Diagnostic accuracy of intraoperative margin assessment techniques in surgery for head and neck squamous cell carcinoma: A meta-analysis

BACKGROUND

Positive margins following head and neck squamous cell carcinoma (HNSCC) surgery lead to significant morbidity and mortality. Existing Intraoperative Margin Assessment (IMA) techniques are not widely used due to limitations in sampling technique, time constraints and resource requirements. We performed a meta-analysis of the diagnostic performance of existing IMA techniques in HNSCC, providing a benchmark against which emerging techniques may be judged.

METHODS

The study was conducted according to Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guidelines. Studies were included if they reported diagnostic metrics of techniques used during HNSCC surgery, compared with permanent histopathology. Screening, manuscript review and data extraction was performed by multiple independent observers. Pooled sensitivity and specificity were estimated using the bivariate random effects model.

RESULTS

From an initial 2344 references, 35 studies were included for meta-analysis. Sensitivity (Sens), specificity (Spec), diagnostic odds ratio (DOR) and area under the receiver operating characteristic curve (AUROC) were calculated for each group (n, Sens, Spec, DOR, AUROC): frozen section = 13, 0.798, 0.991, 309.8, 0.976; tumour-targeted fluorescence (TTF) = 5, 0.957, 0.827, 66.4, 0.944; optical techniques = 10, 0.919, 0.855, 58.9, 0.925; touch imprint cytology = 3, 0.925, 0.988, 51.1, 0.919; topical staining = 4, 0.918, 0.759, 16.4, 0.833.

CONCLUSIONS

Frozen section and TTF had the best diagnostic performance. Frozen section is limited by sampling error. TTF shows promise but involves administration of a systemic agent. Neither is currently in widespread clinical use. Emerging techniques must demonstrate competitive diagnostic accuracy whilst allowing rapid, reliable, cost-effective results.

Uncertainty estimation for margin detection in cancer surgery using mass spectrometry

PURPOSE

Rapid evaporative ionization mass spectrometry (REIMS) is an emerging technology for clinical margin detection. Deployment of REIMS depends on construction of reliable deep learning models that can categorize tissue according to its metabolomic signature. Challenges associated with developing these models include the presence of noise during data acquisition and the variance in tissue signatures between patients. In this study, we propose integration of uncertainty estimation in deep models to factor predictive confidence into margin detection in cancer surgery.

METHODS

iKnife is used to collect 693 spectra of cancer and healthy samples acquired from 91 patients during basal cell carcinoma resection. A Bayesian neural network and two baseline models are trained on these data to perform classification as well as uncertainty estimation. The samples with high estimated uncertainty are then removed, and new models are trained using the clean data. The performance of proposed and baseline models, with different ratios of filtered data, is then compared.

RESULTS

The data filtering does not improve the performance of the baseline models as they cannot provide reliable estimations of uncertainty. In comparison, the proposed model demonstrates a statistically significant improvement in average balanced accuracy (75.2%), sensitivity (74.1%) and AUC (82.1%) after removing uncertain training samples. We also demonstrate that if highly uncertain samples are predicted and removed from the test data, sensitivity further improves to 88.2%.

CONCLUSIONS

This is the first study that applies uncertainty estimation to inform model training and deployment for tissue recognition in cancer surgery. Uncertainty estimation is leveraged in two ways: by factoring a measure of input noise in training the models and by including predictive confidence in reporting the outputs. We empirically show that considering uncertainty for model development can help improve the overall accuracy of a margin detection system using REIMS.

Exploring the potential of rapid evaporative ionization mass spectrometry (Intelligent Knife) for point-of-care testing in aortic surgery

 

OBJECTIVES

Many intraoperative decisions regarding the extent of thoracic aortic surgery are subjective and are based on the appearance of the aorta, perceived surgical risks and likelihood of early recurrent disease. Our objective in this work was to carry out a cross-sectional study to demonstrate that rapid evaporative ionization mass spectrometry (REIMS) of electrosurgical aerosol is able to empirically discriminate ex vivo aneurysmal human thoracic aorta from normal aorta, thus providing supportive evidence for the development of the technique as a point-of-care test guiding intraoperative surgical decision-making.

METHODS

Human aortic tissue was obtained from patients undergoing surgery for thoracic aortic aneurysms (n = 44). Normal aorta was obtained from a mixture of post-mortem and punch biopsies from patients undergoing coronary surgery (n = 13). Monopolar electrocautery was applied to samples and surgical aerosol aspirated and analysed by REIMS to produce mass spectral data.

RESULTS

Models generated from REIMS data can discriminate aneurysmal from normal aorta with accuracy and precision of 88.7% and 85.1%, respectively. In addition, further analysis investigating aneurysmal tissue from patients with bicuspid and tricuspid aortic valves was discriminated from normal tissue and each other with accuracies and precision of 93.5% and 91.4% for control, 83.8% and 76.7% for bicuspid aortic valve and 89.3% and 86.0% for tricuspid aortic valve, respectively.

CONCLUSIONS

Analysis of electrosurgical aerosol from ex vivo aortic tissue using REIMS allowed us to discriminate aneurysmal from normal aorta, supporting its development as a point-of-care test (Intelligent Knife) for guiding surgical intraoperative decision-making.

Real-time authentication of animal species origin of leather products using rapid evaporative ionization mass spectrometry and chemometric analysis

Increasing accounts of fraud and persistent labeling problems have brought the authenticity of leather products into question. In this study, we developed an extremely simplified workflow for real-time, in situ, and unambiguous authentication of leather samples using rapid evaporative ionization mass spectrometry (REIMS) coupled with an electric soldering iron. Initially, authentic leather samples from cattle, sheep, pig, deer, ostrich, crocodile, and snake were used to create a chemometric model based on principal component analysis and linear discriminant analysis algorithms. The validated multivariate statistical model was then used to analyze and generate live classifications of commercial leather samples. In addition to REIMS analysis, the microstructures of leathers were characterized by scanning electron microscopy to provide complementary information. The current study is expected to provide a high-throughput tool with superior efficiency and accuracy for authenticating the identity of leathers and other consumer products of biogenic origin.

Perioperative margin detection in basal cell carcinoma using a deep learning framework: a feasibility study

PURPOSE

Basal cell carcinoma (BCC) is the most commonly diagnosed cancer and the number of diagnosis is growing worldwide due to increased exposure to solar radiation and the aging population. Reduction of positive margin rates when removing BCC leads to fewer revision surgeries and consequently lower health care costs, improved cosmetic outcomes and better patient care. In this study, we propose the first use of a perioperative mass spectrometry technology (iKnife) along with a deep learning framework for detection of BCC signatures from tissue burns.

METHODS

Resected surgical specimen were collected and inspected by a pathologist. With their guidance, data were collected by burning regions of the specimen labeled as BCC or normal, with the iKnife. Data included 190 scans of which 127 were normal and 63 were BCC. A data augmentation approach was proposed by modifying the location and intensity of the peaks of the original spectra, through noise addition in the time and frequency domains. A symmetric autoencoder was built by simultaneously optimizing the spectral reconstruction error and the classification accuracy. Using t-SNE, the latent space was visualized.

RESULTS

The autoencoder achieved an accuracy (standard deviation) of 96.62 (1.35%) when classifying BCC and normal scans, a statistically significant improvement over the baseline state-of-the-art approach used in the literature. The t-SNE plot of the latent space distinctly showed the separability between BCC and normal data, not visible with the original data. Augmented data resulted in significant improvements to the classification accuracy of the baseline model.

CONCLUSION

We demonstrate the utility of a deep learning framework applied to mass spectrometry data for surgical margin detection. We apply the proposed framework to an application with light surgical overhead and high incidence, the removal of BCC. The learnt models can accurately separate BCC from normal tissue.

Rapid evaporative ionization mass spectrometry coupled with an electrosurgical knife for the rapid identification of Mediterranean Sea species

The topic of food analysis and safety has attracted increasing interest in recent decades owing to recent scandals concerning fraudulent activities (mislabeling, sophistication, adulteration, etc.) that can undermine human health. Among them, seafood fraud has probably the strongest relationship with food safety, an activity that goes beyond economic interests. This article explores the capabilities of an innovative instrumental setup, called the "iKnife," as a powerful tool in this specific research area, where until now genomics and proteomics have been the workhorses in analytical approaches. iKnife, which means "intelligent knife," is the name of a recent technology based on rapid evaporative ionization mass spectrometry (REIMS). REIMS is an emerging technique able to characterize different samples rapidly, affording a comprehensive profile usable as a fingerprint, without the need for preliminary extraction or cleanup procedures. In detail, a REIMS source is coupled to a high-resolution tandem mass spectrometer; such coupling allows one to maximize the amount of information (discriminant features) collected for a single analysis, as well as to focus on target analytes to achieve enhanced sensitivity and selectivity. A database was created from 18 marine species typical of the Mediterranean Sea, all caught in the very small area of the Strait of Messina, and reliable identification was achieved for each species with confidence higher than 99%. One big model and three submodels were built by principal component analysis and linear discriminant analysis for unambiguous key variable identification within each class (e.g., Cephalopoda), order (e.g., Perciformes), or family (e.g., Carangidae). Graphical abstract.

Effect of Electrode Geometry on the Classification Performance of Rapid Evaporative Ionization Mass Spectrometric (REIMS) Bacterial Identification

The recently developed automated, high-throughput monopolar REIMS platform is suited for the identification of clinically important microorganisms. Although already comparable to the previously reported bipolar forceps method, optimization of the geometry of monopolar electrodes, at the heart of the system, holds the most scope for further improvements to be made. For this, sharp tip and round shaped electrodes were optimized to maximize species-level classification accuracy. Following optimization of the distance between the sample contact point and tube inlet with the sharp tip electrodes, the overall cross-validation accuracy improved from 77% to 93% in negative and from 33% to 63% in positive ion detection modes, compared with the original 4 mm distance electrode. As an alternative geometry, round tube shaped electrodes were developed. Geometry optimization of these included hole size, number, and position, which were also required to prevent plate pick-up due to vacuum formation. Additional features, namely a metal "X"-shaped insert and a pin in the middle were included to increase the contact surface with a microbial biomass to maximize aerosol production. Following optimization, cross-validation scores showed improvement in classification accuracy from 77% to 93% in negative and from 33% to 91% in positive ion detection modes. Supervised models were also built, and after the leave 20% out cross-validation, the overall classification accuracy was 98.5% in negative and 99% in positive ion detection modes. This suggests that the new generation of monopolar REIMS electrodes could provide substantially improved species level identification accuracies in both polarity detection modes. Graphical abstract.

Ambient Mass Spectrometry in Cancer Research

Ambient ionization mass spectrometry was developed as a sample preparation-free alternative to traditional MS-based workflows. Desorption electrospray ionization (DESI)-MS methods were demonstrated to allow the direct analysis of a broad range of samples including unaltered biological tissue specimens. In contrast to this advantageous feature, nowadays DESI-MS is almost exclusively used for sample preparation intensive mass spectrometric imaging (MSI) in the area of cancer research. As an alternative to MALDI, DESI-MSI offers matrix deposition-free experiment with improved signal in the lower (<500m/z) range. DESI-MSI enables the spatial mapping of tumor metabolism and has been broadly demonstrated to offer an alternative to frozen section histology for intraoperative tissue identification and surgical margin assessment. Rapid evaporative ionization mass spectrometry (REIMS) was developed exclusively for the latter purpose by the direct combination of electrosurgical devices and mass spectrometry. In case of the REIMS technology, aerosol particles produced by electrosurgical dissection are subjected to MS analysis, providing spectral information on the structural lipid composition of tissues. REIMS technology was demonstrated to give real-time information on the histological nature of tissues being dissected, deeming it an ideal tool for intraoperative tissue identification including surgical margin control. More recently, the method has also been used for the rapid lipidomic phenotyping of cancer cell lines as it was demonstrated in case of the NCI-60 cell line collection.

A novel methodology for in vivo endoscopic phenotyping of colorectal cancer based on real-time analysis of the mucosal lipidome: a prospective observational study of the iKnife. Surg Endosc. 2017;31:1361-1370. [PMID: 27501728 PMCID: PMC5315709 DOI: 10.1007/s00464-016-5121-5]

BACKGROUND

This pilot study assessed the diagnostic accuracy of rapid evaporative ionization mass spectrometry (REIMS) in colorectal cancer (CRC) and colonic adenomas.

METHODS

Patients undergoing elective surgical resection for CRC were recruited at St. Mary's Hospital London and The Royal Marsden Hospital, UK. Ex vivo analysis was performed using a standard electrosurgery handpiece with aspiration of the electrosurgical aerosol to a Xevo G2-S iKnife QTof mass spectrometer (Waters Corporation). Histological examination was performed for validation purposes. Multivariate analysis was performed using principal component analysis and linear discriminant analysis in Matlab 2015a (Mathworks, Natick, MA). A modified REIMS endoscopic snare was developed (Medwork) and used prospectively in five patients to assess its feasibility during hot snare polypectomy.

RESULTS

Twenty-eight patients were recruited (12 males, median age 71, range 35-89). REIMS was able to reliably distinguish between cancer and normal adjacent mucosa (NAM) (AUC 0.96) and between NAM and adenoma (AUC 0.99). It had an overall accuracy of 94.4 % for the detection of cancer versus adenoma and an adenoma sensitivity of 78.6 % and specificity of 97.3 % (AUC 0.99) versus cancer. Long-chain phosphatidylserines (e.g., PS 22:0) and bacterial phosphatidylglycerols were over-expressed on cancer samples, while NAM was defined by raised plasmalogens and triacylglycerols expression and adenomas demonstrated an over-expression of ceramides. REIMS was able to classify samples according to tumor differentiation, tumor budding, lymphovascular invasion, extramural vascular invasion and lymph node micrometastases (AUC's 0.88, 0.87, 0.83, 0.81 and 0.81, respectively). During endoscopic deployment, colonoscopic REIMS was able to detect target lipid species such as ceramides during hot snare polypectomy.

CONCLUSION

REIMS demonstrates high diagnostic accuracy for tumor type and for established histological features of poor prognostic outcome in CRC based on a multivariate analysis of the mucosal lipidome. REIMS could augment endoscopic and imaging technologies for precision phenotyping of colorectal cancer.