Pattern Classification Techniques for Lung Cancer Diagnosis by an Electronic Nose - Computational Intelligence in Healthcare: Advanced Methodologies - page 406

subjects (and thus early diagnosis) is possible and that, with a larger and more
balanced dataset, the trained model might achieve even higher performance.
11 Conclusions
A brief overview of pattern analysis algorithms used in the electronic nose field
has been provided, focusing on a specific case study of the use of an electronic nose
based on an array of MOS sensors to detect lung cancer. Many research directions
that are the frontier of sensor based machine olfaction must still be explored, such
as the development of hybrid arrays and the implementation of ensemble meth-
ods where the outputs of different predictors are combined to produce an overall
output. Another cutting edge issue is the development of a transforming mapping
among different devices or sensors in order to transfer the predicted model from
one device to another. Moreover, converting these potential markets to commercial
reality has yet to be achieved: many issues such as requirement for robustness (in
particular when dealing with health of patients), variability of samples, the large
number of environmental and habitual factors that can affect the measurement,
just to mention some, must be faced and solved. Enormous potential exists for the
use of the electronic nose technology in a wide range of applications and the medical
one, as cancer detection, plays a crucial role among them. A very interesting future
development regards the detection of other type of cancers with the electronic nose.
Concerning with lung cancer diagnosis, the achieved results demonstrate that an
instrument as the electronic nose, combined with the appropriate computational
intelligence methodologies, is a promising alternative to current lung cancer diag-
nostic techniques: the obtained predictive errors are lower than those achieved by
present diagnostic methods, and the cost of the analysis, both in money, time and
resources, is lower. Moreover, the instrument is completely non invasive. Current
diagnostic techniques (e.g., CT, PET) are invasive, very expensive, have radiation
risks and a not so good accuracy; moreover, results on early detection and treat-
ment, in last decades, have been very poor and unsatisfying. This calls for the ne-
cessity of a non invasive, accurate and cheap diagnostic technique, able to identify
the presence of lung cancer in its early stages, as the electronic nose demonstrated
to be. Early detection is a major challenge to decrease lung cancer mortality: re-
sults achieved in the early detection analysis are very promising and suggest fur-
ther experimentations and analysis. The introduction of this cutting edge technol-
ogy will lead to very important social and business effects: its low price and small
dimensions allow a large scale distribution, giving the opportunity to perform non
invasive, cheap, quick, and massive early diagnosis and screening.
References
1. Blatt, R., Bonarini, A., Calabro, E., Della Torre, M., Matteucci, M., Pastorino, U.:
Pattern Classification Techniques for Early Lung Cancer Diagnosis using an Elec-
tronic Nose, Frontiers in Artificial Intelligence and Applications. In: 18th European
Conference on Artificial Intelligence (ECAI 2008) - 5th Prestigious Applications of
Intelligent Systems (PAIS 2008), vol. 178, pp. 693-697 (2008)