Synthetic and Realistic Laser–Tissue Interaction Data for Machine-Learning Applications

This dataset provides two complementary CSV files containing simulated laser–tissue interaction measurements, designed to support machine-learning research in biomedical optics and computational modeling:

1- **laser_tissue_data.csv**

- Samples: 1,000,000

Features:

- wavelength (nm)
- pulse_duration (ns)
- energy_density (J/cm²)
- absorption_coeff (cm⁻¹)
- thermal_conductivity (W/(m·K))
- fluence (J/cm²/ns) – derived
- beam_profile – simulated optical absorption
-- success (0/1) – binary target indicating whether the laser pulse meets thermal diffusion and energy thresholds

These data were generated via uniform, log-normal, and normal distributions to represent idealized measurements of laser–tissue interactions under controlled conditions.

2- **realistic_laser_tissue_data.csv**

- Samples: 1,000,000

Same feature set as above, with two added realism factors:

- Measurement noise (configurable noise_level = 0.1) applied to each physical variable
- Label noise (flip_prob = 0.05) randomly inverting 5% of the success outcomes

This version simulates sensor variance and occasional mislabeling, making it suitable for testing robustness of classification and regression models.

-------------------------------------------------

Use Cases & Applications

- Train and benchmark classification models (e.g., logistic regression, random forests, neural networks) on binary outcome prediction.
- Explore feature-engineering strategies (e.g., interaction terms, normalization) for optical parameters.
- Evaluate the impact of measurement and label noise on model performance and generalizability.
- Develop synthetic-data augmentation pipelines for scarce or ethically constrained biomedical datasets.