Premium
Poster ‐ 14: Batch Effect Reduction in in‐vitro Raman Microscopic Radiosensitivity Study Using Ovarian Cancer Cells
Author(s) -
Moradi Hamid,
Murugkar Sangeeta,
Ahmad Abrar,
Shepherdson Dean,
Nyiri Balazs,
Vuong Nhung,
Niedbala Gosia,
Vanderhyden Barbara,
Eapen Libni
Publication year - 2016
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4961788
Subject(s) - radiosensitivity , raman spectroscopy , spectral line , irradiation , biology , materials science , nuclear medicine , physics , optics , medicine , astronomy , nuclear physics
Purpose: To improve classification by reducing batch effect in samples from the ovarian carcinoma cell lines A2780s (parental wild type) and A2780cp (cisplatin cross‐radio‐resistant), before, right after, and 24 hours after irradiation to 10Gy. Methods: Spectra were acquired with a home built confocal Raman microscope in 3 distinct runs of six samples: unirradiated s&cp (control pair), then 0h and 24h after irradiation. The Raman spectra were noise reduced, then background subtracted with SMIRF algorithm. ∼35 cell spectra were collected from each sample in 1024 channels from 700cm‐1 to 1618cm‐1. The spectra were analyzed by regularized multiclass LDA. For feature reduction the spectra were grouped into 3 overlapping group pairs: s‐cp, 0Gy–10Gy0h and 0Gy10–Gy24h. The three features, the three differences of the mean spectra were mapped to the analysis sub‐space by the inverse regularized covariance matrix. The batch effect noticeably confounded the dose and time effect. Results: To remove the batch effect, the 2+2=4D subspace extended by the covariance matrix of the means of the 0Gy control groups was subtracted from the spectra of each sample. Repeating the analysis on the spectra with the control group variability removed, the batch effect was dramatically reduced in the dose and time directions enabling sharp linear discrimination. The cell type classification also improved. Conclusions: We identified a efficient batch effect removal technique crucial to the applicability of Raman microscopy to radiosensitivity studies both on cell cultures and potential clinical diagnostic applications.