Explainable Feature Engineering for Multi-Modal Tissue State Monitoring Based on Impedance Spectroscopy

Abstract

One of the most promising approaches to food quality assessments is the use of impedance spectroscopy combined with machine learning. Thereby, feature selection is decisive for a high classification accuracy. Physically based features have particularly significant advantages because they are able to consider prior knowledge and to concentrate the data into pertinent understandable information, building a solid basis for classification. In this study, we aim to identify physically based measurable features for muscle type and freshness classifications of bovine meat based on impedance spectroscopy measurements. We carry out a combined study where features are ranked based on their F1-score, cumulative feature selection, and t-distributed Stochastic Neighbor Embedding (t-SNE). In terms of features, we analyze the characteristic points (CPs) of the impedance spectrum and the model parameters (MPs) obtained by fitting a physical model to the measurements. The results show that either MPs or CPs alone are sufficient for detecting muscle type. Combining capacitance (C) and extracellular resistance (Rex) or the modulus of the characteristic point Z1 and the phase at the characteristic frequency of the beta dispersion (Phi2) leads to accurate separation. In contrast, the detection of freshness is more challenging. It requires more distinct features. We achieved a 90% freshness separation using the MPs describing intracellular resistance (Rin) and capacitance (C). A 95.5% freshness separation was achieved by considering the phase at the end of the beta dispersion (Phi3) and Rin. Including additional features related to muscle type improves the separability of samples; ultimately, a 99.6% separation can be achieved by selecting the appropriate features.

Keywords: F1-score; bioimpedance spectroscopy; bovine meat; explainable machine learning; machine learning; meat freshness; reliable machine learning.

Figure 1

Bioimpedance measurement setup: (a) electrodes’ design, (b) Agilent 4294 A, together with the examined bovine meat.

Figure 2

LD beef impedance spectrum in Bode and Nyquist plots with 3 characteristic points (CPs) that separate $\alpha ,\beta ,\mathrm{and}\gamma$ regions.

Figure 3

Modified Fricke Model extended by a series resistance.

Figure 4

Fitting the Fricke model to the impedance spectrum of the LD Beef muscle measured on day 14.

Figure 5

Correlation matrix of the model parameters.

Figure 6

Correlation matrix of characteristic point parameters.

Figure 7

Feature correlation map.

Figure 8

(a) Features ranked by their importance, based on their impact on F1-score, and (b) F1-scores following cumulative feature selection for muscle detection based on model parameters.

Figure 9

t-SNE sample separability tests for muscle detection using model parameters, based on (a) C+Rex and (b) C+Rin.

Figure 10

(a) Features ranked by their importance based on their impact on F1-score, and (b) F1-scores following cumulative feature selection for muscle detection based on characteristic points.

Figure 11

t-SNE sample separability test of muscle detection based on characteristic points.

Figure 12

(a) Features ranked by their importance based on their impact on F1-score, and (b) F1-scores following cumulative feature selection for muscle detection based on combinations of model parameters and characteristic points.

Figure 13

t-SNE sample separability test for muscle detection using Z1 and C features.

Figure 14

(a) Features ranked by their importance based on their impact on F1-score, and (b) F1-scores following cumulative feature selection for freshness assessments based on model parameters.

Figure 15

t-SNE sample separability test for freshness assessments using Rin and C features.

Figure 16

(a) Features ranked by their importance based on their impact on F1-score, and (b) F1-scores following cumulative feature selection for freshness assessments based on characteristic points.

Figure 17

t-SNE sample separability test for freshness assessments using characteristic point features.

Figure 18

(a) Features ranked by their importance based on their impact on F1-score, and (b) F1-scores following cumulative feature selection for freshness assessments based on combined model parameter and characteristic point features.

Figure 19

t-SNE sample separability test for freshness assessments using combined model parameter and characteristic point features.

Figure 20

Phi3, C feature projection for sample freshness separability. The arrows indicate the direction of aging.

Figure 21

Rin, C, and Rex feature projection for sample freshness separability. The arrows indicate the direction of aging.