A Survey on Medical Report Generation

PhD Qualifying Examination


Title: "A Survey on Medical Report Generation"

by

Mr. Haibo JIN


Abstract:

Medical image examination and its interpretation plays an important role in 
discovering diseases and abnormalities from the patients. However, manually 
interpreting each examination into a report is both time consuming and 
challenging. It has been shown that a radiologist spends around 10 minutes to 
finish a report on average. On the other hand, the ever increasing demand of 
image examinations far exceeds the increase of radiologists, which further 
brings burdens to radiologists. Due to the complexity of anatomy and the subtle 
variations in lesion areas, writing medical report is also a challenging task 
for radiologists. It has been reported that radiologists often show high inter- 
observer variability in their interpretations and even skilled radiologists can 
miss about 30% abnormalities in their works. To this end, automatic report 
generation has attracted more and more attentions from the researchers as it 
has the potential to largely relieve the burden of radiologists. In this 
survey, we present a comprehensive review of the recent researches on medical 
report generation. We present the existing methods via the taxonomy we build 
and summarize the strategies for improving model performance into five 
categories. Finally, we discuss the primary challenges of this field and point 
out promising future directions.


Date:                   Tuesday, 28 May 2024

Time:                   2:00pm - 4:00pm

Venue:                  Room 2128A
                        Lift 19

Committee Members:      Dr. Hao Chen (Supervisor)
                        Dr. Dan Xu (Chairperson)
                        Dr. Long Chen
                        Dr. Qifeng Chen