Artificial Intelligence and the Haematopathologist

separator

A smart haematology analyser powered by AI is the new aid for a Hematopathologist

The parting gift of the 20th century has been the unparalleled advancement in technology. As a consultant pathologist for the past 15 years with 8 years of specialized practice in haemato-oncopathology, I have witnessed a sea of technological transformation in laboratory and hospital practice. With a keen interest in innovation, I feel that I have sailed through a realm of technological transitions in the last few years.

The humble microscope has always been the faithful companion of a pathologist.  From time immemorial it has always been considered as the “gold standard” of the reporting arena. I started my training and career in pathology with a simple monocular microscope. Cumbersome and challenging in its own way I heaved a sigh of relief when we transitioned to the binocular. That was just the beginning of the journey. The next milestone was with the state of the art haematology analyser. My very first experience was with 3 part haematology analyser which adorned the central position of the lab. As a breakthrough in haematology practice, it provided an efficient alternative to the manual counting of cells on a haemocytometer. Although it took care of the total leucocyte and erythrocyte count, the 5 part differential count was still a challenge and the only resort was manual microscopy.

Replacement of the 3part haematology analyser with a 5 part was the next step in upgradation. By that time, I was heading the haematology division at Bangalore’s tertiary cancer centre – HCG. The 5 part analyser did take away the burden of the existing 3 part but it came with its own set of limitations.  Patients on chemotherapy have fluctuating counts. Cancer in itself can create a disequilibrium in haemostasis and blood counts. 60 to 70% of the samples must be manually verified for presence of blast /atypical cells, disease relapse and dips in platelet counts post therapy. In such situations, a CBC test can provide the most important parameters which can aid decisions in treatment and choosing options in blood transfusion.

With the ever rising multifactorial causes in disease states and increase in laboratory work, the CBC remains as a basic test which can throw light in narrowing down the differential diagnosis. A sea of information can be obtained through this simple screening test. Hence, it becomes equally imperative to have a solution which can provide fast, accurate and equally cost-effective results.

Artificial intelligence heralds an age where well designed and well executed AI algorithms can solve complex medical problems including the interpretation of diagnostic images.  The challenges of expanding medical knowledge, shortage of medical experts and scaling of tests make it the need of the hour.

SHONITTM is one such smart solution for automated analysis of blood smears powered by data driven intelligence, image processing and cloud computing.  All set to revolutionize the world of haematology and hematopathology, SHONIT has found its place in modern day health care system.  It can compute the total leucocyte, erythrocyte and platelet count along with morphological analysis of the same. By achieving an accuracy of over 98% in the 5 part differential count it can not only act as an augmentation to the existing 3 part analyser but has also challenged and surpassed the 5 part haematology analyser with the following advantages:

  • SHONITTM has provided accurate differentials of rare cells in leukopenia cases.
  • It is effective in the categorization of monocytes, eosinophils and basophils.
  • It is highly accurate in enumeration of nRBCS in cases of haemolytic anaemia which is computed as an independent parameter. This provides the added benefit of not performing a corrected total WBC count.
  • Last but not the least it can also enumerate Immature granulocytes.
  • Speaking of the erythrocyte model, it would be important to mention that the classification of anisopoikilocytosis is performed accurately by SHONITTM. For e.g. A report of haemolytic blood picture with the presence of shistocytes and nRBCs can be given with complete confidence. Or the presence of target cells in a case of Thalassemia.
  • Platelets counts have always posed as a troubled zone with conflicting results of analysers and manual reports, exhibiting uncontrolled dynamism and fluctuation in counts due to platelet refractoriness and drugs. SHONITTM has the capability to accurately identify and morphologically analyse the platelets with supporting visual evidence. Platelet count by SHONITTM is just a few validations away.

Haematology analysers come with their own set of limitations. Adoption to new features has always been a challenge and costly affair. Not to mention the costs involved in reagent consumption and hardware maintenance. Most importantly there is no visual evidence of classification.

The “Flags” triggered in a haematology analyser is always an indication for manual review which can be a fatigue driven process.  Manual microscopy has resulted in variability, inaccuracy and inefficiency in reports. Furthermore, the number of cells scanned by manual microscopy is only 200 at a time. With the increase in workload in most laboratories, technicians and pathologists end up counting only 100 cells in most cases. SHONITTM scans around 120 FOVs in contrast to manual microscopy. This enhances the sensitivity in identifying rarer cells on the smear like basophils, monocytes and blasts in cases of pancytopenia which would normally evade the human eye.

Most laboratories today are TAT(turn around time) driven. The turnaround time forms an important quality indicator in an audit process. National and International Accreditation bodies demand a thorough analysis of TAT outliers. SHONITTM provides an optimal solution for this very important requirement. The time taken for a pathologist to authenticate a CBC report on the AI platform is within 2 minutes or less. SHONIT demonstrates a superior TAT to existing methodologies.

With the increasing trend in telepathology, SHONIT can serve as the perfect solution to enable telepathology especially in centres which operate on a hub and spoke model. Needless to say, a second opinion from an expert can be obtained quickly and efficiently through the cloud based platform.

SHONITTM provides quality reports which are accurate and efficient. It saves time, accelerates the turnaround-time and increases productivity of pathologists. It has helped to overcome the time-consuming effort associated with traditional microscopy. It can aid screening, diagnosis and analyse large batches of data. It can leverage multiplexing and drive down stream processes as well.

As pathologists, our professional value comes from our ability to give the most appropriate opinion based on visual images which amalgamates with the clinical background. Artificial intelligence gives this opportunity and the means to overcome bias of the human mind.  Artificial intelligence is here to stay and through SHONITTM I have found a powerful tool as an aid in hematopathology.