Initial validation of mobile Whole Slide Imaging (mWSI) with Northwell Health
Mobile Whole Slide Imaging is our proprietary method of generating digital whole slides. The process was created by Lou
Auguste, co-founder of Alexapath, and automated through the work of Shishir Malav, co-founder of Alexapath, along with
our dedicated team of engineers and developers at NYU Digital Future Labs.
The methods for creating mWSIs was first published in the British Medical Journal (BMJ Innovations, 2015). Following the
publication, we conducted our first clinical validation study with the staff of Northwell Health’s pathology department
led by Dr. Michael Esposito. The goal of the study was to determine if images created by the AiDA device were adequate
for diagnostic interpretation by the pathology team.
The results were presented as a poster at USCAP 2015. Images
created through the mWSI technique were recognized as valid for diagnosis.
Cervical cancer pre-screening on mobile devices with Qualcomm India
Roughly 1 in 5 Pap smears screenings will be abnormal, but for every woman between the ages of 21-65, screening is
recommended once every three years. In low resource areas, these screening programs are the difference between life and
death, but to the lab professionals, it can mean hours of routine work viewing cases that are negative. When coupled
with staffing shortages of diagnostic professionals, mistakes can be made and abnormal smears can be missed.
Our team innovated a workflow to automate the prescreening and increase the throughput of smears called AutoPap.
Developed with the support of engineers from Qualcomm’s Innovation Lab, AutoPap is a CNN capable of determining with 98%
accuracy if there are abnormal cells present in the digital whole slides created by AiDA.
AutoPap was designed using Tensor-Flow which is an interface for expressing machine learning algorithms and deploy the
model on devices. The CNN needed a GPU heavy system for training, but we were able to successfully deploy a lightweight
version of the trained algorithm on a mobile system using the SnapDragon 820.
The CNN and the dataset containing all 120,000 images now open source, and we are using it to teach high school students
how to develop AI. Currently, we are looking for partners with funding to test AutoPap in the field.
Detecting water parasites using HEAD AI with UNAM
Prof. Catalina Maya Rendon
Gustavo Velasquez, MSc
Researchers from UNAM were looking for a way to create whole slide images of rafters (chambered slides for holding water
specimens). All of our competitor whole slide scanners had a design flaw - slides had to be loaded into cartridges
before they were images. Those designs just didn’t hold water! In AiDA 2 they discovered a scalable solution.
We’ve heard this story before, guidelines exist to limit the density of Helminth eggs in various environmental matrices,
but there are not enough qualified technicians for the task of visual identification using conventional microscopy
across all regions in need.
With support from the Bill and Melinda Gates Foundation, we donated an AiDA 2 system to the team at UNAM. AiDA was used
to acquire mobile Whole Slide Images (mWSIs) of various concentrations of Helminth ova. The images were then analyzed by
a pre-classification algorithm in order to verify that illuminance, contrast, and blurriness met the established
threshold values. The testing validated that the AiDA system produced digital images that met the benchmarks needed for
successful analysis by AI.
Expanding Access to parasite testing for farm animals with vHive and Zoetis
Dr. Abel Ekiri (vHive Surrey)
Building on previous research with Qualcomm and UNAM. Our team in partnership with vHive Surrey is developing an
automated solution for the screening of stool samples. In many regions of sub-Saharan Africa, farmers share grazing
lands for their cattle. When a herd with parasites grazes on a common green, they can put all the neighbors' livestock
Facing a similar lack of diagnostic professionals we have identified as a key challenge facing many of our customers.
Together, vHive and Alexapath are set to trial a version of the HEAD algorithm capable of identifying parasite ova
present in stool samples. To speed the uptake of this automated diagnostic, we are employing a team of drivers to travel
from farm to farm to collect specimens.
The farms have access to a mobile application that shows of heat map of grazing areas that parasites may be present at.
The results of the testing are confidential and all information contained on the heat map is anonymous.