I. Sokolov, M. E. Dokukin, V. Kalaparthi, M. Miljkovic, A. Wang, J. D. Seigne, P. Grivas, and E. Demidenko. Noninvasive diagnostic imaging using machine-learning analysis of nanoresolution images of cell surfaces: Detection of bladder cancer. PNAS, 115:12920-12925, 2018. https://doi.org/10.1073/pnas.1816459115
The cystoscopy exam is the current standard for bladder cancer detection and patient follow-up. It is well documented that men at risk for bladder cancer often try to avoid this unpleasant and costly procedure, which considerably contributes to bladder cancer detection at the stage when treatment fails. This work reports on a multi-center (Tufts University, Geisel School of Medicine at Dartmouth, Dartmouth-Hitchcock Medical Center, University of Washington) initiative to develop a novel noninvasive bladder cancer diagnostic tool based on machine learning techniques for discrimination of epithelial tumor cells found in urine using nanoscale cell surface images. Atomic force microscopy (AFM) is applied in combination with a subresonance tapping technique to extract mechanical properties of a cell surface for a subsequent assessment of whether the cell is normal or cancerous. Our approach has been verified on 68 patients (43 controls and 25 bladder cancer patients) and resulted in 94% diagnostic accuracy, a statistically significant improvement compared to the standard cystoscopy exam (p < 0.05). Important advantages of our approach include low cost and the ability to repeat the AFM cell imaging exam from urine samples as often as needed.
Future studies and clinical trials are needed to fully understand the advantages of our approach and find its place in the current practice of clinical oncology. If successful, AFM may have an impact on the early detection of bladder cancer and patient monitoring after treatment.