Mammogram reading assisted by computer is effective


Mammogram reading assisted by computer is effective


UK researchers found that the rate of detection of breast cancer by two experts reading a mammogram was the same as one expert using computer aided detection, but there was a slightly but significantly higher recall rate in the computer assisted method.

The study was the work of lead author Professor Fiona Gilbert of the University of Aberdeen and colleagues from other research centres in the UK and is published online in the 2 October issue of the New England Journal of Medicine.

Research had already shown that two trained readers examining a mammogram are better at detecting small breast cancers than one trained reader. But the small amount of evidence that exists on how effective it might be if the single reader were aided by computer is conflicting.

In the UK normal practice is for two trained readers to examine mammograms, but in many other countries in Europe and in the USA, only one expert is used.

Gilbert said the findings have "huge international significance". Using computer aided detection (CAD) is "likely to improve breast cancer detection in those countries where only a single reader is used," she added.

In the UK, this means better use of human resources, since the same number of experts, aided by computer, will be able to read more mammograms in the same space of time without losing screening effectiveness.

"We want to offer screening to a wider age group, those 47-50 years old and 70-73 years old, an increase of around 30 per cent. The national programme screens over 1.7 million each year. This computer technology will help us achieve this," explained Gilbert.

For the study, Gilbert and colleagues randomly assigned 31,057 women who were having routine mammogram tests at three centres in England to one of three groups: reading by two experts (double reading); reading by one expert aided by computer (single reading with CAD); or both double reading and single reading with CAD.

The primary outcome measures were the proportion of cancers detected by the two reading methods and the recall rate. The researchers then analyzed the results in a matched-pair comparison.

The results showed that:

  • The proportion of cancers detected was 199 of 227 (87.7 per cent) for double reading and 198 of 227 (87.2 per cent) for single reading with CAD (P=0.89).
  • The overall recall rates (where women are recalled for assessment) were 3.4 per cent for double reading and 3.9 per cent for single reading with CAD.
  • The difference between the recall rates was small but significant (P

  • The estimated sensitivity, specificity, and positive predictive value for single reading with CAD were 87.2 per cent, 96.9 per cent, and 18.0 per cent, respectively.
  • For double reading, these estimates were were 87.7 per cent, 97.4 per cent, and 21.1 per cent.
  • The tumors detected by single reading with CAD and double reading were not significantly different in terms of pathology.
The authors concluded that:

"Single reading with computer-aided detection could be an alternative to double reading and could improve the rate of detection of cancer from screening mammograms read by a single reader."

The findings of the study are to be presented at the National Cancer Research Institute's annual conference in Birmingham, UK, on Monday 6th October.

Dr Stephen Duffy, professor of cancer screening at Cancer Research UK, said that earlier studies had come up with conflicting results on CAD, but this larger study means that:

"We can now say for certain that this system is as good at detecting breast cancer as the one used as standard practice."

He said by bringing in computers to help them, radiologists and trained technicians, who currently read all breast X-rays, will find their workload reduced.

"In some areas women do not get their screening invitations as regularly as they should - every three years - because there are simply not enough experts to go round," said Duffy.

"The CAD system would free up radiologists to work on more mammograms as only one instead of two would be required to read each X-ray," he explained.

Dr Lesley Walker, director of information at Cancer Research UK welcomed the study and said it was "good news for women -- particularly for those who live in areas where invitations for screening have been late in arriving".

Computers are not about to take the place of humans though, as Walker explained:

"We will always need the human eye of an expert to read mammograms. In the rare instance when the computer is at odds with the radiologist the human interpretation takes precedence."

"Single Reading with Computer-Aided Detection for Screening Mammography."

Gilbert, Fiona J., Astley, Susan M., Gillan, Maureen G.C., Agbaje, Olorunsola F., Wallis, Matthew G., James, Jonathan, Boggis, Caroline R.M., Duffy, Stephen W., the CADET II Group.

N Engl J Med, Published online October 1, 2008.

DOI: 10.1056/NEJMoa080354

Click here for Article.

Source: University of Aberdeen, NEJM.


The incredible inventions of intuitive AI | Maurice Conti (Video Medical And Professional 2018).

Section Issues On Medicine: Women health