Image Analysis Platform Research Articles

Automated Diagnosis of Knee Osteoarthritis Using ResNet101 on a DEEP:PHI: Leveraging a No-Code AI Platform for Efficient and Accurate Medical Image Analysis

Background: Knee osteoarthritis (OA) is a prevalent degenerative joint disease significantly impacting global health. Early and accurate diagnosis is crucial for effective management, but traditional methods often rely on subjective assessments. This study evaluates the efficacy of a deep learning model implemented through a no-code AI platform for diagnosing and grading knee OA from plain radiographs. Methods: We utilized the Osteoarthritis Initiative (OAI) dataset, comprising knee X-ray data from 1526 patients. The data were split into training (47.0%), validation (26.5%), and test (26.5%) sets. We employed a ResNet101 model on the DEEP:PHI no-code AI platform for image analysis. The model was trained to classify knee OA into five grades (0–4) based on the Kellgren–Lawrence scale. Results: Our AI model demonstrated high accuracy in distinguishing between different OA grades, with particular strength in early-stage detection. The model achieved optimal performance at 20 epochs, suggesting efficient learning dynamics. Grad-CAM visualizations were used to enhance the interpretability of the model’s decision-making process. Conclusions: This study demonstrates the potential of AI, implemented through a no-code platform, to accurately diagnose and grade knee OA from radiographs. The use of a no-code AI platform such as DEEP:PHI represents a step towards democratizing AI in healthcare, enabling the rapid development and deployment of sophisticated medical AI applications without extensive coding expertise. This approach could significantly enhance the early detection and management of knee OA, potentially improving patient outcomes and streamlining clinical workflows.

Single Vesicle Surface Protein Profiling and Machine Learning-Based Dual Image Analysis for Breast Cancer Detection.

Single-vesicle molecular profiling of cancer-associated extracellular vesicles (EVs) is increasingly being recognized as a powerful tool for cancer detection and monitoring. Mask and target dual imaging is a facile method to quantify the fraction of the molecularly targeted population of EVs in biofluids at the single-vesicle level. However, accurate and efficient dual imaging vesicle analysis has been challenging due to the interference of false signals on the mask images and the need to analyze a large number of images in clinical samples. In this work, we report a fully automatic dual imaging analysis method based on machine learning and use it with dual imaging single-vesicle technology (DISVT) to detect breast cancer at different stages. The convolutional neural network Resnet34 was used along with transfer learning to produce a suitable machine learning model that could accurately identify areas of interest in experimental data. A combination of experimental and synthetic data were used to train the model. Using DISVT and our machine learning-assisted image analysis platform, we determined the fractions of EpCAM-positive EVs and CD24-positive EVs over captured plasma EVs with CD81 marker in the blood plasma of pilot HER2-positive breast cancer patients and compared to those from healthy donors. The amount of both EpCAM-positive and CD24-positive EVs was found negligible for both healthy donors and Stage I patients. The amount of EpCAM-positive EVs (also CD81-positive) increased from 18% to 29% as the cancer progressed from Stage II to III. No significant increase was found with further progression to Stage IV. A similar trend was found for the CD24-positive EVs. Statistical analysis showed that both EpCAM and CD24 markers can detect HER2-positive breast cancer at Stages II, III, or IV. They can also differentiate individual cancer stages except those between Stage III and Stage IV. Due to the simplicity, high sensitivity, and high efficiency, the DISVT with the AI-assisted dual imaging analysis can be widely used for both basic research and clinical applications to quantitatively characterize molecularly targeted EV subtypes in biofluids.

Abstract C120: Characterization of the inflammatory tumor microenvironment and its association with cancer health disparities

Abstract Black persons have the highest mortality rates of both cervical cancer (CC) and pancreatic ductal adenocarcinoma (PDAC) in the United States. Differences in access to quality healthcare and socioeconomic status contribute to these cancer health disparities as well as ancestry and biology. In other cancer types, differences in the inflammatory tumor landscape across races and ancestral populations have been reported. However, the inflammatory tumor microenvironment (TME) as it relates to CC and PDAC health disparities is not well defined. Black patients are often underrepresented in PDAC studies. Meanwhile, CC health disparities studies primarily investigate socioeconomic status, screening, and vaccination rates. Therefore, we aim to characterize the inflammatory TME in these two cancer types with the long-term goal of informing prevention, prognostic, and therapeutic strategies. Clinicopathological characteristics were compared between Black (n = 144) and White (n = 198) patients with CC who received care at Johns Hopkins Medicine between 1995 and 2019. Surgically resected formalin-fixed paraffin embedded (FFPE) CC tissue samples were obtained from Non-Hispanic Black (n = 10), Non-Hispanic White (n = 10), Hispanic Black (n = 5), and Hispanic White (n = 5) women matched on grade, age, and subtype. Tissue microarrays (TMAs) were constructed using archival surgically resected FFPE PDAC tissues from Black (n = 45) and White (n = 45) patients, matched on sex, age, stage, and time of surgery. Multiplexing immunohistochemistry (mIHC) was used to assess markers for i) T cells (CD3, CD8, CD4, FoxP3), M2 macrophages (CD163), mast cells (Tryptase), and fibroblasts (FAP) in PDAC tissues and ii) B cells (CD79A), T cells (CD3, CD8, CD4, FoxP3), plasma cells (CD138/syndecan-1), programmed death-ligand 1 protein (PD-L1), macrophages (CD68, CD163), natural killer cells (NKP46/CD56/CD16), neutrophils (CD66ce), and mast cells (Tryptase) in CC tissues. HALO Image Analysis Platform (version 3.6.4134.137; Indica Labs) was used to quantify marker expression. GraphPad Prism Software (version 10.0.03), was used for graphing and statistical analysis. Squamous cell CC was the most frequent histological subtype in both Black (68%) and White (44%) women. Progression through treatment was more common among Black women (24%) compared to White women (7%). CD163+ M2 macrophages were significantly increased in PDAC tumors compared to benign tissues among Black patients (p = 0.0435), and in tumors from Black women compared to White women (p = 0.0067). FAP+ fibroblasts were also significantly increased in PDAC tumors compared to benign tissues among Black patients (p = 0.0001) and in tumors from Black women compared to White women (p = 0.0013). CD8+ cytotoxic T cells and FoxP3+ regulatory T cells were comparable between races. We observed differences in macrophage and fibroblast distribution between races supporting the effort for future studies that examine the significance of characterizing the TME to address cancer health disparities. Citation Format: Jelani Jarrett, Jae W Lee, Ali Dbouk, Vahinipriya Manoharan, Kennedy Rains, Terri Mason, Kimberly Levinson, Laura D Wood, Michael G Goggins, Janielle P Maynard. Characterization of the inflammatory tumor microenvironment and its association with cancer health disparities [abstract]. In: Proceedings of the 17th AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2024 Sep 21-24; Los Angeles, CA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2024;33(9 Suppl):Abstract nr C120.

Abstract C122: Androgen receptor-mediated regulation of innate immunity in prostate cancer in Black men

Abstract Black men are two times more likely to die from prostate cancer (PCa) than White men. We reported that innate immune cell infiltration varies between races, and others have reported differences in androgen receptor (AR) expression by race or ancestry. Herein, we evaluated AR status and immune cell infiltration in primary and metastatic PCa tissues obtained from Black and White men. RNA in situ hybridization (RISH) AR analysis was done on a tissue microarray (TMA) constructed from radical prostatectomy tissues from 120 Black men matched to 60 White men on age, grade and stage. RISH or immunohistochemistry (IHC) analysis of AR, NKX3.1, prostate specific antigen (PSA), mast cells (tryptase), neutrophils (CD66), and macrophages (CD206, CD68, CD80) was done on a TMA containing rapid autopsy tissues from metastatic sites (n=11) collected from Black (n=5) and White (n=16) men. Multiplexing IHC (mIHC) was used to assess AR, neutrophils (CD66ce), mast cells (tryptase), T-cells (CD4, CD8), and M2 macrophages (CD163) on radical prostatectomy whole-tissue sections obtained at Johns Hopkins Medicine, Baltimore, MD and biopsy tissues obtained from the University of Nairobi, Nairobi, Kenya. Tumor regions on TMAs were annotated by a pathologist and signal intensity per unit area measured using TMAJ and FrIDA software. HALO Image Analysis Platform (version 3.6.4134.137; Indica Labs) was used to analyze mIHC stained tissues. Statistical analysis and graphing were done using GraphPad Prism. There was increased AR mRNA expression in primary cancer tissues from Black men compared to White men (p<0.0001) after negative binomial regression with the adjustment of TMA set, age, grade and stage. There was a negative correlation between AR expression and CD66+ neutrophils in tumor and benign tissues from Black men (R = -0.318;p = 0.002) but this correlation was not significant among White men (R= -0.045; p = 0.765). Interestingly, we measured a positive correlation between AR mRNA expression and tryptase+ mast cells among White men (R= 0.320; p=0.016), but a trend toward a negative relationship among Black men (R= -0.252;p = 0.075) We also measured a positive correlation between AR and CD163+(R= 0.324; p<0.001), CD68+(R= 0.514;p<0.001), and CD80+(R= 0.386;p<0.001) macrophages in primary tumor tissues from all men. NKX3.1 and PSA protein were significantly increased in metastatic tissues from White compared to Black men. Conversely, CD68+ and CD80+ macrophages were increased in Black compared to White men. CD66+ neutrophils and CD206+ macrophages were increased in AR- compared to AR intact metastatic tissues. Immune cells appeared to be primarily void of AR expression in primary cancer tissues by mIHC, but there was a subset CD4+ and CD8+ T-cell clusters that express AR. Immune cell infiltration varies by the presence of AR in both primary and metastatic PCa tissues. We observed differences in AR mRNA and AR-related proteins between Black and White patients which may contribute to the differences observed in immune cell infiltration between the two races Citation Format: Kennedy Rains, Katie M. Farkouh, Taylor N. Godwin, Eric Erak, Aakash Parikh, Michael Considine, Elana Fertig, Jiayun Lu, Charles Waihenya, Valerie Odero-Marah, Karen S. Sfanos, Angelo M. De Marzo, Corinne E. Joshu, Janielle P. Maynard. Androgen receptor-mediated regulation of innate immunity in prostate cancer in Black men [abstract]. In: Proceedings of the 17th AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2024 Sep 21-24; Los Angeles, CA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2024;33(9 Suppl):Abstract nr C122.

Towards a Deep Learning-Powered Herbarium Image Analysis Platform

Global digitization efforts have archived millions of specimen scans worldwide in herbarium collections, which are essential for studying plant evolution and biodiversity. ReColNat hosts, at present, over 10 million images. However, analyzing these datasets poses crucial challenges for botanical research. The application of deep learning in biodiversity analyses, particularly in analyzing herbarium scans, has shown promising results across numerous tasks (Ariouat et al. 2023, Ariouat et al. 2024, Groom et al. 2023, Sahraoui et al. 2023). Within the e-Col+project (ANR-21-ESRE-0053), we are developing multiple deep learning models aimed at identifying plant morphological traits. We have developed pipelines and models for cleaning, analyzing, and transforming herbarium images, including models for: i) detecting non-vegetal elements, such as barcodes, envelopes, labels, etc.; ii) detecting plant organs, including leaves, flowers, fruits, etc.; and iii) segmenting to recognize plant parts for image cleaning. We are also developing models for classification tasks related to various morphological traits. To validate these models, improve their generalization, and make them easily usable by end-users, deploying them within a generic platform is crucial. The generic platform called PlantAI, currently under development by the e-Col+ project, should enable easy deployment during development for testing and allow users to load annotations for new traits in order to train a model and add it to the existing catalog. The platform is based on a microservice architecture, allowing users to upload images, create custom datasets, and access various AI models for image analysis. The platform is composed of four main modules, as illustrated in Fig. 1. The first module is the collaborative workspace manager, which allows users to create projects and image datasets and invite other users to collaborate on a project. The second module is the navigation interface and dashboards. This module integrates a search engine using metadata and AI annotations, a navigation interface between projects, datasets, and specimens, as well as dashboards for analysis across datasets, specimens, and AI models. The third module is the dataset manager, which handles metadata and annotations associated with the specimens. These annotations can be produced either by expert users or by AI models. The fourth module is the AI models management module, so that models can be used to generate AI annotations of specimen. During the development lifecycle of an AI model, users can create datasets and annotate them with AI models. These annotations can be in two possible states: validated by experts and non-validated. Users collaborating on a project can indicate errors in the model predictions and leave comments to explain their evaluations. These corrections made by experts can be used to retrain the models and thus improve their performance. This platform, will be highly beneficial for botanists, enhancing the efficiency and effectiveness of biodiversity analyses from herbarium scans. We aim to provide users with a catalog of AI models through this platform and allow them to import their own datasets with their own annotations regarding traits of their choice. Users will be able to select a model from the AI model catalog and train it using their dataset. Ultimately, the model obtained from this training will be automatically deployed to be available for AI annotation. The annotations produced by this model will be automatically available in the filtering and navigation interface, thus allowing for dynamic and automatic integration of the AI annotations into the navigation interface.

Abstract We112: Volumetric imaging to recapitulate cardiac injury and arrhythmias

Advanced understanding of cardiac structure and function is crucial for uncovering the underlying mechanism of injury and arrhythmias. To explore cues to structural and functional abnormalities in extensively established animal models, we have developed optical imaging and computational methods tailored to investigate intact hearts of zebrafish and rodents at cellular resolution. Our cardiac light-field microscopy allows us to capture instantaneous dynamics such as calcium transients, irregular contraction, and blood flow at 200 volumes per second in zebrafish larvae 1 . To improve the spatial resolution and enable the study of rodent models, we have customized two light-sheet imaging systems. One system is empowered by the retrospective synchronization and machine learning for the study of contractile function and focal myocardial mechanics from end-systole to end-diastole 2 , while the other integrates tissue clearing and axially scanning approaches for the exploration of ventricular trabeculation and myocardial compaction 3 . These methods aim to overcome the trade-off in spatial resolution, imaging speed, field of view across the atria and ventricles with minimal photo-damage and maximal penetration depth, enabling long-term investigation of cardiac development and regenerative processes. To further interpret the large volume of datasets, we have customized a successive subspace learning and virtual reality-based platform for interactive image analysis 4 . Collectively, our multi-scale strategy has opened up new landscapes for exploring the cardiac micro-structure and contractile function, both in health and disease.

A quantitative image analysis platform for assessing trophoblast differentiation

Immunofluorescence microscopy is extensively used in characterization of trophoblast differentiation in vitro. However, such data is primarily used to confirm the presence of protein markers or qualitatively compare levels of protein markers across experimental conditions. Imaging data, when processed and analyzed appropriately can provide quantitative and spatial information, and provide biological insight. Towards this end, here we present MATroph, an open-source MATLAB-based computational tool to process images generated by immunofluorescent microscopy. MATroph automatically executes a series of image processing operations, including the classification of red, blue, and green channels from images, background extraction, morphological operations, and image filtering. From the isolated blue channels corresponding to nuclear staining, this tool generates numerical values for cell number. Additionally, relative levels and spatial location of proteins are obtained by mapping red and green channel pixels to blue pixels by assigning minimum pixel distance between the blue and other color objects. Thus, this tool provides information about intracellular protein accumulation areas. Additionally, this tool can also classify cells as single cells or part of colonies, and extract information on protein levels for each; this is particularly useful for quantitative studies on extravillous trophoblast maturation. We provide a user-guide to analyze the relative levels of markers relevant to human trophoblast stem cell self-renewal and differentiation. Importantly, MATroph is composed of a simple MATLAB algorithm, and its implementation requires minimal expertise in programming.

Aramchol improves hepatic fibrosis in metabolic dysfunction-associated steatohepatitis: Results of multimodality assessment using both conventional and digital pathology.

Antifibrotic trials rely on conventional pathology despite recognized limitations. We compared single-fiber digital image analysis with conventional pathology to quantify the antifibrotic effect of Aramchol, a stearoyl-CoA desaturase 1 inhibitor in development for metabolic dysfunction-associated steatohepatitis. Fifty-one patients with metabolic dysfunction-associated steatohepatitis enrolled in the open-label part of the ARMOR trial received Aramchol 300mg BID and had paired pre-post treatment liver biopsies scored by consensus among 3 hepatopathologists, and separately assessed by a digital image analysis platform (PharmaNest) that generates a continuous phenotypic Fibrosis Composite Severity (Ph-FCS) score. Fibrosis improvement was defined as: ≥1 NASH Clinical Research Network (NASH-CRN) stage reduction; "improved" by ranked pair assessment; reduction in Ph-FCS ("any" for ≥0.3 absolute reduction and "substantial" for ≥25% relative reduction). Fibrosis improved in 31% of patients (NASH-CRN), 51% (ranked pair assessment), 74.5% (any Ph-FCS reduction), and 41% (substantial Ph-FCS reduction). Most patients with stable fibrosis by NASH-CRN or ranked pair assessment had a Ph-FCS reduction (a third with substantial reduction). Fibrosis improvement increased with treatment duration: 25% for <48 weeks versus 39% for ≥48 weeks by NASH-CRN; 43% versus 61% by ranked pair assessment, mean Ph-FCS reduction -0.54 (SD: 1.22) versus -1.72 (SD: 1.02); Ph-FCS reduction (any in 54% vs. 100%, substantial in 21% vs. 65%). The antifibrotic effect of Aramchol was corroborated by reductions in liver stiffness, Pro-C3, and enhanced liver fibrosis. Changes in Ph-FCS were positively correlated with changes in liver stiffness. Continuous fibrosis scores generated in antifibrotic trials by digital image analysis quantify antifibrotic effects with greater sensitivity and a larger dynamic range than conventional pathology.

Prediction of Oncotype DX Recurrence Score Based on Systematic Evaluation of Ki-67 Scores in Hormone Receptor-Positive Early Breast Cancer.

Oncotype DX (ODX) predicts the risk of recurrence and benefits of adding chemotherapy for patients with estrogen receptor positive (ER+)/human epidermal growth factor receptor 2 negative (HER2-) early-stage breast cancer. We aimed to develop a simplified scoring system using readily available clinicopathological parameters to predict a high-risk ODX recurrence score (RS) while minimizing reproducibility issues regarding Ki-67 index evaluation methods. We enrolled 300 patients with ER+/HER2- early breast cancer, for whom ODX RS data were available in the test set. Using the QuPath image analysis platform, we systematically evaluated the average, hotspot, and hottest spot Ki-67 scores in the test set. Logistic regression analyses were conducted to establish a predictive scoring system for high-risk ODX RS. An independent validation set comprising 117 patients over different periods was established. Factors such as age ≤ 50 years, invasive ductal carcinoma tumor type, histologic grade 2 or 3, tumor necrosis, progesterone receptor negativity, and a high Roche-analyzed Ki-67 score (> 20) were associated with high-risk ODX RS. These variables were incorporated into our scoring system. The area under the curve of the scoring system was 0.8057. When applied to both the test and validation sets with a cutoff value of 3, the sensitivity of our scoring system was 92%. We successfully developed a scoring system based on the systematic evaluation of Ki-67 scoring methods. We believe that our user-friendly predictive scoring system for high risk ODX RS could help clinicians in identifying patients who may or may require additional ODX testing.

Quantification of Visceral Fat at the L5 Vertebral Body Level in Patients with Crohn's Disease Using T2-Weighted MRI.

The umbilical or L3 vertebral body level is often used for body fat quantification using computed tomography. To explore the feasibility of using clinically acquired pelvic magnetic resonance imaging (MRI) for visceral fat measurement, we examined the correlation of visceral fat parameters at the umbilical and L5 vertebral body levels. We retrospectively analyzed T2-weighted half-Fourier acquisition single-shot turbo spin echo (HASTE) MR axial images from Crohn's disease patients who underwent MRI enterography of the abdomen and pelvis over a three-year period. We determined the area/volume of subcutaneous and visceral fat from the umbilical and L5 levels and calculated the visceral fat ratio (VFR = visceral fat/subcutaneous fat) and visceral fat index (VFI = visceral fat/total fat). Statistical analyses involved correlation analysis between both levels, inter-rater analysis between two investigators, and inter-platform analysis between two image-analysis platforms. Correlational analysis of 32 patients yielded significant associations for VFI (r = 0.85; p < 0.0001) and VFR (r = 0.74; p < 0.0001). Intraclass coefficients for VFI and VFR were 0.846 and 0.875 (good agreement) between investigators and 0.831 and 0.728 (good and moderate agreement) between platforms. Our study suggests that the L5 level on clinically acquired pelvic MRIs may serve as a reference point for visceral fat quantification.

Development of a Secure Web-Based Medical Imaging Analysis Platform: The AWESOMME Project.

Precision medicine research benefits from machine learning in the creation of robust models adapted to the processing of patient data. This applies both to pathology identification in images, i.e., annotation or segmentation, and to computer-aided diagnostic for classification or prediction. It comes with the strong need to exploit and visualize large volumes of images and associated medical data. The work carried out in this paper follows on from a main case study piloted in a cancer center. It proposes an analysis pipeline for patients with osteosarcoma through segmentation, feature extraction and application of a deep learning model to predict response to treatment. The main aim of the AWESOMME project is to leverage this work and implement the pipeline on an easy-to-access, secure web platform. The proposed WEB application is based on a three-component architecture: a data server, a heavy computation and authentication server and a medical imaging web-framework with a user interface. These existing components have been enhanced to meet the needs of security and traceability for the continuous production of expert data. It innovates by covering all steps of medical imaging processing (visualization and segmentation, feature extraction and aided diagnostic) and enables the test and use of machine learning models. The infrastructure is operational, deployed in internal production and is currently being installed in the hospital environment. The extension of the case study and user feedback enabled us to fine-tune functionalities and proved that AWESOMME is a modular solution capable to analyze medical data and share research algorithms with in-house clinicians.

A Dorsal Skinfold Window Chamber Tumor Mouse Model for Combined Intravital Microscopy and Magnetic Resonance Imaging in Translational Cancer Research.

Preclinical intravital imaging such as microscopy and optical coherence tomography have proven to be valuable tools in cancer research for visualizing the tumor microenvironment and its response to therapy. These imaging modalities have micron-scale resolution but have limited use in the clinic due to their shallow penetration depth into tissue. More clinically applicable imaging modalities such as CT, MRI, and PET have much greater penetration depth but have comparatively lower spatial resolution (mm scale). To translate preclinical intravital imaging findings into the clinic, new methods must be developed to bridge this micro-to-macro resolution gap. Here we describe a dorsal skinfold window chamber tumor mouse model designed to enable preclinical intravital and clinically applicable (CT and MR) imaging in the same animal, and the image analysis platform that links these two disparate visualization methods. Importantly, the described window chamber approach enables the different imaging modalities to be co-registered in 3D using fiducial markers on the window chamber for direct spatial concordance. This model can be used for validation of existing clinical imaging methods, as well as for the development of new ones through direct correlation with "ground truth" high-resolution intravital findings. Finally, the tumor response to various treatments-chemotherapy, radiotherapy, photodynamic therapy-can be monitored longitudinally with this methodology using preclinical and clinically applicable imaging modalities. The dorsal skinfold window chamber tumor mouse model and imaging platforms described here can thus be used in a variety of cancer research studies, for example, in translating preclinical intravital microscopy findings to more clinically applicable imaging modalities such as CT or MRI.

Abstract 87: Integrated pipeline for immuno-oncology drug testing: from patient-derived organoid assays with reconstituted TME to fresh ex vivo patient tissue cultures

Abstract Introduction: Successful pre-clinical evaluation of immuno-oncology drugs requires complex models to accurately represent all components of the human tumor microenvironment (TME). Patient-derived organoids (PDO) are known to reflect genetic and morphological profiles of the original patient’s tumors. In our immuno-oncology in vitro assays, PDOs can be combined with immune cells and fibroblasts to re-constitute the TME and study drug efficacy and mechanism of action. Alternatively, we have developed assays with fresh tumor material containing native TME in the 3D Ex Vivo Patient Tissue (EVPT) platform. Both types of assays have advantages and limitations. Here, we will present a case study displaying both capabilities and benchmarking the technologies against each other. Methods: Patient tumor tissues were obtained from tissue providers and processed within 24 hours to preserve the TME. Freshly isolated ex vivo tumor clusters were embedded in a protein-rich hydrogel and exposed to drug panels in a 384-well format for 5-7 days. Phenotypic effects of therapies on morphological features were measured using our proprietary automated high content imaging (HCI) analysis platform. IHC, FACS, or sequencing, were used for in-depth sample characterization. In parallel, PDOs were generated, biobanked and characterized from same ex vivo samples. Allogenic or autologous PBMCs were isolated from blood. PDOs and PBMCs were then co-cultured together in presence of immuno-oncology drugs, and drug effects were analyzed by the same automated HCI analysis platform. Results: Samples from ovarian, head and neck, breast, prostate, lung cancer, and melanoma were used in the pipeline. Fresh ex vivo culture of a melanoma sample presented here has shown mild responses (tumoroid size reduction) to immunomodulatory treatments (SEA and Pembrolizumab) and strong responses to targeted therapy treatment (Dabrafenib and Trametinib). Genetic characterization of the sample identified heterozygous BRAF V600E mutation and IHC analysis showed a large proportion of immune cells (CD3, CD4, CD8, CD80 and CD68 positive cells) present, confirming the observed ex vivo drug responses. Part of the melanoma sample was allocated for PDO generation and biobanking. Established PDOs were used in co-culture with allogeneic and autologous PBMCs and treatment responses to SEA stimulation and CD3/CD28 beads as well as increased PBMC infiltration rates were observed. Conclusion: Choice of the pre-clinical model for immuno-oncology studies relies on multiple factors such as biological relevance, replicability, TME representation, and scalability. Here we present a spectrum of assays ranging from complex ex vivo systems (EVPT) for ‘close to the clinic’ scenarios to more scalable PDOs co-cultured with allogeneic and/or autologous PBMCs for high-throughput mechanistic drug screenings. Citation Format: Ezgi Kaya Aksoy, Sergei Chavez Abiega, Talita Stessuk, Saskia De Man, Marrit Putker, Sander Basten, Niels Meesters, Emma Spanjaard, Leo S. Price, Gera Goverse, Nataliia Beztsinna. Integrated pipeline for immuno-oncology drug testing: from patient-derived organoid assays with reconstituted TME to fresh ex vivo patient tissue cultures [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 87.

Abstract 6181: Digital pathology workflow: Building a bridge for translational biomarker development in humanized mice

Abstract The introduction of immune checkpoint inhibitors (ICIs) directed against programmed death protein 1 (PD-1), as well as cytotoxic T lymphocyte antigen 4 (CTLA-4), has revolutionized the treatment of non-small cell lung cancer (NSCLC) in the last decade. These therapies have shown significant success in a subset of NSCLC patients, leading to durable response and improved overall survival. However, clear predictive biomarkers for response are still yet to be defined. Tumor-infiltrating lymphocytes (TILs) as one important part of the tumor microenvironment (TME) have proven to be predictive biomarkers in other tumor types such as melanoma, whereas not fully understood in NSCLC. The assessment of TILs typically involves techniques such as flow cytometry (FC) or immunohistochemistry (IHC). While the former has the advantage of being quantitative and relatively easy to standardize it lacks translational value as most clinical samples will be examined using IHC. The current study aims to compare both methods at terminal end point and for treatment outcome, using a panel of 14 NSCLC PDX models, implanted in humanized NSG mice and treated with anti-CTLA4, anti-PD1 and the combination thereof. Based on FC analyses of the untreated tumor tissue the PDX models were stratified into seven hot (&amp;gt;5% TILs) and seven cold (&amp;lt;5% TILs) tumors. The IHC samples were analyzed by digital pathology for TIL infiltration by making use of automated segmentation of the three main tissue classes: tumor, stroma and necrosis. Small number of IHC slides were manually annotated to form a training set, which were then used to train a deep learning model using Visiopharm, a commercial AI platform for image analysis. By quantification of TILs in a spatial context of the tumor tissue, across all models, tumor was the most prominent tissue class. The tissue class distribution as well as the TIL rate (FC or IHC measure) is a model immanent feature. Under treatment, the ratio was stable in case of the cold tumors. In the hot tumors the stroma compartment increased while tumor tissue decreased. Treatment sensitive models displayed an increase of TILs confirmed by FC and IHC. The latter revealed that the TILs specifically increased in the tumor tissue whereas decreased or was stable in the stroma. This observation was significantly more pronounced in cold tumors compared to hot tumors (Oneway ANOVA, p &amp;lt; 0.001). Furthermore, a positive correlation between FC and IHC results on CD45 positive cells in the tumors (p&amp;lt; 0.002) could be confirmed. In this study, we proofed the feasibility of software-based image analysis as translational relevant read-out in preclinical animal experiments. The possibility to reproducibly quantify tissue classes, immune cell markers and including spatial information paves the way towards discovery of novel features predicting response in translational immuno-oncology research. Citation Format: Matti Hakkarainen, Timmo Bragge, Jussi Rytkönen, Eva Oswald, Julia B. Schueler. Digital pathology workflow: Building a bridge for translational biomarker development in humanized mice [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 6181.

Digital pathology with artificial intelligence analysis provides insight to the efficacy of anti-fibrotic compounds in human 3D MASH model

Metabolic dysfunction-associated steatohepatitis (MASH) is a severe liver disease characterized by lipid accumulation, inflammation and fibrosis. The development of MASH therapies has been hindered by the lack of human translational models and limitations of analysis techniques for fibrosis. The MASH three-dimensional (3D) InSight™ human liver microtissue (hLiMT) model recapitulates pathophysiological features of the disease. We established an algorithm for automated phenotypic quantification of fibrosis of Sirius Red stained histology sections of MASH hLiMTs model using a digital pathology quantitative single-fiber artificial intelligence (AI) FibroNest™ image analysis platform. The FibroNest™ algorithm for MASH hLiMTs was validated using anti-fibrotic reference compounds with different therapeutic modalities-ALK5i and anti-TGF-β antibody. The phenotypic quantification of fibrosis demonstrated that both reference compounds decreased the deposition of fibrillated collagens in alignment with effects on the secretion of pro-collagen type I/III, tissue inhibitor of metalloproteinase-1 and matrix metalloproteinase-3 and pro-fibrotic gene expression. In contrast, clinical compounds, Firsocostat and Selonsertib, alone and in combination showed strong anti-fibrotic effects on the deposition of collagen fibers, however less pronounced on the secretion of pro-fibrotic biomarkers. In summary, the phenotypic quantification of fibrosis of MASH hLiMTs combined with secretion of pro-fibrotic biomarkers and transcriptomics represents a promising drug discovery tool for assessing anti-fibrotic compounds.

NanoLocz: Image Analysis Platform for AFM, High-Speed AFM, and Localization AFM.

Atomic Force Microscopy (AFM), High-Speed AFM (HS-AFM) simulation AFM, and Localization AFM (LAFM) enable the study of molecules and surfaces with increasingly higher spatiotemporal resolution. However, effective and rapid analysis of the images and movies produced by these techniques can be challenging, often requiring the use of multiple image processing software applications and scripts. Here, NanoLocz, an open-source solution that offers advanced analysis capabilities for the AFM community, is presented. Integration and continued development of AFM analysis tools is essential to improve access to data, increase throughput, and open new analysis opportunities. NanoLocz efficiently leverages the rich data AFM has to offer by incorporating and combining existing and newly developed analysis methods for AFM, HS-AFM, simulation AFM, and LAFM seamlessly. It facilitates and streamlines AFM analysis workflows from import of raw data, through to various analysis workflows. Here, the study demonstrates the capabilities of NanoLocz and the new methods it enables including single-molecule LAFM, time-resolved LAFM, and simulation LAFM.

3D modeling of normal skin and cutaneous squamous cell carcinoma. A comparative study in 2D cultures, spheroids, and 3D bioprinted systems.

The current cancer research and drug testing are primarily based on 2D cell cultures and animal models. However, these methods have limitations and yield distinct drug response patterns. This study addressed this gap by developing an innovativein vitrohuman three-dimensional (3D) normal skin model and a multicellular model of human cutaneous squamous cell carcinoma (cSCC) using 3D bioprinting technology. Comparative analyzes were performed between bioprinted 3D-cSCC model, consisting of HaCaT keratinocytes, primary normal human dermal fibroblasts and A431 cancer cells (tricellular), bioprinted 3D-A431 model composed of A431 cancer cells only (monocellular), A431 cancer cell spheroids, and conventional 2D models. The models were structurally characterized by light microscopy, immunofluorescence (LIVE/DEAD assay, confocal microscopy) and immunohistochemistry (hematoxylin/eosin, p63, vimentin, Ki67, epidermal growth factor receptor stainings). The spatial arrangement of the 3D models was analyzed using the ARIVIS scientific image analysis platform. All models were also functionally assessed by cetuximab (CTX) response testing with the MTS assay. 3D-cSCC models were maintained for up to 16 weeks. Morphological and histological examinations confirmed the presence of skin-like layers in the bioprinted 3D models of normal skin, and the intricate and diverse features of the bioprinted skin cancer model, replicating the criticalin vivocharacteristics. In both mono- and tricellular bioprinted tumor constructs, there was a gradual formation and continuous growth of spheroid-like clusters of cancer cells, significantly influencing the morphology of the models. Cancer cells in the 3D bioprinted constructs showed reduced sensitivity to CTX compared to spheroids and 2D cultures. This study underscores the potential of 3D multicellular models in elucidating drug responses and gaining a better understanding the intricate interplay of cellular components within the tumor microenvironment. Developing the multicellular 3D tumor model paves the way for new research critical to advancing fundamental cancer research and future clinical applications, particularly drug response testing.

Cancer-associated Fibroblast-specific Expression of the Matricellular Protein CCN1 Coordinates Neovascularization and Stroma Deposition in Melanoma Metastasis.

In human patients, the expression of proangiogenic matricellular protein CCN1 in CAFs correlates positively with expression of stroma and angiogenic markers and progressive disease/resistance to checkpoint inhibitor therapy. In an animal model, loss of CCN1 from CAFs impaired metastasis of melanoma cells, neovascularization, and collagen deposition, emphasizing that CAFs coordinate cellular behavior in a tumor microenvironment and that CCN1 may be a novel target.

A smart fruit size measuring method and system in natural environment

The measurement of fruit size (length, width, and area) is one of the key components in the assessment of fruit ripeness for consumption and fruit quality. Due to issues with fruit shape and ambient occlusion, noncontact measurement of fruits in their natural growth state in natural environment is more difficult. This study proposes a fast and low-cost noncontact measurement solution combining deep learning, image analysis and robotic platform. First, based on the YOLOv5 object detection algorithm and converged dataset, a 95.6% detection accuracy and 0.05 s detection speed was achieved. Second, a Cycle GAN model for fruit occlusion recovery was established, and the occlusion of the fruit under different natural conditions was adaptively repaired, with a 5.48% average relative error under different occlusion conditions. Third, through image morphological operations, the size measurements were achieved, with an average relative error of 8.43% for individual fruits and the 10.12% overall error under the complex natural environment. This method and platform provide a systematic, fast (0.2s) and low-cost solution for the noncontact measurement of fruit size.

Analysis of organoid and immune cell co-cultures by machine learning-empowered image cytometry.

Organoids are three-dimensional (3D) structures that can be derived from stem cells or adult tissue progenitor cells and exhibit an extraordinary ability to autonomously organize and resemble the cellular composition and architectural integrity of specific tissue segments. This feature makes them a useful tool for analyzing therapeutical relevant aspects, including organ development, wound healing, immune disorders and drug discovery. Most organoid models do not contain cells that mimic the neighboring tissue’s microenvironment, which could potentially hinder deeper mechanistic studies. However, to use organoid models in mechanistic studies, which would enable us to better understand pathophysiological processes, it is necessary to emulate the in situ microenvironment. This can be accomplished by incorporating selected cells of interest from neighboring tissues into the organoid culture. Nevertheless, the detection and quantification of organoids in such co-cultures remains a major technical challenge. These imaging analysis approaches would require an accurate separation of organoids from the other cell types in the co-culture. To efficiently detect and analyze 3D organoids in co-cultures, we developed a high-throughput imaging analysis platform. This method integrates automated imaging techniques and advanced image processing tools such as grayscale conversion, contrast enhancement, membrane detection and structure separation. Based on machine learning algorithms, we were able to identify and classify 3D organoids within dense co-cultures of immune cells. This procedure allows a high-throughput analysis of organoid-associated parameters such as quantity, size, and shape. Therefore, the technology has significant potential to advance contextualized research using organoid co-cultures and their potential applications in translational medicine.