Further enhancements of machine learning strategies for hematology analytics with near-field microscopy

Alentic Microscience Inc has developed a highly portable device that performs complete blood counts and other in
vitro diagnostic tests based on the lensless microscopic imaging of blood and reagents. The system uses
proprietary sampling techniques and unique analysis software based on advanced machine learning techniques.
The software has been designed as a multistep process of artifact removal, recognition of cell localization, image
enhancement, and classification.
We have previously investigated several possible advancement techniques within a two-term internship of a
Master student. Some of the anticipated improvements such as vision transformers have not shown major
improvements, but we made progress with multi-stage processing for recognition of difficult classes. Also, the
investigation of uncertainty within class shows important potentials that we will further develop in this internship. In
addition, we have devised several new potential improvement strategies specifically on the super-resolution
strategy as well as the processing of context in different clinical settings.

Faculty Supervisor:

Thomas Trappenberg;Dirk Arnold

Student:

Partner:

Alentic Microscience Inc

Discipline:

Computer science

Sector:

Manufacturing; Professional, scientific and technical services

University:

Dalhousie University

Program: