Crater-Spectrum Feature Fusion Method for Panax notoginseng Cadmium Detection Using Laser-Induced Breakdown Spectroscopy

This study from Zhejiang University and Yunnan University of Chinese Medicine developed and validated a laser-induced breakdown spectroscopy (LIBS) method for rapid, non-destructive quantification of cadmium in Panax notoginseng (san qi, tienchi ginseng) powder. The reference cadmium concentrations in 240 commercial powder samples, measured by ICP-MS, ranged from 0.45 to 75.20 µg/g (dry weight) across six brands, confirming substantial cadmium accumulation in commercially available P. notoginseng. The paper’s primary contribution is methodological: a crater-spectrum feature fusion approach that reduces signal uncertainty caused by matrix heterogeneity among brands, achieving a root-mean-square error of prediction (RMSEP) of 2.8556 µg/g using a least-squares support vector machine model.

Study design

Six commercial brands of P. notoginseng powder were obtained from Chinese markets. Brand identifiers are anonymized in the paper; standards compliance ranged across Chinese Pharmacopoeia, primary agricultural product standards, and local pharmacopoeia. Forty samples were collected per brand, yielding 240 samples total. Cadmium-spiked calibration sets were prepared by mixing cadmium nitrate tetrahydrate solution into 4 g portions of dry powder at concentrations of 0, 0.5, 1, 10, 20, 30, 40, 50, 60, and 70 µg/g to produce ten contamination levels (40 samples per concentration per brand). Samples were pressed into 13 mm diameter tablets at 20 MPa. The 240 unspiked samples were divided into calibration (Datasets 1-3, 30 samples per brand) and prediction (Dataset 4, 10 samples per brand) sets.

Cadmium reference values for all samples were obtained by ICP-MS (Agilent 7800ICPMS) following nitric acid digestion. LIBS measurements used a Q-switched Nd:YAG 532 nm pulsed laser with an 8 ns pulse duration, a 2 µs delay time, and 40 mJ pulse energy operating at 1 Hz. Spectra were collected in the 210-231 nm range using a monochromator and intensified CCD detector. Each ablation position was sampled in a 2x3 crater array with nine pulses per position, and spectra were averaged to reduce point-to-point variation.

Key numbers

Cadmium concentrations in the 240 unspiked commercial powder samples, determined by ICP-MS (dry weight, µg/g), are reported by brand in Table 2 of the paper:

Across all 240 samples, minimum cadmium was 0.45 µg/g dw and maximum was 75.20 µg/g dw, with means ranging from 29.90 to 32.88 µg/g dw across brands. The wide within-brand ranges (minimum-to-maximum ratios exceeding 100-fold in some brands) reflect differences in cultivation origin, soil contamination, and processing. These ICP-MS reference values, not the LIBS model predictions, are the occurrence data reported in this paper.

LIBS model performance is reported as RMSEP and R-squared (prediction set). The best model from the crater-spectrum feature fusion approach using a least-squares support vector machine (LSSVM) achieved: Rp2 = 0.9885, RMSEP = 2.8556 µg/g (Table 9, p. 14). For comparison, an uncorrected single-emission-line calibration curve at 228.80 nm yielded RMSEP = 7.6517 µg/g (Table 6, p. 12). After characteristic peak ratio correction (CPRC), the calibration curve model at 228.80 nm achieved a limit of detection of 1.92 µg/g and a limit of quantification of 6.41 µg/g (p. 13). The CPRC-corrected PLSR model reached RMSEP = 3.1889 µg/g (Table 8, p. 13).

LIBS emission lines identified for cadmium in P. notoginseng were cadmium II at 214.44 nm, cadmium II at 226.50 nm, and cadmium I at 228.80 nm. Matrix interference from aluminum (221.00, 226.35, 226.91 nm), silicon (212.41, 220.80, 221.09, 221.17, 221.67, 221.81 nm), calcium (211.28, 212.30, 215.08, 227.55 nm), and iron (213.65, 213.70, 215.24, 215.59, 221.10, 221.71, 226.48, 226.50, 227.71 nm) overlapped with the cadmium II 226.50 nm line in particular (Table 3, p. 7).

Ablation crater morphology varied substantially across brands. The nine crater parameters showed relative standard deviations exceeding 20% across brands: volume RSD 76%, cross-sectional area RSD 46%, surface area RSD 51%, average depth RSD 32%, maximum depth RSD 27%, perimeter RSD 38%, horizontal Feret diameter RSD 24%, vertical Feret diameter RSD 23%, circle equivalent diameter RSD 22% (Table 4, p. 10). This variability reflects differences in physical texture and density of commercial P. notoginseng powders from different origins and processing methods.

Methods (brief)

The analytical approach combined two signal uncertainty reduction strategies. Crater morphology compensation used multiple linear regression (MLR) incorporating ablation crater parameters (volume, average depth, maximum depth, horizontal Feret diameter) selected by stepwise regression as explanatory variables alongside cadmium emission line intensities. Characteristic peak ratio correction (CPRC) divided cadmium emission line intensities by a matrix-variable mean computed from selected wavelength positions highly correlated with cadmium concentration, reducing matrix-driven signal fluctuation. The crater-spectrum feature fusion method combined both approaches, fusing single-shot LIBS spectra with crater morphology parameters and feeding corrected intensities to MLR, LSSVM, and random forest models. Data analysis used MATLAB 2019b. Crater morphology was measured with a shape measurement laser microscope (Keyence VK-X1050). Certified reference material and recovery data are not reported.

Evidence fitness

This paper supports statements about cadmium occurrence in commercial P. notoginseng powder, with ICP-MS as the reference method for concentration determination. The wide cadmium range (0.45-75.20 µg/g dry weight across 240 samples) is a direct occurrence finding applicable to the herbal-botanicals ingredient category and to botanical supplement products containing P. notoginseng as a labeled ingredient. The paper also supports context on LIBS as a rapid screening method for cadmium in complex botanical matrices, with performance benchmarks relative to ICP-MS.

The paper does not support statements about other heavy metals in P. notoginseng, human dietary exposure or risk assessment, or cadmium occurrence in other herbal species. The ICP-MS reference method lacks reported detection limits and certified reference material recovery data, so individual low-end values near the minimum of 0.45-0.98 µg/g carry greater uncertainty than the high-end values.

Limitations

Samples were sourced from Chinese markets only; cadmium concentrations in P. notoginseng sourced from other growing regions may differ. Calibration sets were prepared from spiked powder rather than naturally contaminated field samples, which may not fully represent the chemical matrix of real contamination. The ICP-MS reference method does not report certified reference material verification or spike recovery data. No dietary intake or risk characterization is performed by the authors. The LIBS method achieves a limit of detection of 1.92 µg/g and limit of quantification of 6.41 µg/g after CPRC, which exceeds the Chinese Pharmacopoeia 2020 cadmium limit of 1.0 µg/g for P. notoginseng root, meaning LIBS as described cannot reliably confirm compliance at the regulatory limit.

Regulatory context

The Chinese Pharmacopoeia 2020 specifies a maximum cadmium content of 1.0 mg/kg (1.0 µg/g dry weight) for Panax notoginseng root (Radix et Rhizoma Notoginseng). Based on ICP-MS reference measurements in this study, cadmium means of approximately 30 µg/g across all six brands substantially exceed this pharmacopoeial limit. The authors acknowledge cadmium contamination as a major quality and safety concern for P. notoginseng but do not perform a formal regulatory compliance analysis. No EU, US FDA, or Codex limit specific to P. notoginseng is cited in the paper.

Related evidence

• Cadmium
• Herbal botanicals
• Ingredient class — botanicals/herbs/traditional formulas

Update history

The five most recent substantive edits to this page, classified major (evidence or structure moved), correction (a published value or statement was wrong and has been fixed), or minor (narrative rewritten without changing the underlying evidence). Each description is derived from what the edit did to this page; the linked commit is the authoritative record, routine regeneration passes are excluded, and the full version history lives in git.