Optimisation of vibrational spectroscopy instruments and pre-processing for classification problems across various decision parameters

Joy Sim; Cushla McGoverin; Indrawati Oey; Russell Frew; Biniam Kebede; Joy Sim; Cushla McGoverin; Indrawati Oey; Russell Frew; Biniam Kebede

doi:10.48130/fia-0024-0004

Figures (4) Tables (1)

Figure 1.
Raw spectra from instruments before pre-processing data treatment: (a) DG-NIR, (b) HSI-NIR, (c) ATR-FTIR, (d) Raman. GCP samples are labelled according to the country of origin (B: Brazil, E: Ethiopia, I: Indonesia, R: Rwanda).
Figure 2.
(a) Pre-processed DG-NIR spectra, (b) scores, (c) first loading, (d) second loading of (i) MSC with MNCN, (ii) SNV with MNCN, (iii) EMSC with MNCN pre-processed DG-NIR spectra.
Figure 3.
(a) Pre-processed HSI-NIR spectra measured in reflectance, (b) scores, (c) first loading of (i) Normalisation with MNCN, (ii) SG (1^st der, 2^nd poly, 15 pts) with MNCN, (iii) Normalisation, SG (1^st der, 2^nd poly, 15 pts) with MNCN pre-processed HSI-NIR spectra.
Figure 4.
(a) Pre-processed Raman spectra, (b) scores, (c) first loading of (i) MSC with MNCN, (ii) EMSC with MNCN, (iii) WLS, Normalisation with MNCN pre-processed Raman spectra. Loading plots were produced using R.

Optimised pre-processing		TVar %	RMSECV	RMSEC	RMSEP	MCC, Pred.	Accuracy, Pred	F1, Pred.
DG-NIR	MNCN	98.54	0.338	0.329	0.473	0.383	0.665	0.483
	MSC, MNCN	91.49	0.245	0.243	0.309	0.774	0.882	0.865
	SNV, MNCN	91.49	0.245	0.243	0.309	0.774	0.882	0.865
	SNV, Detrend, MNCN	91.32	0.245	0.243	0.309	0.774	0.882	0.865
	MSC, SG (1^st der, 2^nd poly, 15 pts), MNCN	76.06	0.268	0.265	0.350	0.684	0.835	0.788
	Normalisation, SG (2^nd der, 2^nd poly, 7 pts), MNCN	98.05	0.358	0.352	0.351	0.652	0.812	0.757
	EMSC, MNCN	87.87	0.240	0.238	0.250	0.876	0.929	0.916
HSI-NIR	MNCN	99.69	0.372	0.362	0.421	0.618	0.800	0.681
	Normalisation, MNCN	98.28	0.333	0.322	0.402	0.655	0.767	0.650
	SG (1^st der, 2^nd poly, 15 pts), MNCN	68.87	0.341	0.325	0.364	0.636	0.800	0.728
	MSC, SG (1^st der, 2^nd poly, 15 pts), MNCN	63.41	0.338	0.324	0.403	0.473	0.733	0.605
	Normalisation, SG (1^stder, 2^nd poly, 15 pts), MNCN	85.79	0.324	0.313	0.375	0.612	0.800	0.732
	SNV, SG( 1^st der, 2^nd poly, 15 pts), MNCN	63.38	0.337	0.324	0.402	0.473	0.733	0.605
FTIR	MNCN	99.69	0.335	0.321	0.386	0.253	0.452	0.200
	Normalisation, MNCN	98.17	0.334	0.320	0.391	0.372	0.452	0.179
	Normalisation, SG (1^st der, 2^nd poly, 15 pts), MNCN	97.12	0.402	0.369	0.490	0.141	0.500	0.330
	EMSC, MNCN	71.52	0.409	0.384	0.482	0.286	0.500	0.326
Raman	MNCN	99.91	0.319	0.312	0.329	0.756	0.860	0.819
	SG (2^nd der, 2^nd poly, 7 pts), MNCN	99.66	0.350	0.343	0.369	0.521	0.735	0.648
	Normalisation, SG (1^st der, 2^nd poly, 15 pts), MNCN	98.94	0.321	0.315	0.334	0.554	0.747	0.622
	WLS (2^nd poly), MNCN	96.86	0.343	0.336	0.372	0.611	0.795	0.760
FT, Fourier-Transform; DG, Dispersive; HSI, Hyperspectral Imaging; NIR, near-infrared; TVar, Total explained variance; RMSE(C/CV/P), Root Mean Square Errors of Calibration/Cross-Validation/Prediction; MCC, Matthew's Correlation Coefficient, MNCN, Mean centering; MSC; Multiplicative Scatter Correction; EMSC, Extended Multiplicative Scatter Correction; SG (#der, #poly, #pts), Savitzky-Golay #derivative, #polynomial, #window points; WLS, Weighted Least Squares; Pred., Prediction.

Table 1.

Prediction statistics associated with optimal pre-processing methods for spectral data collected using DG-NIR, HSI-NIR, FTIR and Raman.