Shonit™: From a pathologist's point of view


When I first heard about digital image analysis using artificial intelligence and its intended use in routine hematology reporting, I was very excited but simultaneously I had my critical views. Some of the questions that immediately came to my mind were:

  • can a software actually analyze the cells?
  • can it be fast enough without creating delays in report delivery?
  • can it add value in any way in the quality of reporting ?

Having experienced and witnessed the journey of creation and training of Shonit™ for hematology peripheral blood smears, I not only got the answers to my initial questions but also found multiple promising benefits of it.

Our existing work flow of handling peripheral blood samples include processing of all the samples on a five-part analyzer. Results with “normal” counts are released to patients while for those with “abnormal“ counts (based on our slide review criteria), slides are made by an auto stainer. This is done for approximately 25 percent of the total cases. These slides are studied by a pathologist under the microscope to verify the abnormal counts or abnormal morphology of cells. This review normally includes seeing and counting a minimum of 100 white blood cells, studying the shapes and sizes of RBCs, estimating the adequacy and morphology of platelets and to see if there are hemoparasites. While a lot of information regarding the counts of cells is provided by the blood analyzer, it has its limitations in morphology analysis of cells and mandates viewing them under microscope. With massive workload and pressure of time to deliver reports to patients, it may not always be possible to spend much time with each slide. This increases the risk of missing certain critical details.

Shonit™ analyzes the morphology of a large number of WBCs and RBCs, identifies and quantifies large/immature platelets, identifies malarial parasites and various other features that add value to reporting. High quality images with good resolution are displayed along with the results. Since “seeing is believing”, this, in my opinion is the best feature and gives the pathologist the confidence and comfort that is required in signing out a report. The intended use of this software in our laboratory is to analyze all cases for which slides are made and view them on the computer (instead of microscope). The images and other morphological parameters are correlated with the counts obtained by the cell counter. In most cases, a good correlation can be found. For example when the cell counter flags for monocytosis and Shonit™ shows a high monocytes count along with the images of monocytes, the report can be finalized at this level. When there is a discrepancy or when there is a suspicion of abnormal or atypical cell, the slide will be viewed under microscope.

Sharing images and cases with colleagues to seek a second opinion, to report remote site cases where pathologists may not be available, using images from the archives for teaching purposes or presentations are other advantages Shonit™.

The goal of a laboratory practitioner is to provide high quality, error free reports delivered on or before the committed time. Much of this can be achieved by planning the work flow better, making necessary changes in the system and processes and utilizing available resources best suited for each setup. Having tried and tested Shonit™, I can say with conviction, that image analysis using AI in hematology will help the laboratory achieve higher quality of reporting and will definitely become an integral part of laboratory practice in the near future.