top of page

AI and diagnostics - an evolving relationship

Artificial Intelligence (AI) has gained the general public’s attention and interest back in early November 2021, when ChatGPT was released. ChatGPT is an AI and machine learning-based free chatbot capable of generating human-like text based on the input it receives. But what about AI in healthcare? When has this applied technology entered the medical world and for what purposes?

Within the medical field, AI has seen enormous application in diagnostics. Back in 1972, Stanford University in California witnessed the design of an early AI program, called MYCIN (1). Its purpose was to diagnose and recommend a course of treatment for blood infections. To provide its service, MYCIN relied on the input of various data: symptoms, medical tests, and further information about the patient. MYCN was even capable of providing a logical reasoning behind the diagnosis and treatment recommendation. However, MYCN never saw the clinic due to the limited computational power and the challenging integration of the system into the existing clinical workflow (1).

Today, various AI-based diagnostic systems have reached the clinic, after their passing rigorous evaluation, validation and regulatory approvals. Most of them have implementation in medical imaging diagnostics.

Indeed, the potential of AI in the field is enormous; the human eye can miss undetectable lesions due to their appearance and characterization, while real-time AI for computer aided-detection and computer-aided diagnosis can be instructed to automatically extrapolate complex microimaging structures and identify quantitative pixel-level features to differentiate between neoplastic and non-neoplastic lesions.

A study conducted by Luo et al., back in 2019, aimed to investigate the potential clinical applicability and reliability of artificial intelligence (AI)-assisted endoscopic diagnosis (2).

• Endoscopic images with standard white light were used to develop and validate the Gastrointestinal Artificial Intelligence Diagnostic System (GRAIDS), which demonstrated high sensitivity and diagnostic accuracy, comparable to that of expert endoscopists and more sensitive than non-experts (2).

• Therefore, GRAIDS has the potential to improve efficiency of upper gastrointestinal diagnosis and screening, especially in support of non-expert endoscopists from primary basic or low-volume hospitals (2).

• At the moment, GRAIDS is being routinely utilized at Sun Yat-sen University Cancer Center, China, in the endoscopic clinical workflow and its screening center (2).

This is an example of a well-established, AI-assisted diagnostic system. To be more analytical about this technology-guided advancement, however, it is important to evaluate how healthcare professionals welcomed this advancement and whether they needed any training to integrate it into their clinical workflow.

According to the multi-centre, case-control, diagnostic study about GRAIDS, its diagnostic accuracy had sensitivity comparable to that of expert endoscopists and it was superior to that of non-expert

endoscopists. The paper concludes explaining that GRAIDS could be a precious and helpful tool in assisting community-based hospitals, where staff might be reduced or not of an expert level.

Therefore, is AI a helpful tool in diagnostics and is it going to remain such, or is it going to significantly assist, or eventually, replace, doctors of various specialties? Only time can tell, but hopefully a balanced integration between the two will improve diagnostic techniques for the benefit of patients, enabling early and accurate disease detection.

(1) Encyclopaedia Britannica (2018) [accessed on 10/06/2023].

(2) Luo, Huiyan et al. “Real-time artificial intelligence for detection of upper gastrointestinal cancer by endoscopy: a multicentre, case- control, diagnostic study.” The Lancet. Oncology vol. 20,12 (2019): 1645-1654.

30 views0 comments
bottom of page