Cervical cancer prognosis and diagnosis using electrical impedance spectroscopy

Ping Li; Peter E. Highfield; Zi-Qiang Lang; Darren Kell

doi:10.2478/joeb-2021-0018

.blurhash-client-img { display: none !important; }

Cervical cancer prognosis and diagnosis using electrical impedance spectroscopy

Journal of Electrical Bioimpedance

Volume 12 (2021): Issue 1 (January 2021)

By: Ping Li, Peter E. Highfield, Zi-Qiang Lang and Darren Kell

Open Access

|Dec 2021

Figures & Tables

The ZedScan handset for making the EIS measurements used in this paper. The handset is shown placed on the base.

Comparison between measured and model fitted EIS

Table.1

p-values from MANOVA using EIS data taken from 1704 women for HG CIN detection

Feature combinations	p-values	Feature combinations	p-values
${\bar{R}}_{0}, \bar{α}, Δ R_{0}$	1.1003 × 10⁻³¹	${\bar{R}}_{0}, \bar{α}, Δ R_{0}, Δ f_{c}$	5.0124 × 10⁻³¹
${\bar{R}}_{0}, Δ R_{0}$	1.5861 × 10⁻³¹	${\bar{R}}_{\infty}, {\bar{R}}_{0}, Δ R_{\infty}, Δ R_{0}$	5.3276 × 10⁻³¹
${\bar{R}}_{0}, \bar{α}, Δ R_{\infty}, Δ R_{0}$	2.6287 × 10⁻³¹	${\bar{R}}_{0}, Δ R_{0}, Δ f_{c}$	6.5955 × 10⁻³¹
${\bar{R}}_{0}, Δ R_{\infty}, Δ R_{0}$	3.1665 × 10⁻³¹	${\bar{R}}_{0}, Δ R_{0}, Δ α$	7.2683 × 10⁻³¹
${\bar{R}}_{0}, \bar{α}, Δ R_{0}, Δ α$	3.4687 × 10⁻³¹	${\bar{R}}_{0}, {\bar{f}}_{c}, \bar{α}, Δ R_{0}$	7.4318 × 10⁻³¹

Table.2

p-values from MANOVA using EIS data taken at initial colposcopy of 569 women for evaluation of prognostic value of EIS

Feature combinations	p-values	Feature combinations	p-values
$\bar{α}, Δ α$	0.0168	${\bar{f}}_{c}, Δ α$	0.0286
$\bar{α}, Δ R_{0}$	0.0231	${\bar{R}}_{0}, \bar{α}$	0.0295
${\bar{f}}_{c}, \bar{α}$	0.0256	${\bar{R}}_{\infty}, \bar{α}$	0.0296
$\bar{α}, Δ R_{\infty}$	0.0274	${\bar{f}}_{c}, \bar{α}, Δ α$	0.0314
$\bar{α}, Δ f_{c}$	0.0275	${\bar{R}}_{0}, \bar{α}, Δ α$	0.0335

Table.3

AUC values for testing sets from 10 repeated two-fold cross validation runs with three logistic regression models

Repetitions	$\begin{array}{l} {\bar{R}}_{0}, \bar{α}, Δ R_{0}, \\ CI,Ref \end{array}$	$\begin{array}{l} {\bar{R}}_{0}^{2}, {\bar{α}}^{2}, Δ R_{0}^{2}, \\ CI,Ref \end{array}$	$\begin{array}{l} {\bar{R}}_{0}^{3}, {\bar{α}}^{3}, Δ R_{0}^{3}, \\ CI, Ref \end{array}$
1	0.9127	0.9160	0.9165
2	0.9177	0.9178	0.9190
3	0.9013	0.9034	0.9045
4	0.9165	0.9210	0.9238
5	0.8830	0.8840	0.8858
6	0.9206	0.9222	0.9230
7	0.9164	0.9165	0.9172
8	0.9053	0.9061	0.9075
9	0.9122	0.9146	0.9155
10	0.9181	0.9215	0.9222
Mean AUC	0.9104	0.9123	0.9135

Table.4

Regression coefficient estimates and the associated p-values for the final logistic regression model

β	estimates	p-values
βo	-2.9619	3.9518 × 10⁻³²
β1	−7.4684 × 10⁻¹¹	0.0047
β2	3.3987	0.0090
β1	3.0025 × 10⁻¹¹	0.0044
βCI	2.3621	3.9281 × 10⁻⁴⁷
βRef	2.2241	5.8068 × 10⁻³⁵

ROC comparison between new method, template match method and colposcopy only.

Table.5

Mean AUC values from 100 5-fold cross validation runs with linear logistic regression models

Feature combinations	Mean AUC	Feature combinations	Mean AUC
$\bar{α}, Δ α$	0.5870	${\bar{R}}_{0}, α, Δ α$	0.5723
$\bar{α}, Δ R_{\infty}$	0.5777	${\bar{f}}_{c}, α, Δ α$	0.5716
${\bar{f}}_{c}, \bar{α}$	0.5745	$\bar{α}, Δ R_{0}$	0.5715
$\bar{f_{c}}, Δ α$	0.5744	$\bar{α}, Δ f_{c}$	0.5686
${\bar{f}}_{c}, Δ R_{\infty}, Δ α$	0.5736	${\bar{R}}_{0}, \bar{α}$	0.5678

Table.6

Mean AUC values from 100 5-fold cross validation runs with nonlinear logistic regression models

Feature combinations	Mean AUC	Feature combinations	Mean AUC
${\bar{α}}^{2}, Δ α^{2}$	0.6103	${\bar{f}}_{c}, {\bar{α}}^{2}, Δ α^{2}$	0.5911
$Δ α, {\bar{α}}^{2}$	0.5992	${\bar{R}}_{0}^{2}, α^{2}, Δ α^{2}$	0.5899
$\bar{α}, Δ α^{2}$	0.5989	$Δ R_{\infty}, {\bar{α}}^{2}, Δ α^{2}$	0.5895
${\bar{α}}^{2}, \bar{α} \cdot Δ α, Δ α^{2}$	0.5946	$Δ R_{\infty}^{2}, {\bar{α}}^{2}, Δ α^{2}$	0.5891
$α, Δ α, Δ α^{2}$	0.5939	${\bar{α}}^{2}, Δ f_{c}, Δ α^{2}$	0.5885

2-D histogram of α̅-Δα data points from two groups

An ROC curve of final model for separating two groups with OOP and the associated performance indices

Table.7

Classification performance comparison between the new classifier developed and the previous classifiers

Classifier	AUC	Sensitivity	Specificity
Logistic regression $({\bar{α}}^{2}, Δ α^{2})$	0.628	45.714%	82.022%
Impedance at 152Hz	0.621	38.7%	83.4%
Slope (between 1.22	0.596	45.2%	70.1%
and 2.44kHz) as α

References

Authors

Metrics

Articles in this issue

DOI: https://doi.org/10.2478/joeb-2021-0018 | Journal eISSN: 1891-5469

Journal RSS Feed

Language: English

Page range: 153 - 162

Submitted on: Sep 25, 2021

Published on: Dec 27, 2021

Published by: University of Oslo

In partnership with: Paradigm Publishing Services

Keywords:

Electrical impedance spectroscopy (EIS),

spectrum curve fitting,

logistic regression,

classification

Related subjects:

Engineering,

Bioengineering and biomedical engineering,

Biomedical electronics,

Life sciences,

Biophysics,

Medicine,

Biomedical engineering,

Physics,

Spectroscopy and metrology

© 2021 Ping Li, Peter E. Highfield, Zi-Qiang Lang, Darren Kell, published by University of Oslo
This work is licensed under the Creative Commons Attribution 4.0 License.

Volume 12 (2021): Issue 1 (January 2021)

Cervical cancer prognosis and diagnosis using electrical impedance spectroscopy

Figures & Tables

Fig.1

Fig.2

Table.1

Table.2

Table.3

Table.4

Fig.3

Table.5

Table.6

Fig.4

Fig.5

Table.7

Paradigm

My account