Patholytix AI is the right solution to leverage Deep Learning and integrate AI-based approaches into the toxicologic pathology field. Coupled with a viewer tailored to the evaluation of nonclinical safety studies, it combines a user-friendly annotation process and some powerful AI models, together with visual maps of the generated masks and quantitative pathology capabilities, providing valuable diagnostic decision support to the pathologists.
Patholytix AI seamlessly integrates the contributions of pathologists and AI engineers to create robust deep-learning models. The outputs of these models can be used to generate an easy to interpret visual for your whole slide images which streamline the process of detecting abnormalities caused by drug toxicity.
Patholytix AI combines an ergonomic user interface with machine learning that requires minimal data input. The result is a simpler, more efficient workflow.
Patholytix AI employs powerful, state-of-the-art models including InceptionV3, ResNet50, Xception, DeepLab v3Plus and Segmodel. Such models are trained on large and diverse datasets from different sources which ensures they work exceptionally well at detecting abnormalities on your own slides. This also means there is minimal training data required from you to get these robust outputs.
Our AI Pipeline is optimised and scalable across multiple GPU engines. As a result, the training of your AI classifiers will always operate as fast as possible.
Within our software, we have tools that can provide quantitative outputs for your data.
Interactive segmentation masks allow identification and quantitation of up to 1 million individual features within an image, with feature-specific morphometry, including area, roundness, perimeter and orientation.
Such quantitative outputs can help you to determine if an abnormality is of a level (lesion count or lesion area) indicative of drug toxicity.
This Quantitative Pathology tool allows you to determine how likely a specific area of tissue is abnormal. Individual probability masks views are available for each lesion. In addition to the information on where the abnormality could be on the tissues, it provides information on how probable each lesion is in an identified region.
The quicker the review is complete, the quicker a decision can be made on the safety of the drug.
Visual outputs (heatmaps) from Patholytix AI allow you to quickly find and interrogate results without having to sift through massive amounts of data. These heatmaps can highlight changes in cell density. Variances in cell density can be indicative of cellular changes caused by drug effects.
Set your parameters. Our AI software will then identify and count cells of interest. An example? Inflammatory cells such as macrophages, or necrotic cells, can be detected by adjusting segmentation mask threshold parameters to only detect cells of this morphology.
See and review the slides of concern first.
You can review what the mask is picking up as a lesion. If it’s detecting areas that you feel are not of concern, you can adjust the thresholds accordingly. Then you will only see lesions of concern, and be directed to these slides more quickly.
You can sort entire study cohorts based on the extent of the segmented lesion present. Samples with lesions indicative of drug toxicity may be filtered based on their quantitative parameters to view those highest in lesion count or area first.
SPEED UP YOUR REVIEWS AND DRUG DEVELOPMENT TIME
REVIEW & SCORE DRUG DISCOVERY STUDIES DIGITALLY FOR THE FIRST TIME
ACHIEVE IMAGE QUALITY AT SPEEDS BEYOND HUMAN CAPABILITIES
REACH CRITICAL DECISIONS FASTER, AND WITH GREATER CONFIDENCE
END-TO-END GLP-COMPLIANT DIGITAL STUDIES (A WORLD FIRST)
UNLOCK HISTORICAL PATHOLOGY KNOWLEDGE