How technology is enabling the collaborative radiologist

fhi logo
future health index

May 17, 2018 - reading time 6 mins

By Hugh Harvey


Dr Harvey is a board certified radiologist and clinical academic. He trained in the UK’s NHS and has a research MD from Europe’s leading cancer research institute, the Institute of Cancer Research, where he was twice awarded Science Writer of the Year. He has worked at Babylon Health, heading up the regulatory affairs team, gaining a world-first CE marking for an AI-supported triage service, and is now a consultant radiologist, Royal College of Radiologists informatics committee member, and advisor to AI start-up companies, including Kheiron Medical.

By Hugh HarveyClick here to read less

fhi logo mobile

No role in a hospital has evolved as much as that of radiologists over the past three decades.

Radiologist at work

From slow-paced wet film and written notes to the fully digitized, interconnected and volume-driven job of today, the work of the clinical radiologist has changed far faster than any other medical speciality .


With an estimated one billion radiological examinations performed worldwide every year and the increasing use of diagnostic and interventional radiology services, radiologists have been elevated from a supporting role to centre stage in a hospital. The outdated stereotype of a slightly awkward, unapproachable clinician resigned to a dark basement is obsolete and radiologists now hold significant influence over patients’ care pathways.


Against this background, the continued surge in demand for diagnostic and interventional services is providing radiologists with end-to-end involvement in patient care from the point of admission. Take mechanical thrombectomy, for example, which is used for the treatment of acute stroke. While we must acknowledge the invaluable role of specialist physicians here, the stroke is actually diagnosed by a diagnostic radiologist interpreting a CT scan of the patient’s brain, followed by swift treatment of the stroke by an interventional neuroradiologist.


Increasing demand, though, brings new challenges and responsibilities as radiologists become custodians of a greater proportion of a hospital’s resources, especially with regards to managing a labour force and finances. A central role in patient care also demands active participation in quality improvement, patient safety and data governance.

With great power comes great responsibility

Despite this evolving role, diagnostic image interpretation remains at its heart and still makes up the bulk of a radiologist’s workload. Although technology is constantly improving to make image interpretation more sensitive and specific, discrepancies in radiology reporting still occur to the detriment of patients.


Although the exact definition of a radiological error or discrepancy is a debate in itself, there is some agreement that a discrepancy encompasses any radiological interpretation that differs significantly from that of one’s peers. Such is the ambiguity as to what exactly is deemed as a discrepancy in radiology.


As the Royal College of Radiologists in the UK states: “There are no objective benchmarks for acceptable levels of observation, interpretation or ambiguity discrepancies.” The published literature suggests discrepancy rates range from 3%–30% and, based on the estimated 40 million radiological examinations performed every year in the UK, even if we take the lowest of these figures (3%), there are approximately 1.2 million radiological discrepancies every year in that country alone.


Radiology is a unique speciality in which nearly every radiological opinion is available for re-evaluation. Digital imaging archiving systems mean the loss of imaging data over time is virtually impossible and data is retrievable within minutes. It would be fair to say, then, that radiology is one of the few professions where the owner of the job captures all their mistakes in pictures – which is the essence of radiology department discrepancy meetings that are held on a regular basis in hospitals everywhere.

Correcting errors to prevent mistakes

Comparing the diagnostic accuracy of radiologists with that of other specialities is increasingly difficult. Autopsy results suggest that the final working diagnosis may be inaccurate in up to one in five patients, however, it is not possible to quantify how much of the physician’s final working diagnosis is influenced by the radiology report. Since discrepancies in final working diagnoses cannot always be attributable to either the radiologist or the clinician, accountability should be shared by both. A patient-centric approach, which sees radiologists embedding themselves deeply into the clinical team, will help with this.


The Kim-Mansfield radiological error classification system analysed 1,269 radiological errors and split them into 12 categories. The three most common errors were ‘missed findings’ (42%), ‘satisfaction of search’ (22%) and ‘misinterpretation of findings’ (9%).


A much simpler approach is to broadly separate errors into two categories: perceptual (missed findings) and cognitive (misinterpretation of findings). Although discrepancies in both categories are due to a combination of system and human errors, it is clear the radiologist is as dependent on the clinician to provide accurate clinical information as the clinician is on the accuracy of the final radiological report.


There is a well-documented example of procedural error whereby two paediatric radiologists missed a coin lodged within a child’s oesophagus on a chest X-ray despite no clinical history to suggest this. The process of interpreting a study requires a complex thought process, which is tailored to the clinical history and index of suspicion of a particular pathology as a cause for the patient’s symptoms. In the case of the missed coin, had the suspicion of a swallowed foreign body been raised by the clinical team, at least one of the radiologists would have searched for it.


Although perceptual errors account for a larger proportion of errors, cognitive errors might benefit most from better clinical input. More often than not, radiological findings can represent wide differential diagnoses of limited clinical value. In these situations, the initiative is on the radiologist to actively seek additional clinical information to produce a clinically relevant report.


Radiology departments should ensure that strategies are in place to reduce errors where possible, whether through the use of computer-aided detection or the imminent arrival of AI systems to help identify and eliminate avoidable errors. For example, structured or synoptic reporting has been shown to provide clearer reports while also reducing typographic errors.


Despite the increased exposure to discrepancies that come with the work’s volume and complexity, it is important that radiologists do not allow the fear of an error to alter their practice, as this can lead to non-committal and defensive reports that are often of little clinical value. It is in fact the radiologist’s courage of his or her convictions that has added so much clinical value in the past and will allow them to thrive in the future.


This article was co-written by Dr. Shah Islam

Share this article

  • Link copied


You are about to visit a Philips global content page


You are about to visit the Philips USA website.

I understand

Our site can best be viewed with the latest version of Microsoft Edge, Google Chrome or Firefox.