For early diagnosis of malignancies in the gastrointestina
l
tract, surveillance endoscopy is increasingly used to moni
tor abnormal
tissue changes in serial examinations of the same patient. D
espite suc-
cesses with optical biopsy for
in vivo
and
in situ
tissue characterisa-
tion, biopsy retargeting for serial examinations is challe
nging because
tissue may change in appearance between examinations. In th
is paper, we
propose an inter-examination retargeting framework for op
tical biopsy,
based on an image descriptor designed for matching between e
ndoscopic
scenes over significant time intervals. Each scene is descri
bed by a hierar-
chy of regional intensity comparisons at various scales, off
ering tolerance
to long-term change in tissue appearance whilst remaining d
iscrimina-
tive. Binary coding is then used to compress the descriptor v
ia a novel
random forests approach, providing fast comparisons in Ham
ming space
and real-time retargeting. Extensive validation conducte
d on 13
in vivo
gastrointestinal videos, collected from six patients, sho
w that our ap-
proach outperforms state-of-the-art methods.