AI-driven analysers have ushered into the pathological space of life sciences, reducing the stress of the pathologists and providing accurate results while strategically reducing the analytical turnaround time by up to 89 per cent. Pankaj Rathod, Founder, Genesis Healthcare explains what one needs to know for achieving best from AI analysers in pathology lab
Pathology labs have seen a dramatic crunch in efficiency owing to the increased demand for diagnostics and the time-consuming manual process of obtaining results from samples. Technology, especially Machine Learning (ML) and Artificial Intelligence (AI), has catapulted in taking over many manual operations in life sciences, hence ironing out human errors and saving the turnaround time (TAT) of the reports, hence enhancing the rate of the patient’s recovery.
Referring to the application of modern ML techniques to digital tissue images in order to identify and provide details on constituents of a specific cell/ tissue, AI has ushered into the pathology space and has only enhanced the way diagnoses are reported in the modern-day. Revolutionizing space, AI allows to identify and characterize special tissue and cell structures and report quickly about sensitive biomarkers to prescribe accurate medicinal procedures.
AI can ably cover several diagnostic tasks in pathology and excel at the results which otherwise can be cumbersome for the pathologists. Be it identifying tumor cells/tissues or many other analytical observations, AI-driven analysers have been excelling in pathological applications.
How does AI aid in analysing samples?
With the precision image-processing capability at higher speeds, AI-driven analysers help in identifying and analysing biomarkers at an early stage during testing, hence dismissing the prolonged time spent by the pathologists on analysing samples individually which may also be subject to analytical errors.
These analysers simplify addressing the hassle of several patients who might require subsequent medical consultations. This demands for their timely pathological reports and AI analysers can do the job quickly without burdening the pathologists.
Also, patients revisiting for pathological diagnosis after a prolonged period regarding a pre-existing disease might have to witness higher turnaround times given the complexity to study the case between the timelines. AI analysers help answer this tussle through their storage, by quickly pulling up the data of the patient’s reports analysed months or years ago.
Furthermore, AI analysers can help reduce TAT substantially as the validation of their generated reports can be remotely done by the examiners/ doctor irrespective of the time of the day or holidays, etc.
Merits of AI-driven analysers
As aforementioned, AI analysers’ application in pathology goes above and beyond to be applied to recognise and report early for precise medicinal procedures. Prospective application of AI analysers includes body serums, semen analysis, bone marrow analysis, etc. It has been possible due to the many unique characteristics that AI analysers put forward:
- Reduced physical fatigue for pathologists
- Efficient and error-proof results
- Faster results compared to manual analysis, leading to reduced TAT of up to 89 per cent.
- Can store data of patients on the cloud for paperless future reference.
- Enhances testing efficiency for sensitive and extreme cases.
Limitations of AI analysers
While AI analysers are efficient enough, some degree of limitations is shown by them too. Reportedly, due to repetitive imaging from different angles of rotation, AI-driven analysers couldn’t provide an accurate report on skin lesions interrupted by non-nevus skin lesions. (Young AT, Fernandez K, Pfau J et al. Stress testing reveals gaps in clinic readiness of image-based diagnostic artificial intelligence models). Furthermore, AI analysers which are dependent on data heavily to provide results while drawing from similar data in the past, face trouble in their application in India. The key challenge is directly related to the patient’s consent for data collection, and then corroborating whether the data is clean and uniform. A discrepancy in this can lead to inaccurate results by the AI analysers.
Yet another limitation lies in the gap that exists between the AI analysers as a prototype and the production-ready units. The focus on investing more in prototypes for efficiency in a bid to receive regulatory approvals as compared to the production-ready units of these analysers, make the divide furthermore distinctive, leading to a decreased dependency on the AI analysers for even routine workflows.
Headroom for better regulation of AI analysers for pathological application
Owing to the challenges faced by pathologists in achieving the highest degree of efficiency from the AI analysers, there are several dots that need to be connected. An effective and just regulatory framework for AI in healthcare will help seekers obtain the devices while knowing about the possible risks of their application. The regulatory system will also ensure spotless practices in the development by vendors while eliminating any biased assessment. The framework will also list down standard practices to verify the system’s performance at sites of application. The framework will also ensure real-time assessment of the performance of the AI analyzers over time of clinical usage and suggest standard protocols to dismiss the performance-related issues that come to the fore.
Putting AI analysers to their best use
Pathologists keen on using AI analysers see it as a universal solution to provide increased efficiency and accurate diagnostics in routine tasks. Pathologists can use them to count elements like tumor cells, inflammatory cells, or pathogens, and to present results that may be flagged as examples for reference. The AI analysers can particularly point the pathologists towards selecting priority cases based on the elements in a slide. As per several studies, AI analysers have been used for multiple applications including obtaining results depicting the presence of cancer cells, counting tissues or cells, tumors, and offsetting workload while being analytically efficient by about 89 per cent in turnaround time.