MSACL 2018 US Abstract

Topic: Tissue Imaging & Analysis

An International “iKnife” Network to Validate Tissue-Specific Database Across Multi-Centers

Zsolt Bodai (Presenter)
Imperial College London

Bio: I did my PhD in Hungary, in the field of food analytical chemistry using HPLC-MS/MS. I started working in the United Kingdom at Imperial College London (department of Surgery and Cancer) as a postdoc with Prof. Zoltan Takats at the beginning of September 2015. In 2016, I have been promoted to co-ordinate the the iKnife team`s to do in-vivo, real-time tissue identification during surgery using rapid evaporative ionization mass spectrometry (REIMS). On top of my new role, I also work on different applications (e.g. early detection of cancer, discovering biomarkers from different biological samples, identification of bacteria based on lipidomic profiles, metabolic phenotyping of lung injury etc.) using various mass spectrometric methods (UPLC-MS, direct injection nano-ESI-MS, REIMS MS, etc.).

Authorship: Zsolt Bodai (1), Alvaro Perdones-Montero (1), Emma White (1), Adele Savage (1), Eftychios Manoli (1), Petra Paizs (1), Julia Balog (2), Tiffany Porta (3), Pierre-Maxence Vaysse (3), Martin Kaufmann (4), Seth Chitayat (4), John Rudan (4), Michael Woolman (3), Arash Zarrine-Afsar (5), Steven D. Pringle (2), Ron M.A. Heeren (3), Zoltan Takats (1)
(1) Imperial College London, London, United Kingdom (2) Waters Research Centre, Budapest, Hungary (3) Maastricht University, Maastricht, The Netherlands (4) Queen's University, Kingstone, Canada (5) TECHNA Institute, Toronto, Canada

Short Abstract

Rapid evaporative ionization mass spectrometry (REIMS) has been recently introduced and showed great promises to improve margin assessment in situ, opening up new perspectives in surgery management. REIMS analyses in real time the chemical composition of aerosols produced by electrosurgical devices and provides enough selectivity and specificity for differentiating tissues based on specific molecular signatures. But can other hospitals immediately implement the technology and use the previously built database? What is the inter- hospital and intra- hospital repeatability and reproducibility? In order to answer these questions, an international iKnife network was created. One of the first goals of the network is to monitor and assess reproducibility across different sites, identify pitfalls, and harmonize the protocols for surgical iKnife applications.

Long Abstract


Intraoperative tissue identification has been a long-standing problem in cancer surgery, both in context of margin detection and identification of unknown tissue. While existing intraoperative margin assessment techniques were largely limited to frozen section histology, a number of alternatives have been proposed in course of the last decade, ranging from the fluorescent labelling of cancer cells [1] to the Raman spectroscopic [2] characterisation of surgically exposed tissues. Mass spectrometric methods have also been put forward as a potential solution, such as the recently developed rapid evaporative ionization MS (REIMS) which is utilising on-line chemical analysis of aerosols produced by electrosurgical devices, thereby harnessing the existing role of the electrosurgical tool as a tissue dissection device [3].

REIMS coupled with electrosurgical radiofrequency devices or “diathermy”, have already proved to have high selectivity and specificity in the identification of different type of tissues [3-5] using ex-vivo fingerprint built databases. But can other hospitals immediately implement the technology and use the previously built database? What is the inter- hospital and intra- hospital repeatability and reproducibility? Are any variations in results arising from the biological heterogeneity of the tissue higher than the variation arising from the operator or the technology? Does a site-specific database need to be built for every institution or can we build a single, universal database?

In order to answer these questions, an international iKnife network has been formed between Imperial College London (London, United Kingdom), Maastricht University (Maastricht, The Netherlands), Queens`s University (Kingston, Canada) and is being expanded with TECHNA Institute (Toronto, Canada) with the support of Waters Corporation (Wilmslow, United Kingdom and Budapest, Hungary). One of the first goals of the network is to monitor and assess reproducibility across different sites, identify pitfalls, come up with a consensus that could be apply across different sites, and harmonize the protocols for surgical iKnife applications.


Preliminary experiments were performed by the research group at the Imperial College London; by three research technicians across three different hospitals (Charring Cross, St Mary`s and Hammersmith Hospitals) in order to validate a set of methods, and to prepare standard operating protocols (SOPs) for the preparation, measurement and analysis of data. These initial experiments were performed on calf liver pieces from a single batch, which were sampled at each hospital over the course of 10 days, with 10 individual sampling points each day and at each location. After these initial tests, the SOPs, calf liver and breast cell line (T47D, MDAMB468 and BT474) samples were distributed within the network for the international study.

A surgical diathermy hand-piece with a Covidien ForceTriad generator was used during the sampling. The hand-pieces were equipped with a PTFE tube connected to the mass spectrometer for aerosol evacuation. The captured aerosol was aspirated into custom-built Xevo G2-XS quadrupole – time-of-flight (Q-TOF) MS (Waters, Wilmslow, UK) equipped with a REIMS interface. During the mass spectrometry analysis, 2-propanol was used as a matrix solvent to enhance the ionisation, for lock mass introduction and also to prevent the contamination of the mass spectrometer.

Data was imported to Offline Model Builder (Waters Research Centre, Budapest, Hungary) for lock mass correction and background subtraction. Advanced mass binning (0.1) was used in OMB and the data was exported to R for further statistical analysis. Principal component analysis (PCA), cluster and average class (Pearson) correlation analysis, and intraclass correlation analysis (using single fixed raters) [6] were performed on the data sets to determine if the different setups produced significant differences in the data collected. The average class correlation was done pairwise on each instrument using an average spectrum for each instrument. Average of the correlation coefficient gave the average class correlation. Cluster correlation used an average spectrum for each day.


The PCA score plots did not show any batch effect and the inter-day variation (i.e. on the same instrument with the same operator) was higher than between the instruments. However, some clustering was observed on the score plot based on the user. As the PCA score plot cannot give a metric value to conclude how reproducible the set up is, other statistical methods were also tested.

Average class correlation analysis was performed pairwise on each instrument. The correlation values were between 0.955 and 0.992 (standard deviation 0.014, average 0.967), which suggests a high correlation within the data, and between users and locations. Cluster of the Pearson correlations was also made to visualize the data, and to monitor reproducibility and batch effect between instruments on different days. The dendrogram also showed random branches which indicates no batch effect and high correlation in the data. The average correlation value was 0.958 (standard deviation 0.038, min 0.791, 1st quartile 0.944, 3rd quartile 0.985, max 1.00). Intraclass correlation analysis also showed excellent correlation [7] between the instruments, with a coefficient value at 0.956.

The results of all the tested statistical methods showed that the variation within the instruments and users was low, suggesting good reproducibility of the technology in London.

Conclusions & Discussion

All the tested statistical methods gave high correlation values (average class correlation 0.967, average of the cluster correlation 0.958, and intraclass correlation 0.956), suggesting that the preliminary results on the reproducibility of the surgical REIMS technology (iKnife) is promising, and it might not be necessary to rebuild the database at each hospital where iKnife instruments are used. However, further harmonization and developments may be needed before database building can begin using human tissues. As a next step, human ex-vivo breast samples will be collected and analysed by the international iKnife network participants.

References & Acknowledgements:

1. Gao, X., et al., In vivo cancer targeting and imaging with semiconductor quantum dots. Nat Biotech, 2004. 22(8): p. 969-976.

2. Haka, A.S., et al., In vivo margin assessment during partial mastectomy breast surgery using raman spectroscopy. Cancer Res, 2006. 66(6): p. 3317-22.

3. Schafer, K.C., et al., In vivo, in situ tissue analysis using rapid evaporative ionization mass spectrometry. Angew Chem Int Ed Engl, 2009. 48(44): p. 8240-2.

4. Balog, J., et al., Intraoperative tissue identification using rapid evaporative ionization mass spectrometry. Sci Transl Med, 2013. 5(194): p. 194ra93.

5. Balog, J., et al., Identification of biological tissues by rapid evaporative ionization mass spectrometry. Anal Chem, 2010. 82(17): p. 7343-50.

6. Shrout, P.E. and J.L. Fleiss, Intraclass correlations: uses in assessing rater reliability. Psychological bulletin, 1979. 86(2): p. 420.

7. Cicchetti, D.V., Guidelines, criteria, and rules of thumb for evaluating normed and standardized assessment instruments in psychology. Psychological assessment, 1994. 6(4): p. 284.

Financial Disclosure

Board Memberno

IP Royalty: no

Planning to mention or discuss specific products or technology of the company(ies) listed above: