• Samir Atiya

We discuss the current landscape of artificial intelligence (AI) in healthcare, exploring the advancements and challenges that define its role today. Drawing inspiration from the classic film "The Good, the Bad, and the Ugly," we navigate through the complex interplay of patient rights, data governance, and healthcare equity as they relate to AI. As we look to the future, we address the ongoing debates and the critical steps required to harness AI's full potential while safeguarding patient rights and promoting equitable healthcare. Join for a comprehensive exploration of AI in healthcare, where we will unpack the complexities, celebrate the advancements, and confront the challenges head-on, all in the spirit of driving meaningful progress in the field of laboratory medicine.

• Anil Parwani, OSU

Biobanks play an integral role in the nexus of personalized medicine, big data, and AI. Biobanks, as essential repositories for multimodal data, including digital pathology (DP) data and whole slide images (WSIs), are critical for supporting AI-driven research and clinical applications. This presentation will discuss the critical importance of biobanks in collecting and preserving biological samples, thereby enabling complex analyses and the development of personalized medicine solutions. We will explore how biobanks like the Cooperative Human Tissue Network (CHTN), a nationwide nonprofit offering academic and commercial investigators queried cancer samples from curated collections, are uniquely positioned to facilitate advanced medical research by providing high-quality, well-characterized biospecimens that are used to train and validate AI models. These models are increasingly applied in fields such as genitourinary (GU) oncopathology, where AI-enabled technologies offer greater diagnostic and prognostic precision and new disease insights from previously undiscoverable morphologic and molecular biomarkers. This session will also highlight the key challenges and innovations within the biobanking sector, focusing on aspects such as data management, interoperability, and ethical considerations in the collection and use of human tissue samples.

• Drew Williamson, Emory University School of Medicine, United States

This presentation explores the history of extraction of features from histology images, the images that pathologists use to make gold standard diagnoses for many disease. We will also explore the various options available in 2024 that are being used to extract these features via computational methods.

• Liron Pantanowitz

Artificial Intelligence (AI) has the potential to transform the field of pathology by addressing unmet needs and enhancing the efficiency and accuracy of diagnostic processes. This presentation aims to explore the feasibility of implementing AI in clinical pathology practices. We review the specific unmet needs within pathology that AI technology can potentially fulfill. We discuss the synergy of digital pathology with deep learning technologies. We also present various generative and non-generative AI applications in pathology.

• Ankush Patel
• Quincy Gu

Misunderstandings between pathologists and machine learning experts have hindered the effective integration of computationally-aided diagnostic (CAD) tools. Often, pathologists may misinterpret or misuse CAD tools due to limited comprehension of their design and purpose, leading to significant clinical implications. Conversely, computer scientists might develop algorithms that, while technically proficient, lack clinical relevance, resulting in high metrics but poor clinical utility. Enhanced understanding of computational terminology and processes from pathologists can foster a more accurate evaluation of algorithm strengths and weaknesses. This is crucial as pathologists, custodians of laboratory data, play a pivotal role in curtailing data to train algorithms for broader applicability. Our proposed framework offers a standardized approach for aligning clinical goals with their computational execution. By doing so, it facilitates a shared language between computational scientists and pathologists, bridging the existing gap and ensuring a more synergistic collaboration in the field of digital pathology.

• John Graff

In the era of advancing Artificial Intelligence (AI), leveraging laboratory archival information (AI) emerges as a crucial tool for steering us through the complexities. With historical data serving as the bedrock of diagnostic truth, AI integrates this repository for modeling and testing. Yet, as novel laboratory tests reshape disease classification, AI adapts, accommodating evolving diagnostic paradigms. Harnessing the power of archival data, AI not only safeguards against AI-driven errors but also pioneers dynamic diagnostic frameworks for the future.

• Jason Hipp

• Steven Hart

In this presentation, we delve into the multifaceted nature of clinical AI deployment, emphasizing its intrinsic multimodal characteristics. The integration of artificial intelligence and machine learning within clinical environments transcends simple data analysis, encompassing a complex interplay of multidimensional data, technological frameworks, procedural nuances, and interdisciplinary collaboration. We advocate for the adoption of the Clinical AI Readiness Evaluator (CARE) framework, presenting it as a rigorous methodology to navigate and optimize this intricate landscape, thereby enhancing the efficacy, responsibility, and sustainability of healthcare innovations. This session is an invitation to the scientific community to reconceptualize clinical AI deployment as a rich, multidisciplinary endeavor, necessitating a concerted and informed approach. We will examine strategies to harness this complexity, ensuring that our advancements in AI-driven healthcare are both scientifically robust and aligned with the ultimate goal of enhancing patient outcomes.

• Ben Cahoon, Techcyte, United States

Digital pathology adoption has been slow, but is moving into the next phase. What are the lessons learned from real world deployments and what does the next stage of digital pathology look like?

• Essa Mohamed

The rise and digitization of electronic medical records have enabled biomedical communities to identify cures and treatment of diseases in a relatively short period of time. However, even with this rapid growth of industrialization of “artificial intelligence,” we are still observing the same health inequities as we have seen over the decades. A new method and approach are warranted to ensure that we are not repeating the failures of medicine into this era of rapid growth in computational medicine. We will explore where the principles and scientific lens of social science can contribute to this paradigm for an improved understanding of human disease.

• Samvrit Rao, Thomas Jefferson High School for Science and Technology, United States

Chronic Obstructive Pulmonary Disease (COPD), afflicting over 300 million people globally, constitutes a significant contributor to morbidity and mortality, with patients manifesting severe COPD having a 2-year survival rate of 50%. COPD is characterized by chronic respiratory symptoms like difficulty breathing, cough and fluid filled airways, at times worsening rapidly in episodes known as exacerbations. Despite treatment advancements, generic "one size fits all" approaches fail to prevent COPD exacerbations in most patients, emphasizing the need for personalized treatment strategies due to the disease's varied pathogenesis. The primary goal of my project was to use electronic health record (EHR) data to characterize COPD exacerbations, apply computational methods for identifying patient clusters with similar disease indicators, and develop personalized treatment strategies to enhance patient outcomes. Data from over 8300 participants was extracted from deidentified clinical notes through natural language processing techniques and approximately 20 clinical features were extracted per patient. K-means clustering was then used to categorize the cohort into three distinct and homogenous subgroups based on disease indicators such as serum biomarkers, comorbidities, symptoms, demographics, and respiratory characteristics. Forced Expiratory Volume (FEV1) and breath sounds were the key differential indicators that were identified among these subgroups as identified through Principal Component Analysis. Random forest and Bayesian optimization were employed to build a predictive model in patient subgroups, considering the presence or absence of exacerbations alongside treatment data, aiming to predict personalized treatment approaches while incorporating real-time breath sounds and FEV data for enhanced efficacy. Prognostic matching was then employed alongside the predictive model to further optimize treatment plans based on empirical data. These results were then validated by physicians with a high level of confidence, significantly higher than the physician prescribed treatment plans for the same patient. Further, in collaboration with a telemedicine startup, efforts are underway to integrate these models into EHR modules to enable patient-specific treatments, improving outcomes and survival rates. This methodology marks a significant improvement in the treatment plan of Chronic Respiratory Diseases, helping physicians in their decision making process.

• Jinge Wang, Department of Microbiology, Immunology & Cell Biology, West Virginia University, Morgantown, WV 26506, USA, United States
• Qing Ye, West Virginia University Cancer Institute, West Virginia University, Morgantown, WV 26506, USA, United States
• Li Liu, College of Health Solutions, Arizona State University, Phoenix, AZ 85004, USA, United States
• Nancy Guo, West Virginia University Cancer Institute, West Virginia University, Morgantown, WV 26506, USA, United States
• Gangqing Hu, Department of Microbiology, Immunology & Cell Biology, West Virginia University, Morgantown, WV 26506, USA, United States

Emerging studies underscore the promising capabilities of large language model-based chatbots in conducting basic bioinformatics data analyses. The recent feature of accepting image inputs by ChatGPT, also known as GPT-4V(ision), motivated us to explore its efficacy in deciphering bioinformatics scientific figures. Our evaluation with examples in cancer research, including sequencing data analysis, multimodal network-based drug repositioning, and tumor clonal evolution, revealed that ChatGPT can proficiently explain different plot types and apply biological knowledge to enrich interpretations. However, it struggled to provide accurate interpretations when color perception and quantitative analysis of visual elements were involved. Furthermore, while the chatbot can draft figure legends and summarize findings from the figures, stringent proofreading is imperative to ensure the accuracy and reliability of the content.