|
Medical Statistics & A.I.
Publications link
Projects:
Impact
of maternal age and other biosocial factors on the hospital newborn health in Peshawar
N.W.F.P Pakistan
(Sareer
Badshah, Ken McKelvie, Linda Mason, Roger Payne*)
*
IACR-Rothamsted
Objective:
(i) The main objective of this study is to assess birth health i.e., weight,
length, head circumference, and APGAR scores (heart rate, respiratory effort,
muscle tone, reflex irritability, and colour) at the three hospitals in
Peshawar, NWFP-Pakistan. (ii) To examine the effect of maternal age and other
biosocial factors on birth health (weight, height, head circumference and APGAR
score). (iii) This study will provide an opportunity to generate
yardsticks/parameters and will establish a base, for further comparison in
future with national & international studies. (iii) This study will identify
adverse factors affecting birth health, which will help parents, husbands,
doctors, health department and funding agencies in their decisions.
Brain tumour characterisation with Magnetic Resonance Spectroscopy
Y.Y.B. Lee, Pete Harris, W. El-Deredy, P.J.G. Lisboa and A. Vellido
Magnetic Resonance is rapidly becoming the preferred method for the diagnosis of
brain tumours. This usually involves imaging ‘slices’ of tissue, which is done
non-invasively and therefore can be applied in vivo. However, there are instances
where the image is ambiguous concerning the identification of the type, or grade, of
tumour. In these instances, the standard MR systems can be configured to acquire a
biochemical spectrum with good spatial resolution. Studies in collaboration with the Universitat
Autónoma de Barcelona show that the spectra thus obtained can be resolved into
separate sources, which have a clinical interpretation.
 
Sources, identified by
non-linear signal analysis, as the generators of astrocytic tumours. The
signal on the left approximates typical spectra of infiltrating tumours, and the one on
the right resembles spectra from necrotic tissue. Both are clinically relevant
constituents of brain tumours, forming a description with considerable predictive power
for tumour grading with MR spectroscopy.
 
Display of the spectral
components corresponding to infiltrating tumour (C1) and
necrotic tissue (C2), in the model design study (left) and retrospective test of
predictive performance (right). These plots enable clinicians to more easily visualise the
composition of complex, high-dimensional spectra.
Outcome. In a recent
retrospective study, high-grade glioblastomas were correctly separated from medium- and
low-grade astrocytomas, from the spectrum alone, in 8 of 11 patients. This provides
decision support for a difficult, but important, clinical assignment.
Breast cancer survival following surgery
Ian
Jarman, H. Wong, P. Harris,
Ric
Swindell, and Sue O’Reilly
An accurate prognosis immediately following tumour re-section an
essential requirement to provide the patient with informed options regarding adjuvant
treatment. A joint project with Manchester’s Christie Hospital has
developed prognostic risk models which capture time dependencies and complex interactions
between the clinical and laboratory measurements made at the time of surgery.
Predicted (left) vs. Kaplan-Meier (right)
estimates of survival for a low risk cohort of 931 with 5 years follow-up
during1990-2000, using a model fitted to the corresponding 917 patients followed-up during
1983-89. Notice that among these patients, who have no distant metastases as well as T
& N stages of 0 or 1, there is a group with substantial mortality risk during the
first 5 years following surgery.
Histograms of the predictive patient
attributes in each of risk groups whose survival is shown in the previous
figure. These characteristic features are very specific of mortality risk, showing a
developing pattern from low risk (top-left) to high risk (bottom-right).
Outcome. In a
retrospective study, a group of patients with substantially increasing risk over time, was
identified among the clinically defined low risk cohort, pointing to the importance of
detailed statistical analysis in obtaining maximum information from the clinical data.
Funding:
2002-2005:
EPSRC grant GR/R86782/01 ‘Towards a Disciplined Approach to Integrating
Decision Support Systems for Breast Cancer Care Activities’, £188,456.
1997-2000:
EPSRC RS Pool Studentship 97700129 for ‘Statistical
Analysis of Nuclear Magnetic Resonance Spectra for Tumour Detection’, £20,000.
Publications:
-
Huang,
Y., Lisboa, P.J.G. and El-Deredy, W. ‘Tumour grading from Magnetic
Resonance Spectroscopy: a comparison of feature extraction with variable
selection’. Statistics in Medicine 22(1), 2003, pp.147-164.
-
Lisboa,
P.J.G., Wong, H., Harris, P. and Swindell, R. ‘A Bayesian neural network
approach for modelling censored data with an application to prognosis after
surgery for breast cancer’. In press. Artificial Intelligence in
Medicine, 2003.
- Lisboa,
P.J.G., El-Deredy, W., Lee, Y.Y.B., Huang, Y., Corona Hernandez, A.R.,
Harris, P., Huang, Y. and Arús, C. ‘Characterisation of Brain Tissue from
MR Spectra for Tumour Discrimination’ in H. Yan (ed.) Signal Processing
for Magnetic Resonance Imaging and Spectroscopy, Marcel Dekker, New
York, 569-588, 2002.
- Coombes, N.E., Payne, R. and Lisboa, P.J.G.
‘Comparison of nested simulated annealing and reactive tabu search for
efficient experimental designs with correlated data’, COMPSTAT Proc. In
Computational Statistics, (Ed. W. Haerdle & B. Ronz), 249-254.
Heidelberg: Physica-Verlag. Berlin, 24-28 August, 2002.
- P.J.G.
Lisboa, H. Wong, P. Harris and R. Swindell A retrospective study of breast
cancer prognosis using artificial neural networks, in Papadourakis,
G.M. (ed.) Proc. 4th International Conference on Neural Networks
and Expert Systems in Medicine and Healthcare (NNESMED), Milos, Greece, pp.
125-131, 20-22 June, 2001.
-
Lisboa, P.J.G. and Wong, H. Are neural networks best used to help logistic
regression? An example from breast cancer survival analysis.
Proc. International Joint Conference on Neural Networks, Washington.
D.C., paper 577, 4-19 July, 2001.
- Lee, Y.Y.B., Huang, Y., El-Deredy, W., Lisboa,
P.J.G. and Harris, P. ‘Robust methodology for the discrimination of brain
tumours from in vivo Magnetic Resonance spectra’ Proceedings of
MEDSIP, Bristol, pp 88-95, 4-6 September, 2000. Keynote address.
BOOK: "Artificial Neural
networks in Biomedicine"
Lisboa, P.J.G., Ifeachor, E.C. and Szczepaniak, P.S. (eds.) Springer, London, 2000
BOOK: "Fuzzy Systems in Medicine"
Szczepaniak, P.S., Lisboa, P.J.G. and Kacprzyk, J. (eds.) Physica-Verlag, Berlin, 2000
Click the picture below to view
the list of contents (Springer Web site)
|
