[960] Raman Spectroscopy of Oncocytic Kidney Tumors.

Maria Tretiakova, Shona Stewart, John Maier. University of Chicago, IL; ChemImage Corporation, Pittsburgh

Background: The differential diagnosis of malignant chromophobe renal cell carcinoma (ChRCC) and benign renal oncocytoma (ONC) can be challenging. Even though morphologically ChRCC and ONC significantly overlap, they are characterized by different genetic, ultrastructural and molecular composition. Raman Molecular Imaging (RMI) is a non-destructive optical technique based on the analysis of laser light scattered from a sample. This molecule-specific scattering allows elucidation of relative molecular proportions and biochemical bonds in a tissue, thus generating tissue-specific “fingerprints”. We hypothesized that different molecular make-ups of ChRCC and ONC can be captured by RMI.
Design: Four micron unstained paraffin tissue microarray sections containing 25 cases of ChRCC and 7 ONC were subjected to RMI using Falcon II™ Wide-Field Raman Chemical Imaging System, and analyzed by ChemImage Xpert® software. For each case, represented by three 1 mm tissue cores, multiple RMI spectra were extracted, followed by principal component analysis (PCA). The true diagnoses were kept blinded during the data acquisition and processing steps. Additionally, we performed immunohistochemical staining with CK7 and CD117 antibodies.
Results: Evaluation of 615 spectra from 32 cases was expressed graphically in scatter plots after PCA, which, following several optimization steps, allowed classification of cases in either ChRCC or ONC group based on the values of principal component 3 (t-test, p < 0.001). There was 90% concordance between different spectra derived from each case. However significant overlap existed between two groups of tumors, and after unblinding of histology data 2 ChRCC and 2 ONC were found to be misclassified. Comparison of RMI and immunostaining with CK7 and CD117 is shown in a table.

Cross-validation of RMI vs cytoplasmic CK7 and membranous CD117 staining

Method	Sensitivity	Specificity	PPV	NPV	Accuracy
RMI	85.7%	76%	50%	95%	0.79
CK7	100%	60%	41%	100%	0.8
CD117	100%	64%	44%	100%	0.8

Conclusions: This is the first study attempting molecular “fingerprinting” of ChRCC and ONC. It appears that both tumors show very close, and often overlapping, spectral characteristics by RMI proving their biochemical similarity. However, correct molecular classification of these tumors was possible in 92% of ChRCC and 71% ONC cases. Comparison of RMI with well-established markers CK7 and CD117 demonstrated comparable accuracy, lower sensitivity, but higher specificity. The diagnostic utility of RMI needs to be further elucidated on a larger data set.
Category: Genitourinary (including renal tumors)

Wednesday, March 2, 2011 9:30 AM

Poster Session V # 92, Wednesday Morning