Aiforia Workshop

Enabling Patient Care and Discovery by Embracing AI and Digital Pathology

Please join Aiforia to hear how pathologists and investigators from the Mayo Clinic, MGH/Harvard Medical School, Rutgers University, and Memorial Sloan Kettering Cancer Center are enabling patient care, translational research, and discovery by embracing artificial intelligence (AI) and digital pathology. Each has used Aiforia’s cloud-based digital pathology platform to develop AI models with the goal of advancing research and improving patient care. Speakers will share lessons learned as they developed, published, and applied the AI models to their clinical and research data sets.

The individual presentations will be followed by a panel discussion with the speakers.

3:05 PM – 5:20 Talks
5:20 PM 30min panel discussion
5:50 PM Happy hour with food and drinks, Stay to discuss with Speakers and the Aiforia team

Speakers:

Dr. Nilay Bakoglu, Memorial Sloan Kettering Cancer Center

Artificial Intelligence (AI)-Based Automated Determination of Histomorphological Features and Mitosis Detection in Multiple Organ Systems

AI and other neural networks have made great progress in capturing the detection of morphological diagnoses. In this study, we present our experiences in invasive area detection of breast and colon cancer, tissue elements detection in colon specimens, and atypical and typical mitosis detection in several organs. 344 H&E WSIs and WMIs of 228 cases were trained, tested, and validated in the AI platform (Aiforia Inc). The final AI models showed F1 score of 94.49% for Invasive carcinoma of the breast, 97.40%, and 97.68% for atypical and typical mitosis layers, and 98.02% for vessels and lymph nodes respectively. We aim to use exported data of AI models for defining prognostically significant findings in WSIs to apply on Whole Block imaging including whole-mount samples.

Dr. Alexandra Zuraw, Digital Pathology Place

The Importance of Quality Control in Digital Tissue Image Analysis

Digital tissue image analysis is a powerful research method, enabling the extraction of qualitative datasets from whole-slide images. However, ensuring reliable results requires a robust quality control system.
This system entails clear procedures, comprehensive training for team members, and the involvement of experts to oversee critical stages of the workflow. Pathologists play a crucial role as the ultimate tissue experts, responsible for quantitatively and qualitatively assessing the validity of image analysis solutions.

This talk will cover common endpoints of digital tissue image analysis and discuss the potential sources of errors. Moreover, best practices derived from the Food and Drug Administration’s recommendations for AI and machine learning-based software used in medical devices will be presented.

Dr. Valentina Perosa, MGH/Harvard Medical School

Pathologies of the Medial Temporal Lobe in Alzheimer’s Disease: A holistic perspective

Cerebral small vessel disease(CSVD), including cerebral amyloid angiopathy(CAA) and arteriolosclerosis, commonly co-occur with Alzheimer’s disease and related dementias(AD/ADRD). The medial temporal lobe(MTL), a brain area crucial for memory and global cognition, is prone to harboring multiple pathologies, such as neurofibrillary tangles(NFTs), amyloid-β-plaques, TDP-43, and CSVD. Whether a causal relationship between these pathologies exists remains largely unknown. One possible linking mechanism is the dysfunction of perivascular clearance, the system responsible for the disposal of toxic metabolic waste products in the brain. Using deep-learning (Aiforia®), we aimed to analyze the burden of pathologies in the MTL of an AD-cohort and to determine the relationship between CSVD and enlarged perivascular spaces, a marker of clearance dysfunction.

Dr. Colin Ziabakhsh, Rutgers University

Artificial Intelligence Analysis of Rat Hypoxic Ischemic Encephalopathy Treated with Cord Blood Exosomes

Hypoxic ischemic encephalopathy (HIE) is brain damage at birth. We developed a new bilateral carotid ligation model followed by an hour of 8% hypoxia in 7-day-old rats. This model is more clinically relevant than the Rice-Vanucci model, which permanently occludes one carotid artery followed by hypoxia. We anesthetized 7-day-old rats with isoflurane, ligated both common carotid arteries (CCA), exposed them to 8% oxygen for an hour, and reperfused them. The rats received intraperitoneal human umbilical cord blood (UCB) exosomes at 24 hours. We used two artificial intelligence (AI) programs to do tissue-wide neuron counts and area measurements of Cresyl Violet stained brain sections. The results were remarkably consistent and surprising. We will continue to use Aiforia software to evaluate this HIE model.

Dr. Tom Westerling-Bui, Aiforia, CCO, Americas

Transforming diagnostic pathology by leveraging digital pathology and AI requires a concerted effort by many stakeholders.

A significant impact of AI in diagnostic support requires a broad application strategy across the spectrum of AP (and visual cue-based CP). We will discuss challenges, opportunities, and solutions by highlighting recent work with our various partners. Adoption of the Aiforia Clinical platform and the Create platform as a translational research product can drive AI models to the clinic at scale. Case studies presented include a model for assessing eosinophilic esophagitis developed with Dr. Moreira (Mayo Clinic) and QuantCRC, by Dr. Pai (Mayo Clinic), that provides a powerful AI adjunct to routine pathologic reporting of colorectal cancer and a prognostic model that can improve prediction of recurrence-free survival.

Register Today