Medical school is not for the faint of heart, and just as in other professions, passing an exam to qualify for licensure is a dreaded ritual. The Royal College of Radiologists is not a “college” the way we normally think of such an institution. It is a society chartered to promote the science and clinical practice of oncology and radiology in the United Kingdom. It has over 11,000 members, most of whom are Fellows.
Attaining the title Fellow of the Royal College of Radiologists (FRCR) is the pinnacle of RCR membership. In order to earn it, a member radiologist must go through four Special Training years and pass two grueling exams, one at the end of year 3 and the other at the end of year 4. According to Wikipedia, these exams ensure “…a high quality and standard of radiology consultants. It has been deemed as one of the hardest examinations in the medical profession…”[i]. We can safely assume that, under such challenging education and supervision, FRCRs achieve a very high level of clinical and diagnostic knowledge as evidenced by passing the exams.
Diagnosing lung cancer helped by artificial intelligence
The longstanding method for diagnosing lung cancer has been the chest x-ray. Compared to CT scans, x-ray is a relatively inexpensive technology. It is widely available, and considered safe with less than a 1 in a million chance of later developing cancer from radiation exposure. However, chest x-rays showing abnormal shadows can be difficult to interpret.
A research team from three British hospital centers developed and tested an artificial intelligence (AI) algorithm to identify tumors in the lung.[[ii] The algorithm was trained on a dataset of 400 chest x-rays, including 200 difficult lung cancer cases, and submitted to AI analysis to flag positive cases before sending them to standard reading by three FRCRs. The objective was to explore the role AI could play as a first reader of chest x-rays in order to increase accurate and efficient lung cancer diagnosis.
The FRCRs examined the 400 images and labeled suspected tumors, and their findings were compared to and then combined with what the AI algorithm had identified. It’s interesting to note that as a stand-alone, AI findings were equivalent to the average FRCR findings. Keep in mind that the interpretations by FRCR are authoritative, as evidenced by the two cumulative exams they have passed. So, the AI analysis performed very well.
The performance of the AI algorithm did not disappoint. According to the authors,
The best overall performances were achieved when AI was combined with radiologists, with an average reduction of missed cancers of 60%. Combination with AI also standardised the performance of radiologists. The greatest improvements were observed when common sources of errors were present, such as distracting findings.
Reducing errors and improving the reporting performance of radiologists who interpret chest x-rays is good news for lung cancer patients. When AI is used as a first reader, it offers accurate labeling of suspicious tumors that might have been missed. Its impartiality stems from the fact that it is affected by biases or field conditions that make the subjective interpretation by one radiologist differ from another’s.
Given that lung cancer is the leading cause of cancer death, early diagnosis offers a key to survival. Putting AI to work in the service of reading chest x-rays is a deep breath of fresh air for overworked radiologists.
NOTE: This content is solely for purposes of information and does not substitute for diagnostic or medical advice. Talk to your doctor if you are experiencing pelvic pain, or have any other health concerns or questions of a personal medical nature.
[ii] Tam MDBS, Dyer T, Dissez G, Morgan TN et al. Augmenting lung cancer diagnosis on chest radiographs: positioning artificial intelligence to improve radiologist performance. Clin Radiol. 2021 May 11;S0009-9260(21)00237-3.