SEM Studio

1 Introduksi

1.1 Latar Belakang

SEM Studio adalah expert desk untuk quantitative image analysis SEM micrograph: automated cell segmentation, cell size distribution, area fraction calculations, dan morphological feature extraction (aspect ratio, circularity, solidity). Berbeda dari Quick SEM Structure Check yang fokus eksplorasi visual + manual annotation, SEM Studio menerapkan image processing pipeline lengkap berbasis ImageJ/Fiji methodology (Schindelin et al., 2012; Schneider, Rasband, & Eliceiri, 2012).

Pada konteks microbial research, SEM Studio mengukur: (i) cell count per area (cells/μm²) untuk biomass density, (ii) cell size distribution untuk uniformity assessment, (iii) damaged vs intact cell fraction post-antimicrobial treatment, dan (iv) biofilm coverage sebagai % area covered.

1.2 Tujuan Modul

Modul SEM Studio di SQalytics ditujukan untuk:

• Memproses SEM image dengan standard pipeline: noise reduction, thresholding (Otsu, manual), morphological cleanup, watershed segmentation.
• Mengukur cell features: area, perimeter, equivalent diameter, aspect ratio, circularity, solidity.
• Menampilkan cell size distribution + statistics ($d_{10}$, $d_{50}$, $d_{90}$).
• Mendukung comparison multi-image dengan ANOVA cell density antar treatment.
• Audiens: peneliti S2/S3 dengan SEM analysis untuk publikasi.

1.3 Posisi di Antara Alternatif

Pilih SEM Studio untuk quantitative image analysis. Untuk visual exploration + manual annotation, pakai Quick SEM Structure Check. Untuk statistik cell count antar treatment, ekspor hasil SEM Studio + pakai Compare Many Groups. Untuk PSD non-SEM (laser diffraction), pakai Particle Size Distribution.

2 Metode

2.1 Dasar Teoretis

Image processing pipeline (Gonzalez & Woods, 2018):

1. Pre-processing: Gaussian smoothing ($\sigma$ 1–2 pixel) untuk noise reduction; background subtraction (rolling ball $r$ 50 pixel) untuk uniform illumination.
2. Thresholding — pisahkan foreground (cells) vs background:
   • Otsu's method (Otsu, 1979) — automatic threshold maximizing between-class variance:
   $$\sigma_B^2(t) = w_0(t)\, w_1(t)\, [\mu_0(t) - \mu_1(t)]^2$$
   • Adaptive thresholding untuk illumination gradient.
3. Morphological operations: Opening (erosion + dilation) untuk remove noise dots; Closing (dilation + erosion) untuk fill holes within cells.
4. Watershed segmentation untuk separasi cells touching: $$\text{Markers} = \text{ultimate eroded points}; \quad \text{Watershed lines} = \text{ridges in distance transform}$$

Cell features (Russ & Neal, 2017):

• Area $A$ — sum pixels × scale² (μm²).
• Perimeter $P$ — boundary length (μm).
• Equivalent diameter: $d_{\text{eq}} = \sqrt{4 A / \pi}$ — diameter circle dengan area sama.
• Aspect ratio: $AR = L_{\text{major}} / L_{\text{minor}}$ (fit ellipse).
• Circularity: $C = 4\pi A / P^2$, range 0–1, 1 = perfect circle.
• Solidity: $S = A / A_{\text{convex hull}}$, indikasi kompaktness; 1 = convex.

Cell density per area:

dengan $A_{\text{FOV}}$ field of view area dalam μm².

Damage fraction (post-antimicrobial):

Damaged identifikasi via threshold pada circularity ($C < 0.7$) atau solidity ($S < 0.8$) — sel rusak biasanya irregular shape.

2.2 Persamaan Inti

Otsu threshold: $t^* = \arg\max_t \sigma_B^2(t)$

Equivalent diameter: $d_{\text{eq}} = \sqrt{4A/\pi}$

Circularity: $C = 4\pi A / P^2$

Solidity: $S = A / A_{\text{convex hull}}$

Aspect ratio: $AR = L_{\text{major}} / L_{\text{minor}}$

Cell density: $\rho = N / A_{\text{FOV}}$

2.3 Asumsi & Batas Validitas

3 Cara Kerja

3.1 Step-by-Step di SQalytics

1. Buka SEM Studio dari domain Mikrobiologi Pangan (Expert Desks).
2. Upload SEM image (rekomendasi TIFF 16-bit grayscale).
3. Kalibrasi scale (auto dari TIFF metadata atau manual via scale bar).
4. Atur pre-processing: Gaussian sigma, background subtraction radius.
5. Pilih thresholding: Otsu (auto), adaptive, atau manual slider.
6. Atur morphological: opening/closing kernel size.
7. (Opsional) Aktifkan watershed untuk separasi touching cells.
8. Atur minimum cell size (filter noise, default 0.1 μm²).
9. Klik Run Image Analysis.
10. Tinjau hasil:
    • Tab Segmented Image — overlay outline + numbered.
    • Tab Cell Statistics — features table per cell.
    • Tab Size Distribution — histogram + $d_{10/50/90}$.
    • Tab Damaged vs Intact — classification + fraction.
    • Tab Multi-Image Comparison — bila lebih dari 1 image.

3.2 Template Input + Contoh

Input: 3 SEM images S. aureus (5,000×):

• control.tif — no treatment.
• low_dose.tif — Singkil 5 mg/mL.
• high_dose.tif — Singkil 10 mg/mL.

3.3 Contoh Luaran

Cell features summary (per image, mean ± SD):

ANOVA cell density: $F_{2, 12} = 24.5$, $p < 0.001$ — perbedaan signifikan antar treatment.

4 Kesimpulan

4.1 Relevansi Real-World

• Antimicrobial research publication — quantitative damage markers untuk paper.
• Biofilm coverage quantification — % area covered pre/post sanitizer.
• Probiotik viability + morphology — uniformity batch.
• Food contact surface — bacterial adhesion area fraction.
• Spore germination study — count + morphology kinetics.

4.2 Where to Go from Here

• Quick SEM Structure Check — quick visual untuk first look.
• Compare Many Groups — Tukey post-hoc dari cell density data.
• Particle Size Distribution — paralel data laser diffraction.
• Fiji/ImageJ untuk advanced custom pipeline.

⚙ Troubleshooting Cepat

r Riwayat Revisi