrics were assessed through Bland-Altman analysis, intraclass correlation coefficient, and interscan and intersubject coefficient of variation (CoV). All ROI-QSM metrics exhibited good to excellent consistency and test-retest agreement with no proportional bias. Interscan CoV was higher for in comparison to the other metrics where it was below 15%, in both 3DMEGE1 and 3DMEGE2 datasets. Intersubject CoV for and exceeded 50% in all ROIs. Among the evaluated ROI-QSM metrics, and estimates were less reliable, whereas separating positive and negative values (using ) improved the reproducibility within, and the comparability between, subjects, even when reducing the slice thickness. These preliminary findings may offer valuable insights toward standardizing ROI-QSM metrics across different patient cohorts and imaging settings in future clinical MRI studies.

Ultrasonic Assessment of Liver Fibrosis Using One-Dimensional Convolutional Neural Networks Based on Frequency Spectra of Radiofrequency Signals with Deep Learning Segmentation of Liver Regions in B-Mode Images: A Feasibility Study.

The early detection of liver fibrosis is of significant importance. Deep learning analysis of ultrasound backscattered radiofrequency (RF) signals is emerging for tissue characterization as the RF signals carry abundant information related to tissue microstructures. However, the existing methods only used the time-domain information of the RF signals for liver fibrosis assessment, and the liver region of interest (ROI) is outlined manually. In this study, we proposed an approach for liver fibrosis assessment using deep learning models on ultrasound RF signals. The proposed method consisted of two-dimensional (2D) convolutional neural networks (CNNs) for automatic liver ROI segmentation from reconstructed B-mode ultrasound images and one-dimensional (1D) CNNs for liver fibrosis stage classification based on the frequency spectra (amplitude, phase, and power) of the segmented ROI signals. The Fourier transform was used to obtain the three kinds of frequency spectra. Two classical 2D CNNs were employed for liver ROI segmentation: U-Net and Attention U-Net. ROI spectrum signals were normalized and augmented using a sliding window technique. Ultrasound RF signals collected (with a 3-MHz transducer) from 613 participants (Group A) were included for liver ROI segmentation and those from 237 participants (Group B) for liver fibrosis stage classification, with a liver biopsy as the reference standard (Fibrosis stage: F0 = 27, F1 = 49, F2 = 51, F3 = 49, F4 = 61). In the test set of Group A, U-Net and Attention U-Net yielded Dice similarity coefficients of 95.05% and 94.68%, respectively. In the test set of Group B, the 1D CNN performed the best when using ROI phase spectrum signals to evaluate liver fibrosis stages ≥F1 (area under the receive operating characteristic curve, AUC: 0.957; accuracy: 89.19%; sensitivity: 85.17%; specificity: 93.75%), ≥F2 (AUC: 0.808; accuracy: 83.34%; sensitivity: 87.50%; specificity: 78.57%), and ≥F4 (AUC: 0.876; accuracy: 85.71%; sensitivity: 77.78%; specificity: 94.12%), and when using the power spectrum signals to evaluate ≥F3 (AUC: 0.729; accuracy: 77.14%; sensitivity: 77.27%; specificity: 76.92%). The experimental results demonstrated the feasibility of both the 2D and 1D CNNs in liver parenchyma detection and liver fibrosis characterization. The proposed methods have provided a new strategy for liver fibrosis assessment based on ultrasound RF signals, especially for early fibrosis detection. The findings of this study shed light on deep learning analysis of ultrasound RF signals in the frequency domain with automatic ROI segmentation.

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.

A Modified Bio-Inspired Optimizer with Capsule Network for Diagnosis of Alzheimer Disease

Recently, Alzheimer’s disease (AD) is one of the common neurodegenerative disorders, which primarily occurs in old age. Structural magnetic resonance imaging (sMRI) is an effective imaging technique used in clinical practice for determining the period of AD patients. An efficient deep learning framework is proposed in this paper for AD detection, which is inspired from clinical practice. The proposed deep learning framework significantly enhances the performance of AD classification by requiring less processing time. Initially, in the proposed framework, the sMRI images are acquired from a real-time dataset and two online datasets including Australian Imaging, Biomarker and Lifestyle flagship work of ageing (AIBL), and Alzheimer’s Disease Neuroimaging Initiative (ADNI). Next, a fuzzy-based superpixel-clustering algorithm is introduced to segment the region of interest (RoI) in sMRI images. Then, the informative deep features are extracted in segmented RoI images by integrating the probabilistic local ternary pattern (PLTP), ResNet-50, and Visual Geometry Group (VGG)-16. Furthermore, the dimensionality reduction is accomplished by through the modified gorilla troops optimizer (MGTO). This process not only enhances the classification performance but also diminishes the processing time of the capsule network (CapsNet), which is employed to classify the classes of AD. In the MGTO algorithm, a quasi-reflection-based learning (QRBL) process is introduced for generating silverback’s quasi-refraction position for further improving the optimal position’s quality. The proposed fuzzy based superpixel-clustering algorithm and MGTO-CapsNet model obtained a pixel accuracy of 0.96, 0.94, and 0.98 and a classification accuracy of 99.88%, 96.38%, and 99.94% on the ADNI, real-time, and AIBL datasets, respectively.

Infrared thermogram image guided discontinuous appearances of hyaluronic acid for classification of arthritic knee joints

The liveliness of a human potentially depends on his/her smooth movability. To accomplish the work of daily life, the joints of the body need to be healthy. However, the occurrence of Rheumatoid arthritis and Osteoarthritis has a significant prevalence towards the immovability of humankind. Rheumatoid arthritis (RA) and Osteoarthritis (OA) mostly affect the joints of the hand and knee which result in lifelong pain, inability to climb, walk, etc. In the early stages, these diseases attack the synovial membrane and synovial fluid, and further it destroys the soft tissues and bone structure. By early diagnosis, we can start the treatment in the early stage which may cure these diseases with such extreme consequences. As per clinical studies of previous literature, it is observed that synovial fluid imbalance appears in the early stage of such diseases and Hyaluronic Acid (HA) concentration also decreases for that. Therefore, estimation of HA is a significant key to arthritis disease classification and grading. In this paper, we proposed a hybrid framework for classification of arthritic knee joints based on the analysis of the discontinuous appearances of the HA concentration using infrared imaging technology. To meet up the specific necessities, firstly we have proposed a modified K-Means clustering algorithm for extraction of the region of interest (ROI) i.e., the knee joint surface. Secondly, a mathematical formulation is proposed to calculate the concentration of HA from the segmented ROIs. This experimental process was implemented on the publicly available IR (Infrared) Knee Joint Dataset and for further evaluation of the novelty of mathematical formulation, we have extended the proposed work to the classification of healthy and arthritis affected knee joints depending on significant discriminative characteristics of the HA concentration with respect to the existing significant imaging features. Experimental results and analysis demonstrates that concentration of HA has the dominant potential for classifying healthy and arthritic knee joints using infrared holistic images. Our experimental analysis reveals that estimation and combination of the HA concentration features with conventional handcrafted and deep features increases the classification performance with an average accuracy of 91% and 97.22% respectively as compared to the each individual feature sets.

Quantifying lung fissure integrity using a three-dimensional patch-based convolutional neural network on CT images for emphysema treatment planning.

Evaluation of lung fissure integrity is required to determine whether emphysema patients have complete fissures and are candidates for endobronchial valve (EBV) therapy. We propose a deep learning (DL) approach to segment fissures using a three-dimensional patch-based convolutional neural network (CNN) and quantitatively assess fissure integrity on CT to evaluate it in subjects with severe emphysema. From an anonymized image database of patients with severe emphysema, 129 CT scans were used. Lung lobe segmentations were performed to identify lobar regions, and the boundaries among these regions were used to construct approximate interlobar regions of interest (ROIs). The interlobar ROIs were annotated by expert image analysts to identify voxels where the fissure was present and create a reference ROI that excluded non-fissure voxels (where the fissure is incomplete). A CNN configured by nnU-Net was trained using 86 CT scans and their corresponding reference ROIs to segment the ROIs of left oblique fissure (LOF), right oblique fissure (ROF), and right horizontal fissure (RHF). For an independent test set of 43 cases, fissure integrity was quantified by mapping the segmented fissure ROI along the interlobar ROI. A fissure integrity score (FIS) was then calculated as the percentage of labeled fissure voxels divided by total voxels in the interlobar ROI. Predicted FIS (p-FIS) was quantified from the CNN output, and statistical analyses were performed comparing p-FIS and reference FIS (r-FIS). The absolute percent error mean (±SD) between r-FIS and p-FIS for the test set was 4.0% (), 6.0% (), and 12.2% () for the LOF, ROF, and RHF, respectively. A DL approach was developed to segment lung fissures on CT images and accurately quantify FIS. It has potential to assist in the identification of emphysema patients who would benefit from EBV treatment.

Intraoperative OCT-Guided Volumetric Measurements of Subretinal Therapy Delivery in Humans.

Purpose: To evaluate a recently developed technique using intraoperative optical coherence tomography (OCT) to measure subretinal tissue plasminogen activator (tPA) volumes in patients with submacular hemorrhage secondary to exudative age-related macular degeneration (AMD). Methods: Three patients (72 to 83 years old) had 25-gauge pars plana vitrectomy, subretinal tPA, and a partial gas fill. An investigational intraoperative OCT system with a modified widefield noncontact indirect viewing apparatus was used to image subretinal tPA blebs. Using the recently developed technique, the volume and surface area in the segmented region of interest were determined. Results: In each case, the delivered tPA volume measured from the syringe differed from the intraoperative OCT-measured subretinal tPA volume: Patient 1, 130 µL from syringe, 118 µL based on intraoperative OCT, 9% difference; Patient 2, 140 µL, 50 µL, 64%; Patient 3, 110 µL, 122 µL, 11%. The total bleb surface area was 129 mm2 in Patient 1, 55 mm2 in Patient 2, and 106 mm2 in Patient 3. Conclusions: This was the first human study to implement and evaluate intraoperative OCT image-based methods to obtain volumetric bleb measurements in patients receiving subretinal tPA for exudative AMD. This proof-of-concept study showed that intraoperative OCT-obtained bleb volume differed from intraoperative recordings, which could be explained by tPA delivery into the vitreous, efflux through the retinotomy, or human error. Intraoperative OCT can provide visualization and quantification of subretinal tPA bleb volume and surface area, which has implications for improved safety, efficacy, and analysis of the effects of subretinal drug delivery.

Abstract 4172: Pathology supervised approaches for ROI placement and segmentation to overcome caveats and pitfalls in high-plex spatial profiling

Abstract Background: Tissue-based high-plex platforms evaluate numerous biomarkers in limited tissue sections. Certain tissue features and assay technical limitations may jeopardize the accuracy of this analysis. In this study, we aim to determine tissue specific features that influence image-based region of interest (ROI) selection and segmentation for spatial profiling of carcinomas, and to develop pathology approaches to overcome them. Methodology: A total of 317 samples from different carcinoma types processed with the GeoMx Digital spatial profiling protein assay protocol were assessed to identify deviations from original strategies for ROI placement and segmentation (Table 1). ROI were selected by a pathologist. The initial strategy for all cases was to place rectangles ROIs of 660*785 µm in tumor area and to segment based on morphology biomarkers as indicated in Table 1. Strategies to overcome technical and tissue specific features were recorded and quantified. Results: A total of 2068 ROIs and 4422 AOIs were evaluated. In all the cases, tissue displayed specific features that limited planned ROI placement and segmentation. Table 1 describes these features by tissue type and profiling strategies. A total of 970 ROIs were solved using different annotation type or changes in segmentation strategy, 584 bypassed the limitations by changing the processing protocol; and 322 were not solved due to non-compliance with assay requirements. Conclusions: We identified tissue features that limit the planned ROI selection and segmentation in spatial high-plex profiling, mainly due to tissue processing, distinct histology, type of sample, and presence of adjacent non-neoplastic tissue. A deep understanding of tumor architecture, biology and platform related technical limitations is needed to select areas of analysis for spatial profiling. Table 1. ROI and segmentation strategy methodology and tissue specific features with pathology appro Histological type (Profiling segments) Anal Carcinoma (Tu/TME) Lung NSCLC (Tu/TME) Lung NSCLC TMA (Tu/Tcells/Mϕ) Lung NSCLC TMA (Tu/TME) CRC MT Tu/Immune Breast Ca Tu/Immune/else Pathology Approach (ROIs) Methodology Samples 48 33 80 80 6 70 Biopsy type Incisional (Core/other), Whole tissue Incisional (Core/other)Whole tissue TMA TMA Whole tissue Core Morphology markers/mIF Channels Syto13: FITC/525 PanCK: Cy3/568 CD45: Texas red/615 CD68: Cy5/666 Syto13: FITC/525PanCK: Cy3/568CD20: Texas red/615CD3: Cy5/666 Syto13: FITC/525 PanCK: Cy3/568 CD68: Texas red/615 CD3: Cy5/666 Syto13: FITC/525 PanCK: Cy3/568 CD68: Texas red/615 CD3: Cy5/666 Syto13: FITC/525 PanCK: CD45: Texas red/615 Syto13: FITC/525 PanCK: Cy3/568 CD45: Texas red/615 Strategy for segmentation Tumor (PanCK+) TME (PanCK-) Tumor (PanCK+)TME (PanCK-) Tumor (PanCK+) T cell (CD3+) Mϕ (CD68+) Tumor (PanCK+) TME (PanCK-) Tumor (PanCK+) Immune (CD45+) Tumor (PanCK+) Immune (CD45+) Else (PanCK-CD45-) Total ROIs/AOIs 224/435 144/252 80/218 82/162 944/1573 594/1782 Tissue specific features (ROI number/percentage) Folds (ROIs) 10 1 0 0 9 5 Draw Polygons (25) Adjacent* Non-Tumoral tissue 22 7 9 7 388 40 Draw Polygons (473) Tissue Derived Autofluorescence 3 105 0 1 0 45 Draw Polygons (13). New segment to exclude autofluorescence (141) Sample Size 26 19 1 2 5 138 Polygons (191) Necrosis 13 0 1 1 217 6 Polygons (238) Biological absence of morphology marker 10 0 0 0 0 5 Polygons non segmented for Tu and TME (15) Solid tumors with scant TME 0 21 0 0 0 9 Polygons non segmented for Tu and TME (30) Eosin autofluorescence 0 0 0 0 0 584 Stained new section with Validated PanCK in Cy5/666 channel (584) Failure to reach the minimum number of cell requirements 13 36 22 2 249 43 Excluded segments (322) New scan to select more areas (43) Citation Format: Sharia Hernandez, Claudio J. Arrechedera, Pedro Rocha, Larisa Kostousov, Sean Barnes, Khan Khaja, Luisa M. Solis-Soto. Pathology supervised approaches for ROI placement and segmentation to overcome caveats and pitfalls in high-plex spatial profiling [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 4172.

Abstract 3529: Leveraging deep learning for fully automated analysis of pre-clinical mouse positron emission tomography

Abstract Turning early-cancer diagnosis into actionable treatment often relies on highly accurate and sensitive imaging techniques to guide surgical intervention or to monitor therapeutic efficacy. Earli is developing a highly sensitive, orthogonal approach that uses genetic constructs to usurp dysregulated cancer pathways to force the tumor to produce a PET reporter gene, enabling the use of radiotracers amenable to positron emission tomography/computed tomography (PET/CT) to guide precision imaging. However, analysis of PET/CT images using standard manual and semi-automated methods is challenging resulting in reduced experimental throughput and user variability, highlighting the need for a fully automated PET/CT analysis pipeline using Deep Learning (DL). To increase pre-clinical throughput and reduce inter-user variability, we developed an automated DL processing pipeline in Python to perform three major tasks: separation of multi-mouse bed data for PET uptake quantification by mouse, segmentation of tumors and background organs, and quantification of PET signal for PK/PD analysis. Preclinical PET/CT data was acquired using a 4 animal, multi-mouse bed and separated into individual mice by co-registering the reconstructed CT image against a multi-mouse bed reference mask using a rigid 3D Euler transformation algorithm and cropping at fixed indices. Following mouse separation, a 3D nnUNet architecture is used to perform semantic segmentation of regions of interest (ROI) on CT for bi-hemispheric subcutaneous tumors, lungs, liver, kidneys, spleen, and the bladder. nnUNet was trained on PET/CT of 311 mice consisting of a mixture of naïve and tumor-bearing animals subcutaneously implanted with H1299 cells using a five-fold cross-validation strategy with 1000 epochs per fold. Model performance was evaluated by comparing the network prediction to the reference manually annotated masks using the Dice coefficient. PET uptake is reported as percent injected dose per milliliter (%ID/mL) or standardized uptake value (SUV). Segmentation performance results for the hold-out set identified mean Dice scores greater than 0.80 for all ROIs, reflecting the similarity between model predictions and human annotations. The automated pipeline reduced the analysis time from 5.5-6 hours per mouse when using traditional manual/semi-automatic approaches to approximately 15 minutes. In summary, we developed a fully automated DL-based PET/CT quantification pipeline for pre-clinical mouse studies that provides accurate ROI segmentation and PET uptake quantification, significantly increasing experimental throughput and reducing inter-user variability. Further improvements to model inference include loss functions weighted for poor performing ROIs and addition of lung nodule segmentation for eventual translation to humans as an accompaniment to Earli’s theranostic programs. Citation Format: Hung-Yu Henry Lee, Mohammed Goryawala, Tim Sproul, Maggie Louie, David Suhy. Leveraging deep learning for fully automated analysis of pre-clinical mouse positron emission tomography [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 3529.

An efficient breast cancer classification model using bilateral filtering and fuzzy convolutional neural network

BC (Breast cancer) is the second most common reason for women to die from cancer. Recent workintroduced a model for BC classifications where input breast images were pre-processed using median filters for reducing noises. Weighed KMC (K-Means clustering) is used to segment the ROI (Region of Interest) after the input image has been cleaned of noise. Block-based CDF (Centre Distance Function) and CDTM (Diagonal Texture Matrix)-based texture and shape descriptors are utilized for feature extraction. The collected features are reduced in counts using KPCA (Kernel Principal Component Analysis). The appropriate feature selection is computed using ICSO (Improved Cuckoo Search Optimization). The MRNN ((Modified Recurrent Neural Network)) values are then improved through optimization before being utilized to divide British Columbia into benign and malignant types. However, ICSO has many disadvantages, such as slow search speed and low convergence accuracy and training an MRNN is a completely tough task. To avoid those problems in this work preprocessing is done by bilateral filtering to remove the noise from the input image. Bilateral filter using linear Gaussian for smoothing. Contrast stretching is applied to improve the image quality. ROI segmentation is calculated based on MFCM (modified fuzzy C means) clustering. CDTM-based, CDF-based color histogram and shape description methods are applied for feature extraction. It summarizes two important pieces of information about an object such as the colors present in the image, and the relative proportion of each color in the given image. After the features are extracted, KPCA is used to reduce the size. Feature selection was performed using MCSO (Mutational Chicken Flock Optimization). Finally, BC detection and classification were performed using FCNN (Fuzzy Convolutional Neural Network) and its parameters were optimized using MCSO. The proposed model is evaluated for accuracy, recall, f-measure and accuracy. This work’s experimental results achieve high values of accuracy when compared to other existing models.

Diffusion tensor imaging metrics as natural markers of multiple sclerosis-induced brain disorders with a low Expanded Disability Status Scale score

Non-invasive and effective differentiation along with determining the degree of deviations compared to the healthy cohort is important in the case of various brain disorders, including multiple sclerosis (MS). Evaluation of the effectiveness of diffusion tensor metrics (DTM) in 3T DTI for recording MS-related deviations was performed using a time-acceptable MRI protocol with unique comprehensive detection of systematic errors related to spatial heterogeneity of magnetic field gradients. In a clinical study, DTMs were acquired in segmented regions of interest (ROIs) for 50 randomly selected healthy controls (HC) and 50 multiple sclerosis patients. Identical phantom imaging was performed for each clinical measurement to estimate and remove the influence of systematic errors using the b-matrix spatial distribution in the DTI (BSD-DTI) technique. In the absence of statistically significant differences due to age in healthy volunteers and patients with multiple sclerosis, the existence of significant differences between groups was proven using DTM. Moreover, a statistically significant impact of spatial systematic errors occurs for all ROIs and DTMs in the phantom and for approximately 90 % in the HC and MS groups. In the case of a single patient measurement, this appears for all the examined ROIs and DTMs. The obtained DTMs effectively discriminate healthy volunteers from multiple sclerosis patients with a low mean score on the Expanded Disability Status Scale. The magnitude of the group differences is typically significant, with an effect size of approximately 0.5, and similar in both the standard approach and after elimination of systematic errors. Differences were also observed between metrics obtained using these two approaches. Despite a small alterations in mean DTMs values for groups and ROIs (1–3 %), these differences were characterized by a huge effect (effect size ∼0.8 or more). These findings indicate the importance of determining the spatial distribution of systematic errors specific to each MR scanner and DTI acquisition protocol in order to assess their impact on DTM in the ROIs examined. This is crucial to establish accurate DTM values for both individual patients and mean values for a healthy population as a reference. This approach allows for an initial reliable diagnosis based on DTI metrics.