Patent information
IP India · live recordAbstract
This invention presents a novel method for enhancing autofocus accuracy in Whole Slide Imaging systems through a spiral based equidistant sampling strategy for initial focus point selection. Traditional autofocus methods employ a regular grid of focus points, which can introduce aliasing artifacts, poorly adapt to non rectangular scan geometries, and fail to account for surface irregularities in glass slides. To overcome these limitations, the proposed approach utilizes Archimedean or Fermat spiral geometries to distribute focus points equidistantly across the scanning region. This spiral sampling pattern improves spatial uniformity, reduces interpolation distortion in the generated 3D depth map, and offers greater adaptability to the topological variations commonly encountered in biological specimens. The method is computationally efficient and especially advantageous for low-cost and GPU accelerated whole slide imaging systems, providing a robust initialization for adaptive autofocus algorithms without increasing system complexity
Applicants
1Inventors
2Claims
5Specification
Show full textCollapse
Description:DETAILED DESCRIPTION OF THE INVENTION:
In conventional Whole Slide Imaging (WSI) systems, the initial generation of a 3D depth map for autofocusing is critical for ensuring sharp and accurate digital slide images. A common method involves selecting a set of 9 or more initial focus points arranged in a regular rectangular grid pattern across the scanning area. While simple, this method can lead to aliasing artifacts in the interpolated depth surface due to the symmetric and repetitive structure of the sampling grid, particularly when the slide contains non-planar regions or subtle topological undulations.
Spiral-Based Sampling Innovation We propose an enhanced sampling scheme based on a spiral pattern with equidistant spacing along the spiral path rather than fixed Cartesian grid points. This non-linear, quasi-random sampling technique offers superior spatial distribution and reduces interpolation bias and aliasing effects when generating the initial 3D surface model.
Technical Description of Spiral-Based Sampling 1. Spiral Geometry The proposed sampling follows an Archimedean spiral or a Fermat spiral layout, defined by the polar equation: r(θ) = a + bθ (Archimedean) or rₙ = c√n, θₙ = n·φ (Fermat) where φ is the golden angle (~137.5°), ensuring even angular separation. 2. Equidistant Point Placement
Spiral placement distributes points more uniformly across the slide area, especially in circular or elliptical regions common in WSI scan windows. This ensures better topological coverage and reduces interpolation distortion. 3. Algorithmic Advantage
The spiral placement algorithm is efficient and allows dynamic adjustment of point numbers based on scan area. It complements adaptive sampling phases and minimizes the need for post-correction.
Benefits over Grid Sampling
Feature Rectangular Grid Spiral Equidistant Aliasing Resistance Low High Surface Fit Accuracy Moderate High Topological Adaptability Poor Good Visual Uniformity Rigid Natural Application in Autofocusing
In our proposed WSI system, the initial spiral-based samples are used to create a first-pass 3D depth map. During scanning, focus inference is run at each spiral point. If the focus score is acceptable, it is accepted; otherwise, a precise focus step is triggered. The depth map is refined but initialized robustly using spiral-sampled data. , Claims:1. A method for initial focus point selection in WSI systems using spiral-based equidistant sampling rather than a conventional grid. 2. The use of Archimedean or Fermat spiral geometries to place focus points with uniform angular separation and equidistant radial spacing. 3. An autofocusing mechanism that utilizes spiral distributed points for first pass depth map generation, followed by conditional refinement. 4. A sampling algorithm that dynamically adjusts the number of spiral points based on the scanning area. 5. An imaging system incorporating the above method to enhance auto focusing accuracy, minimize aliasing, and improve interpolation in 3D depth modeling.