Evaluation of an automated bacterial growth system.
Think of an existing example where Artificial Intelligence (AI), or any of its branches such as machine learning, is a useful tool in healthcare. Dermatology, robotic caregivers, or preventive medicine for example (CloudMedX, DeepMind Health iCarbonX…) . It is evident that the use of AI in such fields is a useful tool, providing real advantages in speed and quality in the diagnostics.
As mentioned in our previous blog post The new utility in laboratory automation, a smart integration of Artificial Intelligence in the microbiology laboratories is key to consolidate its automation and fulfil the “new utility” in these labs.
However, its suitability may differ between cases. To help you gauge the suitability of an automated dishes growth AI system in microbiology, note these following relevant 5 questions that should be considered when conducting the assessment, to dictate whether AI is a suitable solution to the identified problem.
Q1 – Does Artificial Intelligence in microbiology laboratories help professionals to focus mainly on more complex and added-value activities?
Clinical microbiology laboratories are facing increased demand for shortening and improving results.
It has been dramatically proven during the pandemic. Nowadays, the increased urgency for testing is exhausting laboratories, with many of them pushed to their limit. The bottleneck has shown to be in specialized laboratory professionals.
· It is possible to manufacture and install a diagnostic instrument in a few months but not to train microbiology experts.
· There is a strong need to automate repetitive and non-added values tasks, such as removing negative cultures from the workflow.
On the other hand, it is essential to allow the professionals to concentrate on the positive and more complex samples, where their accurate intervention in the diagnosis has a direct impact on the patient.
The integration of AI as a complementary tool will improve both speed and quality of the diagnosis, that is the core goal of any automation in the laboratory.
An interesting comparison can be read in this sense in the following article (This Is How A.I. Will Transform Medicine: The Same Way It Has Transformed Chess (linkedin.com)
Q2 – Does Artificial Intelligence in Microbiology laboratories improve human performance when interpreting images?
Traditionally, assessment of growth in cultures is performed optically by following semi-fixed protocols (both in clinical and industrial applications).
However, although automation instruments exist, there is an important path for improvement by implementing image systems with higher sensitivity. Consequently, high-resolution image analysis systems can detect small and mixed colonies, which a human eye cannot.
Quality is also a key factor that can be improved by AI, assuring a higher repeatability and uniformity of the processes.
Q3 – Does the AI system improve access to image analysis?
Digital microbiology and tele bacteriology are clear trends in the laboratories. Having access to all sort of digital images clearly improves the workflow traceability, increases the possibility to interconnect the results with other steps in the workflow, allows remote-microbiology assessment in remote geographies and opens new clinical opportunities.
A smart combination of AI (or any of its branches) with digital microbiology will become a key topic to consolidate laboratories automation.
Q4 – An AI simple model is enough, or should we consider a more complex AI model?
As remarked throughout this document, there are different branches of AI (machine learning, cognitive services, chatbots…) which provides different levels of complexity and outputs. In addition, the laboratory application AI needs to be used for, is very diverse (clinical/industrial, colony counting, growth/no growth discrimination, colony identification, diversity of consumables, …).
All the above topics (in combination with the traditional economical KPI) need to be considered during the assessment.
Although the use of basic models will be enough for mature and easy detection applications, there are a lot of advantages that can be achieved by using complex systems, such as having a real-time growth sequencing report, different light sources images… to, in the end, achieve a better identification and shorter diagnosis time.
Q5 – Which are the main challenges faced during an Artificial Intelligence system implementation?
As any other implementation process, implementing AI usually faces several concerns. Some of the most known are fear from the technicians or a poor analysis of its integration in the workflow are factors that could drive the application out of routine and in consequence to the failure.
The technician must be the first system ambassador.
The worst scenarios are false negatives, and the application needs to focus on avoiding them, giving confidence to the technician.
An automated AI system requires a relevant investment and without an excellent integration on the laboratory workflow the investment payback will be difficult to achieve. It is important to achieve workflow optimization with the application.
Would you like to know more about Artificial Intelligence in microbiology laboratories? Contact us >
