Classification of Cervical Samples, Southend University Hospital, UK.

Tumour Trace technology classifies healthy/cancerous cervical smears.

Research Study was conducted by:

Southend University Hospital team: Khalil Razvi MD, Consultant Gynaecology (Medical Director of Study) Dr Anu Malhotra, Consultant Pathology, (Medical Study Advisor) and Sarah Hemingway MD, Consultant BMS-Cytology.

Tumour Trace Ltd team: Prof. Dr Djuro Koruga, Chief Scientist, Aleksandra Dragicevic, MSc, research assistant and Ivan Djuricic, MSc, research assistant) and Ass. Prof. Dr Milena Papic-Obradovic, MD, Gynaecologist:

Period of study:

November 16-19 2015

Location:

Southend University Hospital NHS Foundation Trust, UK

In 1887, to celebrate Queen Victoria’s Jubilee, a public fund was started with the aim of building a hospital, and the site for Southend’s first hospital was bought for £350 (in Warrior Square, near to Southend High Street). Built at a cost of £1,287 4s 6d, Southend Victoria Hospital was opened in May 1888. By Christmas, with eight beds and two cots, it had treated 61 patients at an average weekly cost of 4s 6d (22.5p).

Further building work began in 1930, and the new building was officially opened by the Rt. Hon Earl of Iveagh in 1932. In 1948, when the hospital was taken over by the Ministry of Health under the newly formed NHS, there were 24 consultants and 11 resident medical staff.

Southend Hospital now has 157 consultants providing various services, and serves a catchment area with a population of about 350,000. It has officially been designated cancer centre status, and has also gained Foundation Trust status under the name Southend University Hospital NHS Foundation Trust.

Existing Methods

Cervical cancer is the fourth most common cancer worldwide. The incidence rates are higher in less developed countries, where screening programs are not available to the same extent as in developed countries.

PAP Group Bethesda Classification
Pap II Samples that have no cell abnormalities, reported as negative for intraepithelial lesion or malignancy, and they can also include certain benign findings (infections, inflammations)
Pap IIIa ASC-US/ASC-H/AGUS
Pap IIIb LSIL/HSIL
Pap IV HSIL/AIS
Pap V Invasive carcinoma (Squamous/glandular/other cell type carcinoma)

Existing accuracy based on PAP test: 65.6%

Material

Cervical LBC (Liquid Based Cytology) 132 samples were prepared by standard procedure at Southend University Hospital [1]. Samples are separated in for main groups: Normal (N), Low Grade (LG), High Grade (HG) and Cancer (C) with next numbers of sub-groups for training set (TS) and predicting set (PS):

Normal (N) 23 (TS: 16, PS:7)
Low Grade (LG)– BL&CIN1 46 (TS:32, PS:14)
High Grade (HG)- CIN2&CIN3-cancer in situ 35 (TS: 26, PS:9)
Cancer (C) 28 (TS:18, PS:10)

Total number of TS samples was 92 and PS was 40, what means that TS samples were present with 70%, while PS samples were present with 30%.

Method

An Experiment was conducted using the Tumour Trace device based on optomagnetic imaging spectroscopy, which has sensors for measuring temperature and humidity were the sample is placed. During the experiment for each sample, temperature and humidity was measured by device sensors, and average conditions for all 132 samples were: temperature 21 ± 0.4 0C and humidity 48 ± 0.8. This was conducted as a blind study.

Results

  1. The Tumour Trace algorithm classified Normal and Cancer samples with an accuracy of 95%. Sensitivity was 100%, which means that all sick people were correctly identified as sick, while specificity was 54% which means that 3 normal samples was incorrectly identified as sick.
  2. Classification of Normal and High Grade (CIN2 and CIN3) was with an accuracy of 89%. Sensitivity was 100%, which means that all sick people were correctly identified as sick, while specificity was 48% which means that 4 normal samples was incorrectly identified as sick.
  3. Classification of Normal and Low Grade samples was with an accuracy of 74%. Sensitivity was 96%, which means that 1 sick person was incorrectly identified as normal, while specificity was 62% which means that 4 normal samples was incorrectly identified as sick.

According to results of this initial clinical trial of cervical LBC samples, the Tumour Trace device is excellent for screening, because accuracy of normal/cancer stage is very high (about 95%). Also, there is good classification (with about 89% accuracy) between normal and high grade stage (CIN2&CIN3cancer in situ), what means that cancer in early stage can be identified.

Download a pdf of this case study