Mikhail Silk BS, Presenter: Nothing to Disclose

Dmitry Dylov, Abstract Co-Author: Nothing to Disclose

Siavash Yazdanfar, Abstract Co-Author: Nothing to Disclose

Tarik Silk, Abstract Co-Author: Nothing to Disclose

Stephen Barnett Solomon MD, Abstract Co-Author: Research Grant, General Electric Company Research Grant, AngioDynamics, Inc Consultant, Johnson & Johnson Consultant, Covidien AG Director, Devicor Medical Products, Inc Director, Aspire Bariatrics, Inc

Jeremy C. Durack MD, Abstract Co-Author: Scientific Advisory Board, Investor - Adient Medical Research Grants - Society of Interventional Radiology Foundation, Prostate Cancer Foundation

To develop and validate an instrument to rapidly discriminate between renal cancer neoplastic subtypes and normal core biopsy tissue using elastic light scatter spectroscopy.

We performed an Institutional Review Board approved prospective study of surgically resected kidney tumors with a clear pathologic diagnosis from 1/2013 - 2/ 2014. Visible tumors and surrounding normal kidney were biopsied using 18G side-notch core needles. Core biopsy specimens were analyzed using a specialized light spectroscopic scatter device that rapidly scans (less than 1 minute) core needle biopsy samples while still on the needle. Spectra were normalized and distributed against geometrical means and outliers were rejected. The spectral data was decomposed into 25 principal components and a machine learning algorithm was used to differentiate between tumor subtypes and normal tissue. Receiver operating characteristic (ROC) curves were generated using pathology as the gold standard for all samples.

Fifty-three kidneys were biopsied during the study period resulting in 3076 usable spectra after outlier rejection (1272 normal and 1804 tumor samples). The final pathologic diagnoses included clear cell carcinoma (1130/1804,63%), papillary carcinoma (248/1804, 14%), chromophobe carcinoma (226/1804, 13%) and oncocytoma (200/1804, 11%). Principal component analysis using the Random Forest algorithm resulted in a sensitivity of 92.6%, specificity of 93.3%, 95.2% PPV, and 89.9% NPV. Despite overall high accuracy for renal tumor subtyping, the device performed least well differentiating papillary from clear cell carcinoma and normal renal tissue from chromophobe carcinoma.

Rapid tissue-preserving optical spectroscopy analysis of core biopsy samples is feasible and can successfully differentiate renal tumor subtypes with a high degree of classification accuracy. This instrument offers the potential to improve on-site biopsy assessment.

Automated workflow-integrated pathologic assessment of core needle biopsies using optical spectroscopy is possible and has the potential to improve on-site biopsy assessment. 

Silk, M, Dylov, D, Yazdanfar, S, Silk, T, Solomon, S, Durack, J, Rapid Pathologic Subtyping of Kidney Tumors after Ex Vivo Core Needle Biopsy Using Optical Spectroscopy .  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14006949.html