AI-based Analysis Research Articles

PATH-39. AI-BASED IDENTIFICATION OF GLIOMA IDH MUTATIONAL STATUS FROM H&E-STAINED WHOLE SLIDE IMAGES

Abstract INTRODUCTION Isocitrate-dehydrogenase (IDH) mutational status is diagnostically critical in adult gliomas, with prognostic and therapeutic implications. IDH status cannot be determined by sole histologic assessment and requires molecular testing. We hypothesize that AI-based analysis can accurately predict IDH mutational status in glioma, by retrieving sub-visual cues in H&E-stained digitized tissue sections. MATERIALS AND METHODS A retrospective discovery cohort of 1,534 cases (756/778 IDH-wildtype/IDH-mutant, including all available grades) from the TCGA-LGG/TCGA-GBM collections was used for model development with 10-fold cross-validation (CV). A replication cohort of 114 cases (82/32 IDH-wildtype/IDH-mutant) from the University of Pennsylvania Health System (UPHS) was used for independent hold-out evaluation. Each 20X magnification whole slide image (WSI) underwent comprehensive curation for elimination of artifactual content, followed by partitioning into 256x256 patches, and feature extraction using pre-trained i) ImageNet weights (baseline), and ii) self-supervised vision transformer (ViT). A weakly-supervised attention-based multiple-instance-learning framework distinguished cases as IDH-wildtype or IDH-mutant, while generating visually interpretable attention heatmaps. RESULTS Evaluation on discovery and replication cohorts with the ViT yield accuracy of 90.84% (AUCDiscovery=97.16%) and 93.94% (AUCReplication=97.36%), respectively, with high sensitivity (i.e., high confidence in correctly predicting IDH-wildtype glioma). Comparison of the ViT with the baseline model on the replication cohort indicates the ViT superiority by improvements of 8% and 9.5% in accuracy and AUC, respectively. Histologic assessment of heatmaps indicates IDH-wildtype tumors exhibiting distinct regions of significant pleomorphism and microvascular proliferation, while IDH-mutant tumors exhibit dense nodular cell concentrations, microcystic architecture, and gemistocytic cells. CONCLUSION Our accurate H&E-based computational determination of glioma IDH status, with interpretations aligned with human-identifiable features, can contribute to obviating the need for molecular analyses and enable expedited diagnosis even in community healthcare settings. This study robustly identifies morpho-IDH correlates and could enable expedited patient prognostic stratification.

AI-Based Analysis of Ziehl–Neelsen-Stained Sputum Smears for Mycobacterium tuberculosis as a Screening Method for Active Tuberculosis

Tuberculosis is the primary cause of death due to infection in the world. Identification of Mycobacterium tuberculosis in sputum is a diagnostic test, which can be used in screening programs—especially in countries with a high incidence of tuberculosis—to identify and treat those persons with the highest risk of disseminating the infection. We previously developed an algorithm which is able to automatically detect mycobacteria on tissue; in particular, our algorithm identified acid-fast bacilli on tissue with 100% specificity, 95.65% sensitivity, and 98.33% accuracy. We tested this algorithm on 1059 Ziehl–Neelsen-stained sputum smears to evaluate its results as a possible tool for screening. The results were displayed as a heat map of 32 × 32 pixel patches. Analysis of the positive patches revealed a good specificity (86.84%) and 100% sensitivity for patches with a level of confidence over 90; furthermore, the accuracy remained over 95% for all levels of confidence over 80, except the class (95–100]. The modest specificity is caused by the peculiarities of smears (uneven thickness, dust contamination, lack of coverslip). We will train the algorithm on sputum smears to increase the specificity to over 95%. However, as our algorithm showed no false negatives, it is suitable for screening.

A novel management challenge in age-related macular degeneration: Artificial intelligence and expert prediction of geographic atrophy

PurposeThe progression of geographic atrophy (GA) secondary to age-related macular degeneration (AMD) is highly variable among individuals. Prediction of the progression is critical to identify patients who will benefit most from the first treatments currently approved. The aim of this study was the investigation of the value and difference in predictive power between ophthalmologists and artificial intelligence (AI) to reliably assess individual speed of GA progression. DesignProspective, expert and AI comparison study. ParticipantsEyes with natural progression of GA from a prospective study (NCT02503332). MethodsOphthalmologists predicted yearly growth speed of GA, as well as selecting the potentially faster growing lesions from two eyes based on fundus autofluorescence (FAF), near-infrared reflectance (NIR), and optical coherence tomography (OCT). A deep learning algorithm predicted progression solely on the baseline OCT (Spectralis, Heidelberg Engineering, Germany). Main Outcome MeasuresAccuracy, weighted kappa (κ), and concordance index (c-index) between the prediction made by ophthalmology specialists, ophthalmology residents and the AI. ResultsA total of 134 eyes of 134 patients from a phase II clinical trial were included, among those 53 were from the sham arm and 81 from untreated fellow eyes. 2880 gradings were performed by four ophthalmologists. Human experts reached an accuracy of 0.37, 0.43, 0.41 and a κ of 0.06, 0.16, 0.18 on FAF, NIR+OCT and FAF+NIR+OCT, respectively. On a pairwise comparison task, human experts achieved a c-index of 0.62, 0.59 and 0.60. Automated AI-based analysis reached an accuracy of 0.48 and a κ of 0.23 on the first task, and a c-index of 0.69 on the second task solely utilizing OCT imaging. ConclusionsPrediction of individual progression will become an important task for patient counseling, most importantly with treatments becoming available. Human gradings improved with the availability of OCT. However, automated AI performed better than ophthalmologists in several comparisons. AI-supported decisions improve clinical precision, access to timely care for the community, and socioeconomic feasibility in the management of the leading cause for irreversible vision loss.

Uso De Inteligência Artificial Na Análise De Conteúdo: Uma Nova Capacidade Pública?

Background: This article addresses an applied study on the use of Artificial Intelligence for content analysis. The problem at hand involved identifying the reasons for the low execution of federal public funds transferred to states and municipalities in Brazil, aimed at extending students' school hours under the Full-Time School Program. As of June 2024, there was R$ 3.8 billion in the accounts of these entities, representing approximately 20% of the funds utilized, while 37.5% of the execution period had already elapsed. The need for a specific assessment targeting 26 states, one federal district, and 5,568 municipalities, without available personnel or time for analysis, led to the methodological approach chosen. Materials and Methods: A questionnaire was administered to 5,595 agents, containing objective questions focused on elementary aspects that would provide relevant information for the results and serve as a consistency check for the analysis. To capture unforeseen situations that could hinder the study, open-ended questions were also included, whose analysis would require significant dedication and substantial time. Therefore, prior to sending the questionnaire, the decision was made to use Artificial Intelligence, specifically the paid version of ChatGPT, to conduct the analyses, which required appropriate structuring of the questionnaire. Results: The survey generated nearly three thousand responses, which were consolidated into a structured document of 2,947 pages. This document was built directly from the response spreadsheet and transformed into a continuous document using the mail merge feature in Microsoft Word, then converted to PDF for analysis. The objective responses reflected optimism from the managers regarding the utilization and contribution of the funds; however, the execution forecast indicated that approximately six more months would be needed to significantly increase fund usage. Hypotheses were confirmed that procurement processes, challenges in school meal provision, and legal difficulties in hiring personnel were major limiting factors. The analysis of the openended questions also revealed specific problems related to school transportation, introducing a new element that was not previously hypothesized. The adopted methodology also enabled regional analyses, which enhanced the Brazilian Federal Government’s capacity to provide targeted technical support. Conclusion: The combination of conventional methods, such as the objective questions in the questionnaire, with open-ended questions designed to facilitate AI-based analysis, provided a fast-paced and balanced evaluation. This reinforces the importance of using AI tools to enhance public institutions' ability to listen to multiple managers, while maintaining a balance with other methods, in order to generate greater reliability alongside innovation.

EASY: Energy-Efficient Analysis and Control Processes in the Dynamic Edge-Cloud Continuum for Industrial Manufacturing

AbstractAccording to the guiding principles of Industry 4.0, edge computing enables the data-sovereign and near-real-time processing of data directly at the point of origin. Using these edge devices in manufacturing organization will drive the use of industrial analysis, control, and Artificial Intelligence (AI) applications close to production. The goal of the EASY project is to make the added value of edge computing available by providing an easily usable Edge-Cloud Continuum with a runtime environment and services for the execution of AI-based Analysis and Control processes. Within this continuum, a dynamic, distributed, and optimized execution of services is automated across the entire spectrum from centralized cloud to decentralized edge instances to increase productivity and resource efficiency.

The modified physics-informed neural network (PINN) method for the thermoelastic wave propagation analysis based on the Moore-Gibson-Thompson theory in porous materials

This paper presents novel contributions to both theory and solution methodology in AI-based analysis of solid mechanics. The physics-informed neural network (PINN) method is developed for thermoelastic wave propagation and Moore-Gibson-Thompson (MGT) coupled thermoelasticity analysis of porous media, a first in the field. The coupled thermoelasticity governing equations, based on the MGT heat conduction model, are derived for a porous half-space, with the thermal relaxation coefficient and strain relaxation factor being considered. Mechanical and thermal shock loading boundary conditions are imposed. The behavior of a magnesium-made porous body is analyzed using the PINN method, with highly accurate results being achieved for the system of coupled PDEs. An adaptive hyperparameter tuning approach, integrating a generalized subset design (GSD) and Bayesian optimization algorithm, is used to automatically select the optimal structure based on the L2 relative error. This hybrid methodology eliminates manual adjustment concerns. The proposed method is verified through a thorough comparison with the Lord-Shulman theory of coupled thermoelasticity. The strength of the methodology lies in its ability to operate without domain data, with only boundary and initial points being required. Four example sets are examined to demonstrate the capabilities of the modified PINN, and high-quality predictions of dimensionless fields’ variables over an extended time interval are obtained, confirming its extrapolation abilities.

Time budgets and weight shifting as indicators of pain in hospitalized horses.

Pain assessment in horses presents a significant challenge due to their nonverbal nature and their tendency to conceal signs of discomfort in the presence of potential threats, including humans. Therefore, this study aimed to identify pain-associated behaviors amenable to automated AI-based detection in video recordings. Additionally, it sought to determine correlations between pain intensity and behavioral and postural parameters by analyzing factors such as time budgets, weight shifting, and unstable resting. The ultimate goal is to facilitate the development of AI-based quantitative tools for pain assessment in horses. A cohort of 20 horses (mean age 15 ± 8) admitted to a university equine hospital underwent 24-h video recording. Behaviors were manually scored and retrospectively analyzed using Loopy® software. Three pain groups were established based on the Pain Score Vetmeduni Vienna : pain-free (P0), mild to moderate pain (P1), and severe pain (P2). Weight shifting emerged as a reliable indicator for discriminating between painful and pain-free horses, with significant differences observed between pain groups (p < 0.001) and before and after administration of analgesia. Additionally, severely painful horses (P2 group) exhibited lower frequencies of feeding and resting standing per hour compared to pain-free horses, while displaying a higher frequency of unstable resting per hour. The significant differences observed in these parameters between pain groups offer promising prospects for AI-based analysis and automated pain assessment in equine medicine. Further investigation is imperative to establish precise thresholds. Leveraging such technology has the potential to enable more effective pain detection and management in horses, ultimately enhancing welfare and informing clinical decision-making in equine medicine.

AI-Based multimodal Multi-tasks analysis reveals tumor molecular heterogeneity, predicts preoperative lymph node metastasis and prognosis in papillary thyroid carcinoma: A retrospective study.

Papillary thyroid carcinoma (PTC) is the predominant form of thyroid cancer globally, especially when lymph node metastasis (LNM) occurs. Molecular heterogeneity, driven by genetic alterations and tumor microenvironment components, contributes to the complexity of PTC. Understanding these complexities is essential for precise risk stratification and therapeutic decisions. This study involved a comprehensive analysis of 521 patients with PTC from our hospital and 499 patients from The Cancer Genome Atlas (TCGA). The real-world cohort 1 comprised 256 patients with stage I-III PTC. Tissues from 252 patients were analyzed by DNA-based next-generation sequencing, and tissues from four patients were analyzed by single-cell RNA sequencing (scRNA-seq). Additionally, 586 PTC pathological sections were collected from TCGA, and 275 PTC pathological sections were collected from the real-world cohort 2. A deep learning multimodal model was developed using matched histopathology images, genomic, transcriptomic, and immune cell data to predict LNM and disease-free survival (DFS). This study included a total of 1,011 PTC patients, comprising 256 patients from cohort 1, 275 patients from cohort 2, and 499 patients from TCGA. In cohort 1, we categorized PTC into four molecular subtypes based on BRAF, RAS, RET, and other mutations. BRAF mutations were significantly associated with LNM and impacted DFS. ScRNA-seq identified distinct T cell subtypes and reduced B cell diversity in BRAF-mutated PTC with LNM. The study also explored cancer-associated fibroblasts and macrophages, highlighting their associations with LNM. The deep learning model was trained using 405 pathology slides and RNA sequences from 328 PTC patients and validated with 181 slides and RNA sequences from 140 PTC patients in the TCGA cohort. It achieved high accuracy, with an AUC of 0.86 in the training cohort, 0.84 in the validation cohort, and 0.83 in the real-world cohort 2. High-risk patients in the training cohort had significantly lower DFS rates (P<0.001). Model AUCs were 0.91 at 1 year, 0.93 at 3 years, and 0.87 at 5 years. In the validation cohort, high-risk patients also had lower DFS (P<0.001); the AUCs were 0.89, 0.87, and 0.80 at 1, 3, and 5 years. We utilized the GradCAM algorithm to generate heatmaps from pathology-based deep learning models, which visually highlighted high-risk tumor areas in PTC patients. This enhanced clinicians' understanding of the model's predictions and improved diagnostic accuracy, especially in cases with lymph node metastasis. The AI-based analysis uncovered vital insights into PTC molecular heterogeneity, emphasizing BRAF mutations' impact. The integrated deep learning model shows promise in predicting metastasis, offering valuable contributions to improved diagnostic and therapeutic strategies.

Electrocardiogram Signal Analysis With a Machine Learning Model Predicts the Presence of Pulmonary Embolism With Accuracy Dependent on Embolism Burden

ObjectiveTo develop an artificial intelligence deep neural network (AI-DNN) algorithm to analyze 12-lead electrocardiogram (ECG) for detection of acute pulmonary embolism (PE) and PE categories. Patients and MethodsA cohort of patients seen between January 1, 1999, and December 31, 2020, from across the Mayo Clinic Enterprise with computed tomography pulmonary angiogram (CTPA) and ECG performed ±6 hours was identified. Natural language processing algorithms were applied to radiology reports to determine the diagnosis of acute PE, acute right ventricular strain pulmonary embolism (RVSPE), saddle pulmonary embolism (SADPE), or no PE. Diagnostic performance parameters of the AI-DNN reported were area under the receiver operating characteristics curve (AUROC), sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). ResultsA cohort of patients with CTPA report and ECG consisted of 79,894 patients including 7423 (9.3%) with acute PE, among whom 1138 patients had RVSPE or SADPE. Artificial intelligence deep neural network predicted acute PE with a modest accuracy of AUROC of 0.69 (95% CI, 0.68-0.71), sensitivity of 63.5%, specificity of 64.7%, PPV of 15.6%, and NPV of 94.5%. The AI-DNN prediction using the same algorithm for RVSPE or SADPE was higher (AUROC, 0.84; 95% CI, 0.81-0.86) with a sensitivity of 80.8%, specificity of 64.7.8%, PPV of 3.5%, and NPV of 99.5%. ConclusionAn AI-based analysis of 12-lead ECG shows modest detection power for acute PE in patients who underwent CTPA, with higher accuracy for high-risk PE. Moreover, with the high NPV, it has the clinical potential to exclude high-risk PE quickly and correctly.

New Tools for Data Harmonization and Their Potential Applications in Organ Transplantation.

In organ transplantation, accurate analysis of clinical outcomes requires large, high-quality data sets. Not only are outcomes influenced by a multitude of factors such as donor, recipient, and transplant characteristics and posttransplant events but they may also change over time. Although large data sets already exist and are continually expanding in transplant registries and health institutions, these data are rarely combined for analysis because of a lack of harmonization. Promoted by the digitalization of the healthcare sector, effective data harmonization tools became available, with potential applications also for organ transplantation. We discuss herein the present problems in the harmonization of organ transplant data and offer solutions to enhance its accuracy through the use of emerging new tools. To overcome the problem of inadequate representation of transplantation-specific terms, ontologies and common data models particular to this field could be created and supported by a consortium of related stakeholders to ensure their broad acceptance. Adopting clear data-sharing policies can diminish administrative barriers that impede collaboration between organizations. Secure multiparty computation frameworks and the artificial intelligence (AI) approach federated learning can facilitate decentralized and harmonized analysis of data sets, without sharing sensitive data and compromising patient privacy. A common image data model built upon a standardized format would be beneficial to AI-based analysis of pathology images. Implementation of these promising new tools and measures, ideally with the involvement and support of transplant societies, is expected to produce improved integration and harmonization of transplant data and greater accuracy in clinical decision-making, enabling improved patient outcomes.

Abstract GS03-04: Novel Mechanisms of CDH1 Inactivation in Breast Invasive Lobular Carcinoma Unveiled by the Integration of Artificial Intelligence and Genomics

Abstract Background: Invasive lobular carcinoma (ILC) of the breast is the second most common histologic subtype of breast cancer (BC), following invasive ductal carcinoma of no special type (IDC-NST). The hallmark histologic feature of ILC is cellular discohesiveness, the result of bi-allelic inactivation of CDH1, and represents an important genotypic-phenotypic correlation in BC. Although most ILCs harbor CDH1 loss-of function mutations associated to loss-of-heterozygosity (LOH) of the wild type-allele, a subset of ILCs lack these alterations despite displaying a typical lobular phenotype. Here, we sought to identify alternative molecular mechanisms converging on CDH1 inactivation by employing an integrative artificial intelligence (AI) and genomics approach. Materials and Methods: A genomics-driven AI-based algorithm using hematoxylin and eosin (H&amp;E) whole slide images (WSIs) as input, previously developed to detect bi-allelic CDH1 mutations (inactivating mutation associated to LOH) in BC was employed. WSIs of 1,057 BCs including ILCs (n=187) and non-lobular BCs (n=870) previously subjected to FDA-cleared tumor/normal targeted sequencing were subjected to analysis with the AI-based algorithm. Cases predicted to harbor CDH1 bi-allelic mutations by the AI-model but lacking CDH1 bi-allelic mutations by targeted sequencing were assessed through targeted sequencing data re-analysis, CDH1 gene promoter methylation evaluation and/or whole genome sequencing analysis. Results: AI-based analysis WSIs corresponding to 1,057 BCs resulted in the identification of 34 cases found to lack CDH1 bi-allelic mutations by targeted sequencing but predicted to harbor these genetic alterations by the AI-based model. CDH1 gene promoter methylation assessment revealed CDH1 promoter methylation in 18 cases. Targeted sequencing data reanalysis revealed other genetic mechanisms of CDH1 inactivation including CDH1 homozygous deletions (n=3), intragenic deletion with LOH (n=1), and likely pathogenic non-coding CDH1 alterations associated with LOH (n=2). WGS analysis of an ILC revealed a novel deleterious CDH1 fusion stemming from translocation t(13;16), resulting in loss of the 5’UTR, transcription start site and exons 1 and 2 of CDH1, associated with complete loss of E-cadherin protein expression. Taken together, we identified alternative/novel mechanisms of bi-allelic CDH1inactivation in 74% (25/34) cases analyzed. Conclusions: By applying an AI-based algorithm trained to detect a genetic alteration (i.e., CDH1 bi-allelic mutations), we were able to identify alternative epigenetic and genetic molecular mechanisms of CDH1 inactivation in ILCs, including novel non-coding CDH1 genetic alterations and a new inactivating CDH1 fusion gene. These findings indicate that molecular mechanisms affecting a single gene or process converging on the same phenotype can be unveiled by the integration of AI and genomics, highlighting the robustness of this approach for the discovery of novel biology. Citation Format: Fresia Pareja, Higinio Dopeso, Yikan Wang, Andrea Gazzo, David Brown, Pier Selenica, Jan Bernhard, Fatemeh Derakhshan, Edaise M. da Silva, Lorraine Colon-Cartagena, Thais Basili, Antonio Marra, Jillian Sue, Qiqi Ye, Arnaud Da Cruz Paula, Selma Yeni, Xin Pei, Hunter Green, Kaitlyn Gill, Yingjie Zhu, Matthew Lee, Ran Godrich, Adam Casson, Britta Weigelt, Nadeem Riaz, Hanna Y Wen, Edi Brogi, Matthew Hanna, Diana Mandelker, Jeremy Kunz, Brandon Rothrock, Sarat Chandarlapaty, Christopher Kanan, Gerard Oakley III, David Klimstra, Thomas Fuchs, Jorge Reis-Filho. Novel Mechanisms of CDH1 Inactivation in Breast Invasive Lobular Carcinoma Unveiled by the Integration of Artificial Intelligence and Genomics [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr GS03-04.

The value of AI-based analysis of fractional flow reserve of volume computed tomographically detected coronary artery stenosis with regard to their hemodynamic relevance.

The aim of our work was to demonstrate the importance of artificial intelligence-based analysis of fractional flow reserves of computed tomographically detected coronary artery stenosis with regard to their hemodynamic relevance in patients with unclear chest pain and suspected stable coronary heart disease with a low to medium pre-test probability.The collective of our retrospective analysis includes 63 patients in whom coronary artery stenosis was detected by volume computed tomographic examination in "one beat, whole heart" mode in the period from March to October 2022. In these patients, the fractional flow reserve was also determined by computed tomography, which was modulated by the use of artificial intelligence.The calculated values of the fractional flow reserve and the degrees of stenosis determined by computed tomography showed a moderate and significant negative correlation for all three coronary vascular territories (LAD/CX/RCA) (correlation coefficient rho = 0.54/0.54/0.6; p < 0.01 respectively). In just over a third (37.6 %) of all stenoses classified as high-grade by computed tomography, the assessment of hemodynamic relevance by calculating the fractional flow reserve deviated from the severity of the stenosis diagnosed by computed tomography, while the results in the peripheral areas "no stenosis/vascular occlusion" were 100 % consistent in each case.The present results of this work illustrate that the calculation of the fractional flow reserve based on artificial intelligence as a supplement to volume computed tomography of the heart can make a decisive contribution to further therapy planning by increasing the specificity of the purely morphological method by the physiological aspect. · Calculation of fractional flow reserve is a useful addition to computed tomography of the heart.. · It provides possibility to dispense with unnecessary further diagnostics by increasing specificity.. · The combination of both procedures leads to therapy optimization for patients.. · Noblé H, Mühlbauer N, Ehling J et al. The value of AI-based analysis of fractional flow reserve of volume computed tomographically detected coronary artery stenosis with regard to their hemodynamic relevance. Fortschr Röntgenstr 2024; DOI: 10.1055/a-2271-0887.

Diagnosis of soil-transmitted helminth infections with digital mobile microscopy and artificial intelligence in a resource-limited setting.

Infections caused by soil-transmitted helminths (STHs) are the most prevalent neglected tropical diseases and result in a major disease burden in low- and middle-income countries, especially in school-aged children. Improved diagnostic methods, especially for light intensity infections, are needed for efficient, control and elimination of STHs as a public health problem, as well as STH management. Image-based artificial intelligence (AI) has shown promise for STH detection in digitized stool samples. However, the diagnostic accuracy of AI-based analysis of entire microscope slides, so called whole-slide images (WSI), has previously not been evaluated on a sample-level in primary healthcare settings in STH endemic countries. Stool samples (n = 1,335) were collected during 2020 from children attending primary schools in Kwale County, Kenya, prepared according to the Kato-Katz method at a local primary healthcare laboratory and digitized with a portable whole-slide microscopy scanner and uploaded via mobile networks to a cloud environment. The digital samples of adequate quality (n = 1,180) were split into a training (n = 388) and test set (n = 792) and a deep-learning system (DLS) developed for detection of STHs. The DLS findings were compared with expert manual microscopy and additional visual assessment of the digital samples in slides with discordant results between the methods. Manual microscopy detected 15 (1.9%) Ascaris lumbricoides, 172 (21.7%) Tricuris trichiura and 140 (17.7%) hookworm (Ancylostoma duodenale or Necator americanus) infections in the test set. Importantly, more than 90% of all STH positive cases represented light intensity infections. With manual microscopy as the reference standard, the sensitivity of the DLS as the index test for detection of A. lumbricoides, T. trichiura and hookworm was 80%, 92% and 76%, respectively. The corresponding specificity was 98%, 90% and 95%. Notably, in 79 samples (10%) classified as negative by manual microscopy for a specific species, STH eggs were detected by the DLS and confirmed correct by visual inspection of the digital samples. Analysis of digitally scanned stool samples with the DLS provided high diagnostic accuracy for detection of STHs. Importantly, a substantial number of light intensity infections were missed by manual microscopy but detected by the DLS. Thus, analysis of WSIs with image-based AI may provide a future tool for improved detection of STHs in a primary healthcare setting, which in turn could facilitate monitoring and evaluation of control programs.

Improving sensitivity through data augmentation with synthetic lymph node metastases for AI-based analysis of PSMA PET-CT images.

We developed a fully automated artificial intelligence(AI)AI-based-based method for detecting suspected lymph node metastases in prostate-specific membrane antigen (PSMA)(PSMA) positron emission tomography-computed tomography (PET-CT)(PET-CT) images of prostate cancer patients by using data augmentation that adds synthetic lymph node metastases to the images to expand the training set. Synthetic data were derived from original training images to which new synthetic lymph node metastases were added. Thus, the original training set from a previous study (n = 420) was expanded by one synthetic image for every original image (n = 840), which was used to train an AI model. The performance of the AI model was compared to that of nuclear medicine physicians and a previously developed AI model. The human readers were alternately used as a reference and compared to either another reading or AI model. The new AI model had an average sensitivity of 84% for detecting lymph node metastases compared with 78% for human readings. Our previously developed AI method without synthetic data had an average sensitivity of 79%. The number of false positive lesions were slightly higher for the new AI model (average 3.3 instances per patient) compared to human readings and the previous AI model (average 2.8 instances per patient), while the number of false negative lesions was lower. Creating synthetic lymph node metastases, as a form of data augmentation, on [18F]PSMA-1007F]PSMA-1007 PETPET-CT-CT images improved the sensitivity of an AI model for detecting suspected lymph node metastases. However, the number of false positive lesions increased somewhat.

Abstract 4877: Morphological detection of chemotherapy resistance in cancer cell lines using AI-based analysis

Abstract Chemoresistance can develop during the course of treatment of many cancers, which can limit the utility of standard therapeutics and lead to poor outcomes. The mechanism by which resistance arises is not well understood, and developing models to study resistance can be challenging. Resistance can spontaneously arise in cell lines, however, it is difficult to separate resistant from sensitive cells using traditional biomarkers, hampering their utility for studying mechanisms. Morphology presents another modality to examine resistance, independently of protein markers, and may solve a previously intractable problem of separating resistant from sensitive cells for further study. Here we sought to study the mechanism of chemoresistance in cell lines using multi-parameter morphology in combination with gene expression and other standard molecular approaches. We induced resistance to specific chemotherapeutic drugs in parent cancer cell lines from lung, ovarian, and melanoma cancers. IC-50s of each drug increased by ≥1 order of magnitude in the resistant cells. Whole exome sequencing and bulk RNAseq were performed to examine differentially expressed genes and pathways between the parent and resistant cell lines. Each cell line was analyzed using multi-dimensional morphology using the Deepcell platform, which extracts &amp;gt;100 morphological features from images of unlabeled single cells using artificial intelligence (AI), advanced imaging, and microfluidics. The morphology features represent both Deep Learning (DL)-derived features and human-interpretable morphometric features, such as size and shape. Unique morphotypes were observed differentiating each combination of parent and resistant cell lines, with notable changes in granularity distinguishing them. A subset of images was used to train a random forest classifier to predict the resistant state of each cell image, which performed with up to 85% accuracy on independent cell images. Strikingly, the combination of DL plus morphometric features outperformed the use of morphometric features alone, demonstrating the increased utility of combined analysis. The top gene pathways and top morphological features were examined to infer potential mechanisms underlying chemoresistance. This work demonstrates that morphology alone can be used to distinguish drug resistant vs. parent populations without the use of markers. Morphology analysis can be applied to understanding complex phenotypes, and future platform improvements will enable sorting of these resistant populations to better identify molecular pathways in both cell lines and in primary tumor samples. Citation Format: Andreja Jovic, Manisha Ray, Ryan Carelli, Kiran Saini, Tiffine Pham, Christian Corona, Stephane C. Boutet, Chassidy Johnson, Mahyar Salek, Maddison Masaeli, Matt Barnes, Cyril Y. Ramathal. Morphological detection of chemotherapy resistance in cancer cell lines using AI-based analysis [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 4877.

Abstract 4933: A workflow for cloud-based AI development of multiplex IF image analysis using the OMERO Plus platform

Abstract Background: The advancement of AI in digital pathology is essential for analyzing image data containing billions of cells. This is true for digital pathology workflows using bright-field images, but this is even more critical for multiplex immunofluorescence (mIF) whole-slides due to the increased number of fluorescent channels and bit-depth per pixel. These complexities necessitate improved data management, and efficient visualization for both the mIF image and AI-based analysis results. We present a workflow using OMERO Plus, AWS SageMaker, and Ultivue software for AI model development in mIF. Method: Images were uploaded to AWS S3 and imported to the data management system OMERO Plus, where they were organized into training and validation groups. Annotations: PathViewer was used to place rectangular fields of view (FOV) in locations selected by the pathologist where all positive cells for each target marker were manually annotated. Data Preparation: Using the OMERO Plus API, annotation coordinates, image tiles, and image metadata were extracted and formatted for model training. All training data is under a version control system using Gitlab and Data Version Control tools using AWS S3. Additionally, strict quality checks were performed to ensure data integrity. Training: For both semantic segmentation and cell object detection, the training data was used in a customized training workflow using AWS SageMaker, including modern deep learning architectures, data augmentation, training logging, evaluation, and deployment. All AI models were trained using Pytorch and deployed with the open model format ONNX. Results Visualization: AI segmentation results of cells and regions are visualized using PathViewer as overlays on the original images by converting them to the OME-NGFF label image format that uses Zarr multidimensional arrays. Results: We utilized 58 and 24 mIF images for training and validation, respectively, on 15 markers using ISP Ultivue technology with 1681 and 296 FOVs, encompassing over 60K cells. Using this data and methodology we trained 3 semantic segmentation models and over five object detection models suitable for all twenty-four OmniVUE markers. For instance, the CD3 positive cell detection model, trained with 390 FOVs, achieved highly accurate results: f1-score of 0.915 evaluated on 34 FOVs. Conclusions: The combination of OMERO Plus, PathViewer, AWS SageMaker, and internal tools can successfully address mIF Big Data data management complexities in a cloud environment for digital pathology. The architecture is highly scalable, cost effective, and reliable, enabling a very efficient AI workflow. Citation Format: Ruben Cardenes, Erin Diel, Douglas Wood, Je Lee, Angela Vasaturo, Martin Schulze, Lorenz Rognoni. A workflow for cloud-based AI development of multiplex IF image analysis using the OMERO Plus platform [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 4933.

Deforestation detection and reforestation potential due to natural disasters—A case study of floods

Forests play a crucial role in our planet’s health and well-being, providing numerous ecological benefits and mitigating climate change. This research addresses the pressing issue of deforestation in flood-affected regions, focusing on the impact of climate change. Our study explores the potential of employing remote sensing data and AI-based analysis to develop a novel unsupervised learning approach for addressing the problem of deforestation detection, an area that has seen limited prior exploration. Utilizing Sentinel-2 data, unsupervised learning techniques are employed to identify deforested areas caused by floods in Pakistan. The quantitative results show that the forests in Pakistan reduced from 3.9% before the floods to about 3.41% after the floods. The study also explores and identifies the potential areas for reforestation which is estimated to be nearly 10.7% of the total area of Pakistan apart from the area covered by the forests. The findings underscore the urgent need for effective conservation strategies and reforestation efforts to mitigate the devastating loss of forest cover and safeguard our planet’s future.

Triple-negative breast cancer survival prediction using artificial intelligence through integrated analysis of tertiary lymphoid structures and tumor budding.

Triple-negative breast cancer (TNBC) is a highly heterogeneous and clinically aggressive disease. Accumulating evidence indicates that tertiary lymphoid structures (TLSs) and tumor budding (TB) are significantly correlated with the outcomes of patients who have TNBC, but no integrated TLS-TB profile has been established to predict their survival. The objective of this study was to investigate the relationship between the TLS/TB ratio and clinical outcomes of patients with TNBC using artificial intelligence (AI)-based analysis. The infiltration levels of TLSs and TB were evaluated using hematoxylin and eosin staining, immunohistochemistry staining, and AI-based analysis. Various cellular subtypes within TLS were determined by multiplex immunofluorescence. Subsequently, the authors established a nomogram model, conducted calibration curve analyses, and performed decision curve analyses using R software. In both the training and validation cohorts, the antitumor/protumor model established by the authors demonstrated a positive correlation between the TLS/TB index and the overall survival (OS) and relapse-free survival (RFS) of patients with TNBC. Notably, patients who had a high percentage of CD8-positive T cells, CD45RO-positive T cells, or CD20-positive B cells within the TLSs experienced improved OS and RFS. Furthermore, the authors developed a comprehensive TLS-TB profile nomogram based on the TLS/TB index. This novel model outperformed the classical tumor-lymph node-metastasis staging system in predicting the OS and RFS of patients with TNBC. A novel strategy for predicting the prognosis of patients with TNBC was established through integrated AI-based analysis and a machine-learning workflow. The TLS/TB index was identified as an independent prognostic factor for TNBC. This nomogram-based TLS-TB profile would help improve the accuracy of predicting the prognosis of patients who have TNBC.

An Automated Comparative Analysis of the Exudative Biomarkers in Neovascular Age-Related Macular Degeneration, The RAP Study: Report 6

PurposeTo explore the differences in volume and distribution of exudative biomarkers across macular neovascularization (MNV) types and subtypes using artificial intelligence (AI). DesignCross-sectional study. MethodsWe applied an AI-based analysis of 34.528 OCT b-scans of 281 (250 unifocal, 31 multifocal) MNV3, 55 MNV2, and 121 (30 polypoidal, 91 non-polypoidal) MNV1 treatment-naïve eyes. We compared the means (standard deviations) and heat maps of cystic intraretinal fluid (IRF), subretinal fluid (SRF), pigment epithelial detachments (PED) and hyperreflective foci (HRF) volumes, as well as retinal thickness (RT) among MNV (sub)types. ResultsMNV3 had the highest mean of IRF with 291 (290)nl, RT with 357 (49)µm, and HRF with 80 (70)nl, p≤0.05. MNV1 showed the greatest mean of SRF with 492 (586)nl, MNV3 the lowest with 218 (382)nl, p≤0.05. Heat maps showed IRF confined to the center while SRF scattered in all types. SRF, HRF and PED were more distributed in the temporal macular half in MNV3. Means of IRF, HRF and PED were higher in multifocal than in unifocal MNV3 with 416 (309)nl,114 (95)nl, and 810 (850)nl, p<0.05. Compared to the non-polypoidal subtype, the polypoidal subtype had greater means of SRF with 695 (718)nl, HRF 69 (63)nl, RT 357 (45)µm and PED 1115 (1170)nl, p<0.05. ConclusionThis novel quantitative AI analysis showed that SRF is a biomarker of choroidal origin in MNV1, whereas IRF, HRF and RT are rather biomarkers of retinal origin in MNV3. Polypoidal MNV1 and multifocal MNV3 present with higher exudation compared to other subtypes.

Street view imagery: AI-based analysis method and application

Street view imagery is an emerging form of geographic big data. It presents urban visual environments from the perspective of urban residents and also contains non-visual environment of cities, such as urban human activities and socio-economic development. However, traditional digital image processing has its limitations, and the continuous development of artificial intelligence, especially computer vision and deep learning, provides strong technical support for exploring the rich semantic information in street view imagery. This paper reviews the related research on street view imagery and its artificial intelligence analysis methods and applications. It outlines the acquisition, storage, and common data sources of street view imagery. Then it introduces computer vision, deep learning, and commonly used open-source datasets in street view imagery analysis. It also detailed three aspects of AI-based street view imagery applications, namely quantification of the physical space, urban perception, and spatial semantic speculation. Finally, issues like data acquisition, domain adaption and deep learning black box are discussed. The hotspots and prospects for the development of this research topic are also prospected.