Differential diagnosis of Crohn’s disease and intestinal tuberculosis based on ATR-FTIR spectroscopy combined with machine learning

Figure 1 Argentum film on glass slides.

Figure 2 Schematic of XGBoost.

Figure 3 Pathological images of Crohn’s disease and intestinal tuberculosis. A: Typical pathological images of Crohn’s disease (CD); B: Pathological images of typical intestinal tuberculosis (ITB); C: Atypical pathological images of CD; D: pathological images of atypical ITB.

Figure 4 Background absorption spectra of quartz and paraffin. A: Quartz; B: Paraffin.

Figure 5 Absorption spectra of dewaxed and unwaxed tissue sections.

Figure 6 Optimization of dewaxing duration. A: Absorption spectra of tissues during the dewaxing process for 6-42 h; B: Absorption spectra of tissues during the dewaxing process for 24-30 h.

Figure 7 Fourier transform infrared spectrometers equipped with accessories for attenuated total reflectance spectral analysis of intestinal tuberculosis and Crohn’s disease. A: Original spectrum; B: Spectral intensity of the characteristic bands. CD: Crohn’s disease; ITB: Intestinal tuberculosis.

Figure 8 Derivative spectrum. A: First-derivative spectra; B: Second-derivative spectra. CD: Crohn’s disease; ITB: Intestinal tuberculosis.

Figure 9 Derivative spectral characteristic intensity analysis. CD: Crohn’s disease; ITB: Intestinal tuberculosis.

Figure 10 Schematic diagram of XGBoost used to identify intestinal tuberculosis and Crohn’s disease. A: Original spectra; B: Derivative spectra; C: Feature extraction; D: Modeling analysis; E: Classification results. CD: Crohn’s disease; ITB: Intestinal tuberculosis; PCA: principal component analysis; ICA: Independent component analysis.

Figure 11 XGBoost model optimization results based on first derivative spectral data. A: Characteristic component; B: Max_depth; C: N_estimators; D: Min_child_weight; E: Gamma; F: Subsample; G: Alpha; H: Learning_rate.