PathAI
Axial partners with great founders and inventors. We invest in early-stage life sciences companies such as Appia Bio, Seranova Bio, Delix Therapeutics, Simcha Therapeutics, among others often when they are no more than an idea. We are fanatical about helping the rare inventor who is compelled to build their own enduring business. If you or someone you know has a great idea or company in life sciences, Axial would be excited to get to know you and possibly invest in your vision and company . We are excited to be in business with you - email us at info@axialvc.com
Who leads PathAI?
PathAI was founded by Andrew Beck (CEO) and Aditya Khosla (CTO). Beck earned his medical degree at Brown and did his fellowship at Stanford and also completed his PhD there in biomedical informatics. He then joined the faculty of the Harvard Medical School as a professor. Khosla earned his PhD in computer vision and machine learning at MIT.
Beck’s research lab at Harvard was focused on the application of machine learning to pathology. He had read an article about Khosla’s research and emailed him to get coffee. After getting to know each other, PathAI was founded in the basement of Beck’s Brookline, MA home to bring AI to pathology and diagnosis of disease.
What does PathAI do?
The company applies computer vision and deep learning to pathology images. During a medical diagnosis, if a tumor is found after a radiology scan, a tumor biopsy is taken and sent to a pathologist who will make the final call on whether a patient has cancer or not. They analyze the biopsy, 100Ks of cells, under a microscope - in a clinical setting, a pathologist can analyze ~300 of these slides per day. PathAI is working to make the 10Ms of cells pathologists look at every day more manageable. Because the stakes are so high for patients and the likelihood of error present (error rates of 3%-9%), the company’s software can analyze the large amount of biopsy image data per day and accurately distinguish normal cells from cancerous ones. The approach of empowering pathologists with AI can create standardization for the diagnostic process, reduce error, and increase the leverage of an individual pathologist.
By digitizing the pathology lab, PathAI can not only improve cancer diagnosis but inform the treatment of the disease. Major themes for the company are:
Automating monotonous tasks for pathologists
Integrating histopathology with genomic data to inform therapeutic use
Using machine learning to find new cancer gene expression signatures, comprehensively label cell types for a biopsy, tumor grade, and mutational subtypes. The idea is to discover new features that have predictive power for a patient’s response to a medicine and survival.
Working to build a software that can help pathologists exceed the current gold standard in diagnostic sensitivity and specificity
Source: https://www.biorxiv.org/content/10.1101/2020.08.02.233197v1.full.pdf
What makes PathAI unique?
PathAI is not only making pathology more efficient but is building a technology platform that can unlock new capabilities for the field. Current pathologists focus on figuring out whether a biopsy is cancerous or not. Combining their expertise with PathAI’s software can free up time and add new features to also focus on what treatment options are best for an individual patient. As PathAI improves their cancer and cell classification models with more patient samples, the combination of their product with pathologists could soon exceed the standards in the field.
With models that have thousands of features across a database with millions to billions of images, the company has the potential to implement an end-to-end model that makes a diagnosis from raw images. Issues around bias, adversarial attacks, interpretability, among others is a major barrier to completing this vision. PathAI is on track to one day automate the field of pathology; in some use cases PathAI already is such as in detecting metastases in lymph node samples from breast cancer patients.
The 3 main applications for PathAI’s products are:
Helping pathologists make better diagnoses
Improving drug development through more informed therapeutic use during clinical trials and post-approval
Expanding gold standard pathology services to regions that currently do not have access
Source: http://www.globalengage.co.uk/pathology/docs/Lee.pdf
Why I like what PathAI is doing?
In short, PathAI is digitizing an important part of healthcare. By augmenting pathology, the company can improve accuracy of diagnosis and patient outcomes. The revenue potential for this alone is well above $10B. So PathAI still has a lot of work to do as they transform pathology.
The next step is informing drug development and use. A case study of this is PathAI’s work with Bristol Myers Squibb (BMS). The collaboration used PathAI’s software to score PD-L1 expression in cancer biopsies to assess their ability to respond to BMS’ anti-PD-1 antibody, Opdivo. The work found that PathAI was able to find more PD-L1-positive patients that would respond than manual inspection in melanoma and urothelial carcinoma. In a later study, retrospective analysis of PD-L1 expression in 2 trials (n=293) using Opdivo to treat advanced urothelial cancer, PathAI was found to be comparable to current pathologists and reach closer to the gold standard in the field. Ultimately, PathAI has the ability to help with patient selection for trials and therapeutic use. This market opportunity is worth $10Bs.
Overtime, PathAI is positioned to open up new opportunities within pathology (i.e. complete automation one day) and in other fields of medicine and healthcare. They could move into radiology. PathAI could bring on genomics capabilities. What else?
You can find PathAI here.