en, English

Research Papers

The Use of a Novel Optical Algorithm in the Diagnosis of Cervical Pre-Invasive Pathology - A Preliminary Proof of Principal Study.

Gynecol Obstet Invest. 2016;81(6):523-528. Epub 2016 Mar 10.

 2016;81(6):523-528. Epub 2016 Mar 10.

The Use of a Novel Optical Algorithm in the Diagnosis of Cervical Pre-Invasive Pathology - A Preliminary Proof of Principal Study.

Eitan R1Krissi HPeled YBraslavsky DBinyamin LPeretz-Davidi YSeadia OLandesman I.

Author information

1
Gynecologic Oncology Division, Rabin Medical Center, Beilinson Hospital, Petach Tikva, Israel.

Abstract

OBJECTIVE:

To develop a novel optical probe monitoring cervical tissues in real-time and to compare the new imaging technique to actual cervical pathologic findings on resected cone biopsy specimens.

METHODS:

A loop electro-excisional procedure was performed on 15 women with a biopsy diagnosis of dysplasia. The conization specimen was then assessed with the novel optical system and results recorded. The 'normal' and 'abnormal' areas were tested by the optical setup at several points. Extracted parameters were used as the input of the classifier function of a logistic regression algorithm model to assess for system accuracy vs. clinical examination.

RESULTS:

Ninety-seven samples were taken - forty-five samples from 'abnormal zones' and 42 samples from 'normal zones', as defined by the surgeon. The pathologist diagnosed 58 samples as dysplastic and 39 samples as normal. The novel optical method predicted 58 sample points as abnormal and 39 points as normal. The sensitivity of the system was 90% with a specificity of 77%. The probability of correct differentiation of dysplastic cervical tissue from normal cervical tissue was 85%.

CONCLUSIONS:

The optical probe and the algorithms of image processing in combination with the logistic regression algorithm correlated well with pathology results for cervical dysplasia ex-vivo.

© 2016 S. Karger AG, Basel.

PMID:
26960003
DOI:
10.1159/000444584

MeSH terms

MeSH terms

LinkOut - more resources

Full Text Sources

Medical