Edward St John (Presenter)
Imperial College
Bio: I live in London, UK and work at Imperial College. I qualified from medical school in 2006 with a merit in surgery and a 1st class BSc in medical sciences with honours. I became a member of the Royal College of Surgeons of England in 2009. I am specialising in breast and general surgery and was appointed to the role of Clinical Research Fellow to Professor the Lord Darzi in April 2014. Professor Zoltan Takats is supervising my PhD research project entitled Rapid Evaporative Ionisation for Examination of Circumferential Surgical Excision Margins in Breast Surgery. As a surgeon I am fascinated by the translation of mass spectrometric analysis to the clinical environment.
Authorship: St John ER (1), Rossi M (1), Mckenzie J (1), Balog J (2) , Gildea L (1), Ramakrishnan R (1), Shousha S (1), Veselkov K (1), Leff DR (1), Darzi A (1), Takats Z (1)
(1) Imperial College, London, UK, (2) Waters, Budapest, Hungary.
| Short Abstract Rapid Evaporative Ionization Mass Spectrometry (REIMS) measures the tissue specific ionic content of the electrosurgical smoke plume for the rapid identification of dissected tissues. Aerosol was aspirated from histologically validated ex-vivo breast samples for mass spectrometric analysis. Multivariate statistics were used for computational analysis of the data. A combination of “Cut” and “Coag” electrosurgical modalities corresponding to 280(cut)/281(coag) normal spectra, 80(cut)/59(coag) tumour spectra gave sensitivities of 92.5%(cut)/ 93.2%(coag) and specificity of 97.9%(cut)/95.7%(coag). A combined Cut & Coag model together with recognition software has enabled the real-time classification of ex-vivo breast tissues with a very high accuracy. The iKnife has been developed for real-time analysis of heterogeneous breast tissue in both cutting and coagulation electrosurgical modalities. |
Long Abstract
Introduction:
Positive tumour margins following attempted breast-conserving surgery (BCS) is an important risk factor for local recurrence. Published re-operation rates for positive margins are between 20-25% in both the United States and United Kingdom. Traditional techniques such as specimen X-ray, frozen section and imprint cytology to optimise margin clearance have significant limitations. Rapid Evaporative Ionisation Mass Spectrometry (REIMS) uses mass spectrometric analysis of the tissue specific ionic content of the surgical diathermy smoke plume for the rapid identification of dissected breast tissues as an intelligent knife (iKnife). Diathermy devices in common operative use provide a choice of cutting (Cut) and coagulation (Coag) modalities to the surgeon. Cut utilizes a low voltage and continuous radiofrequency wave to produce heat rapidly. Coag utilizes a high voltage and pulsed radiofrequency wave to produce less over all heat thereby forming a coagulum, which subsequently reduces bleeding. Our experience is that surgeons prefer to use both modes interchangeably whilst operating on the breast. We investigate the ability of the “iKnife” to rapidly analyze heterogeneous breast tissue in both cut and coagulation mode using mass spectrometric techniques.
Method:
Smoke aerosol produced as a result of “Cut” or “Coag” electrosurgical diathermy from a variety of frozen and fresh ex-vivo breast samples was aspirated into a mass spectrometer via a modified surgical handpiece. Surgical aerosol was combined with isopropyl alcohol before impact with a heated collision surface. The m/z value of negatively charged ions was analysed by a modified Xevo G2-S instrument. Tissue diagnosis was confirmed by subsequent histopathological validation. Spectra were assigned a classification of normal or tumour based upon histopathology. Multivariate statistics –predominantly Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA), along with leave one patient out cross-validation was used for computational analysis and building of an ex-vivo classification model. Recognition software was designed utilizing peak alignment and background subtraction algorithms combined with multivariate statistical models (PCA and LDA). Both cut and coag electrosurgical modalities were used to analyze ex-vivo breast tissues (normal and tumour) and the recognition software was used to determine the classification in real-time. The iKnife classification was compared to permanent section histopathology. The iKnife was used in the operating theatre during breast surgeries. Ethical approval was obtained from the Research Ethics Committee.
Results
40 patients contributed breast samples (normal and cancerous) for method optimisation to enable analysis of high intensity spectra from heterogeneous breast tissue. Following optimisation an ex-vivo database was constructed from two different diathermy modes: “Cut” and “Coag”. 108(Cut)/101(Coag) excised fresh breast tissue samples from 76(Cut)/73(Coag) patients using 360(Cut)/340(Coag) spectra corresponding to 280(Cut)/281(Coag) normal (B1) spectra, 80(Cut)/59(Coag) tumour spectra consisting of in-situ (ductal) and invasive cancerous breast tissues (ductal, lobular, mucinous). Multivariate statistical analysis of data revealed classification of tumour compared to normal tissue with sensitivities of 92.5%(Cut)/ 93.2%(Coag) and specificity of 97.9%(Cut)/95.7%(Coag) in determining neoplastic from normal breast tissue. Overall cancerous tissues exhibit statistically different spectral profiles with a typical predominance of ions in the 600-800m/z range corresponding mainly to phospholipid compounds compared to normal tissue that has a predominance of ions in the 800-1000m/z range corresponding mainly to triglycerides. However gross spectral differences are observed between the Cut and the Coag modes. Spectra in the triglycerides range are visualized with increased abundance using the coagulation modality compared to the cutting modality. The coagulation mode also exhibits more background noise throughout the spectra (m/z 150-1000) than the cutting mode. Real-time recognition software using a combined Cut and Coag model was used to classify 15 (10 normal and 5 tumour) ex-vivo breast tissue samples that were analysed with the iKnife in both cut and coag modes. The results were displayed on screen within 2 seconds. For these samples the recognition software achieved a level of 100% sensitivity and 100% specificity compared to validated histopathological diagnosis. Preliminary intraoperative breast data show that high intensity spectra can be recorded throughout an operation in both Cut and Coag modes.
Conclusions:
The iKnife has been successfully developed for analysis of heterogeneous breast tissue in both cutting and coagulation electrosurgical modalities. Preliminary data suggests that this technique is suitable with high accuracy for the separation of normal and cancerous breast tissues among a variety of diathermy settings. The different diathermy settings result in different spectral patterns. Advances in model building and classification have led to the recent development of recognition software capable of both prospective and retrospective analysis of iKnife datasets. REIMS spectra can now be collected, analysed, classified and displayed within 1-2 seconds. This model suggests that surgeons will not need to modify their technique or usage of electrosurgical modes whilst using the iKnife for surgical excision. We are one step closer to the development of an intraoperative mass spectrometric margin detection tool. Larger numbers are required in the ex-vivo datasets to provide statistically significant diagnostic accuracy data among a variety of tumour types. Prospective intraoperative data is required to see if this recognition model can detect positive margins in real-time during breast surgery.
References & Acknowledgements:
We acknowledge and thank Waters corporation for their help and financial support with this project.
| Description | Y/N | Source |
| Grants | no | |
| Salary | no | |
| Board Member | no | |
| Stock | no | |
| Expenses | no |
IP Royalty: no
| Planning to mention or discuss specific products or technology of the company(ies) listed above: | no |