Nasopharyngeal cancer (NPC) is a scourge which affects South China and Southeast Asia. Driven by the Epstein-Barr virus, this cancer progresses more quickly than most other cancers. Delaying detection by months can be the difference between a highly treatable and a lethal tumor. Unlike most cancers which tend to present in older age (~65 years), the population most at risk for NPC are men aged 45 to 50, a period when their contributions to family and society are particularly important. Currently, we can detect this disease with high accuracy using a test called the immunofluorescence assay (IFA). This however has not changed the fact that most patients are still detected in the late stages (3 & 4). One reason for this seeming paradox is that IFA is not scalable and requires time-consuming evaluation by highly-trained pathology staff. In short, we have the perfect test, yet lives are needlessly lost. Pathnova proposes to solve this problem by automating the IFA using pattern recognition to simultaneously detect and quantitate NPC disease load for clear cut cases. Borderline cases will still be evaluated by a human pathology expert. Our team comprises Emeritus Professor Chan Soh Ha who pioneered the IFA in the 1970s and Dr. Ian Cheong's lab, which has devised machine learning technology specifically for automated IFA analysis.

Goh S, Swaminathan M, Lai J, Anwar A, Chan SH, Cheong I. (2017) Increasing the accuracy and scalability of the Immunofluorescence Assay for Epstein Barr Virus by inferring continuous titers from a single sample dilution. Journal of Immunological Methods 440: 35-40.


  1. Image analysis may be used to infer continuous EBV titers from IFA.

  2. Only a single sample dilution is required, hence increasing IFA scalability.

  3. Titer inference is accurate even if performed 3 days after IFA.

Read our paper here.

Swaminathan M, Yadav PK, Piloto O, Sjöblom T, Cheong I. (2017) A New Distance Measure for Non-Identical Data with Application to Image Classification. Pattern Recognition 63: 384–396.


  1. Empirical evidence is provided that real-world data is non-identically distributed.

  2. PBR, the first distance measure to account for non-identical data is proposed.

  3. PBR was tested in 6 test applications using 12 benchmark data sets.

  4. PBR outperforms state-of-the-art measures for most data sets.

  5. Avoiding the identical distribution assumption can improve classification.

Read our paper here.

All rights reserved. Contents © 2017 Pathnova Laboratories Pte Ltd

Disclaimer : The information contained in this website is provided for informational purposes only, and should not be construed as medical advice on any matter. Please contact us for specific queries.​