Aiforia Workshop

An afternoon affair with Aiforia: Advancing preclinical and diagnostic pathology with AI

Presentations will be given by:

Thomas Westerling-Bui PhD, VP of AI Solutions & Strategy at Aiforia

Practical AI applications from research to diagnostics
Anatomical Pathology is a diverse field with a deep complexity of visual cue tasks. An entire profession with sometimes narrow subspecialties exists to support the correct and most effective healthcare delivery. In addition they support Rx, Dx, and clinical development from translational research, trials, to deployment in the most cost efficient fashion. Enabling the automation of (some) these tasks is a daunting undertaking that involves all pathology stakeholders. Aiforia’s mission is to provide both a platform solution to enable the large-scale participation of the experts (pathologists) as well as the framework from which to deploy our own core AI model solutions and collaborative efforts at a global scale. We will discuss our vision throughout value chain and illuminate these with practical use cases.

Take home messages:
• How do you develop, validate (analytical), and deploy Clinical Grade AI solutions.
• Lessons in going to scale at an enterprise or global level
• Can any one (organization) cover all Anatomical pathology needs? (hint: We think we have the answer)
• We always listen to pathologists, come and challenge our assumptions!

Lucas Stetzik PhD, Parkinson’s Disease Center, Department of Neurodegenerative Science, Van Andel Institute

Fully automated quantification of histopathology in preclinical Parkinson’s’ Disease research.
Parkinson’s disease is a common and progressive neurodegenerative disease with annual societal costs exceeding $14 billion. Currently, there is a need for automated methods to quantify neuropathology across research settings with a high degree of consistency and reproducibility. Deep learning technology offers solutions for quantifying histopathology, but for many researchers there exists a gap between neuroscience and machine learning that is not usually possible to bridge without extensive computer science training. Fortunately, commercially available deep learning tools can meet this need and allow any research group with Internet access the ability to bridge this gap. The methods presented here demonstrate how we utilize these tools to quantify brain region-specific microglia activation, α-synuclein aggregation, and TH+ cell loss in a way that is fast, flexible and fully automated.

Alexander Klimowicz PhD, Senior Principal Scientist, Immunology & Respiratory Disease Research, Boehringer-Ingelheim

AI-based image analysis and large-scale data analysis for DSS Colitis

Valentina Perosa MD, Postdoctoral Fellow, J. Philip Kistler Stroke Research Center, Department of Neurology, Massachusetts General Hospital, Harvard Medical School.

Deep learning assisted quantitative assessment of histopathological markers of Alzheimer’s disease and cerebral amyloid angiopathy
Analysis of neuropathological markers in neurodegenerative diseases has relied on visual assessments of stained sections. Resulting semiquantitative scores often vary between individual raters and research centers, limiting statistical approaches. To overcome these issues, we have developed six deep learning-based models, that identify some of the most characteristic markers of Alzheimer’s disease (AD) and cerebral amyloid angiopathy (CAA). The deep learning-based models are trained to differentially detect parenchymal amyloid β (Aβ)-plaques, vascular Aβ-deposition, iron and calcium deposition, reactive astrocytes, microglia, as well as fibrin extravasation. The models were trained on digitized histopathological slides from brains of patients with AD and CAA, using Aiforia, a software, that allows neuropathology experts to train convolutional neural networks (CNNs) on a cloud-based graphical interface. Validation of all models indicated a very good to excellent performance compared to three independent expert human raters. Furthermore, the Aβ and iron models were consistent with previously acquired semiquantitative scores in the same dataset. The measures allowed the use of more complex statistical approaches. For example, linear mixed effects models could be used to confirm the previously described relationship between leptomeningeal CAA severity and cortical iron accumulation, to explore the association between neuroinflammation and Aβ pathologies, and that between CAA and enlarged perivascular spaces, a marker of impaired perivascular clearance in the brain. Moreover, ongoing work utilizes these measures to assess the relative spatial distribution of different markers, such as iron density and neuroinflammatory cells. In summary, the implementation of deep learning-assisted methods allows to address specific pathophysiological questions in neurodegenerative disease in a more objective way and possibly on a larger scale across different centers.

Refreshments will be provided.

Register Today