Keeping our eye on AI in Hematology

In today’s world it is all eyes on AI with Artificial Intelligence having catapulted us…
Read more
Avatar
By Gil Ben-Horin

In today’s world it is all eyes on AI with Artificial Intelligence having catapulted us into a new fast-paced reality. This tenet holds especially true within the healthcare setting, whereby this transformative force holds the golden promise of heightened diagnostic accuracy, reduced costs, and improved patient response times and treatment outcomes.

The use of AI in hematology diagnostics is increasing, and it could significantly simplify hematology diagnosis by integrating findings from various diagnostic techniques. AI blood test analyzers using deep learning models have demonstrated encouraging results in decreasing variability among observers and improving diagnostic accuracy. Machine learning and deep learning models have made it possible to detect and categorize patterns within data, resulting in the creation of AI systems that have been utilized across different aspects of digital hematology.

The challenge in developing clinical AI systems lies in ensuring their accuracy and reliability by generating consistent quality results and the successful integration into clinical practice.

One such challenge is the quality and availability of data. AI algorithms are heavily dependent on the data they are trained on. Any biases or lack of representation in the patient population can result in inaccurate and biased outcomes. To overcome this, it is crucial to ensure the data used for AI training is of high quality, free from biases, and encompasses variations in patient backgrounds and clinical conditions.

The successful adoption of clinical AI systems also hinges on their user-friendliness. Designing user interfaces that allow for intuitive and easy navigation and presenting AI system outputs in a clear, understandable format, are critical for decision-making.

Healthcare and laboratory professionals, with diverse levels of technical expertise, need to be adequately equipped to effectively utilize these systems in their everyday workflow, thereby increasing confidence and user comfort level.

In the relatively new field of AI integration into healthcare, unfamiliarity with the technology can act as a barrier to effective implementation. Choosing appropriate AI programs may be daunting, with some laboratories relying on vendor marketing information that might be exaggerated. Understanding the specific tasks AI can assist with, can aid in overcoming such concerns.

For individuals less acquainted with AI, entrusting an AI model with diagnostic tasks typically managed by specialized staff can pose a significant trust hurdle. Alternatively, introducing AI in areas perceived as lower risk, such as assessing and streamlining workflow, reducing manual processes,  or identifying patterns in test utilization that require improvement may be a useful first step. This gradual integration of AI into less critical areas can foster trust and confidence, creating a smoother and fuller integration into the laboratory setting.

While AI holds immense potential, its purpose is not to replace highly specialized professionals or laboratory technicians but to augment their capabilities, creating positive change in clinical practice. Automation in hematology that allows for the pre-processing and assessment of digital microscopic images could enhance reproducible results and enable hematologists and laboratory professionals to turn their attention and time to difficult diagnoses and outlier cases that deviate from the usual pattern, thus decreasing the overall workload as well as improving overall turnaround time in today’s busy laboratory environment.

Understanding, addressing, and consistently scrutinizing the challenges associated with AI in clinical laboratory practice are crucial to unlocking the transformative potential of this groundbreaking technology and ensuring we keep our eye on the golden promise of heightened diagnostic accuracy, improved patient response times, and treatment outcomes.

References

  • Gedefaw L, Liu CF, Ip RKL, Tse HF, Yeung MHY, Yip SP, Huang CL. Artificial Intelligence-Assisted Diagnostic Cytology and Genomic Testing for Hematologic Disorders. Cells. 2023 Jun 30;12(13):1755. doi: 10.3390/cells12131755.
  • Blum K. A Status Report on AI in Laboratory Medicine. Association for Diagnostics and Laboratory Medicine. 1 January 2023. Available from https://www.myadlm.org/cln/articles/2023/janfeb/a-status-report-on-ai-in-laboratory-medicine?utm_source=cln-email&utm_medium=email&utm_c%E2%80%A6 [Accessed 16 January 2024]
  • Radakovich N, Nagy M, Nazha A. Artificial Intelligence in Hematology: Current Challenges and Opportunities. Curr Hematol Malig Rep. 2020 Jun;15(3):203-210. doi: 10.1007/s11899-020-00575-4. PMID: 32239350.
  • Walter W, Haferlach C, Nadarajah N, Schmidts I, Kühn C, Kern W, Haferlach T. How artificial intelligence might disrupt diagnostics in hematology in the near future. Oncogene. 2021 Jun;40(25):4271-4280. doi: 10.1038/s41388-021-01861-y