senoquant 1.0.0b2__py3-none-any.whl → 1.0.0b3__py3-none-any.whl

senoquant/tabs/spots/models/udwt/model.py

@@ -1,482 +0,0 @@
-"""
-UDWT (a trous) B3-spline wavelet spot detector.
-
-Implements the Olivo-Marin (2002) workflow: build a trous smoothing scales,
-compute wavelet planes, WAT-threshold per scale, multiply planes, and label
-connected components. Supports 2D, 3D, and per-slice 2D for 3D stacks.
-
-Algorithm (high level)
-----------------------
-1) Build smoothing scales A1..AJ with the dilated B3-spline kernel.
-2) Form wavelet planes Wi = A_{i-1} - Ai.
-3) Apply WAT thresholding per scale (with per-scale sensitivity).
-4) Multiply thresholded planes to form the multiscale product PJ.
-5) Threshold |PJ| with ld and label connected components.
-"""
-
-from __future__ import annotations
-
-from dataclasses import dataclass
-from functools import lru_cache
-import numpy as np
-from scipy import ndimage as ndi
-from skimage.measure import label
-
-from ..base import SenoQuantSpotDetector
-from senoquant.utils import layer_data_asarray
-
-
-BASE_KERNEL = np.array(
-    [1.0 / 16.0, 1.0 / 4.0, 3.0 / 8.0, 1.0 / 4.0, 1.0 / 16.0],
-    dtype=np.float32,
-)
-"""numpy.ndarray
-Base 1D B3-spline smoothing kernel (separable).
-
-Shape
------
-(5,)
-"""
-
-MAX_SCALES = 5
-"""int
-Maximum number of a trous scales.
-"""
-
-EPS = 1e-12
-"""float
-Small epsilon to avoid zero thresholds.
-"""
-
-
-@dataclass(frozen=True)
-class _Params:
-    """Internal parameter bundle.
-
-    Parameters
-    ----------
-    num_scales : int
-        Number of a trous scales (J), derived from enabled scales.
-    ld : float
-        Threshold on the multiscale product magnitude. Higher values reduce
-        detections (more conservative).
-    force_2d : bool
-        For 3D inputs, apply the 2D detector per slice (ignore Z correlations).
-    scale_enabled : tuple[bool, ...]
-        Per-scale enable flags for scales 1..MAX_SCALES (non-contiguous ok).
-    scale_sensitivity : tuple[float, ...]
-        Per-scale sensitivity for scales 1..MAX_SCALES (higher is more permissive).
-    """
-
-    num_scales: int = 3
-    ld: float = 1.0
-    force_2d: bool = False
-    scale_enabled: tuple[bool, ...] = (True, True, True)
-    scale_sensitivity: tuple[float, ...] = (100.0, 100.0, 100.0)
-
-
-def _min_size(num_scales: int) -> int:
-    """Minimum size per dimension for the requested number of scales.
-
-    Parameters
-    ----------
-    num_scales : int
-        Number of a trous scales (J).
-
-    Returns
-    -------
-    int
-        Minimum required size along each dimension.
-    """
-    # 5 + 4*2^(J-1)
-    return 5 + (2 ** (num_scales - 1)) * 4
-
-
-def _ensure_min_size(shape: tuple[int, ...], num_scales: int) -> None:
-    """Raise if any dimension is too small for the requested scales.
-
-    Parameters
-    ----------
-    shape : tuple[int, ...]
-        Image shape (2D or 3D).
-    num_scales : int
-        Number of a trous scales (J).
-    """
-    min_size = _min_size(num_scales)
-    if any(dim < min_size for dim in shape):
-        raise ValueError(
-            f"UDWT à trous requires each dimension >= {min_size} "
-            f"for {num_scales} scales (got shape={shape})."
-        )
-
-
-@lru_cache(maxsize=None)
-def _b3_kernel(step: int) -> np.ndarray:
-    """Return the dilated 1D B3-spline kernel for the given step.
-
-    Parameters
-    ----------
-    step : int
-        Step between non-zero taps (2**(j-1) for scale j).
-
-    Returns
-    -------
-    numpy.ndarray
-        1D kernel of length 1 + 4*step (float32).
-    """
-    if step <= 0:
-        raise ValueError("UDWT step must be positive.")
-    if step == 1:
-        return BASE_KERNEL
-    kernel = np.zeros(1 + 4 * step, dtype=np.float32)
-    kernel[::step] = BASE_KERNEL
-    return kernel
-
-
-def _atrous_smoothing_scales(image: np.ndarray, J: int) -> list[np.ndarray]:
-    """Compute a trous smoothing scales A1..AJ for a 2D/3D array.
-
-    Parameters
-    ----------
-    image : numpy.ndarray
-        Input 2D or 3D array.
-    J : int
-        Number of scales to compute.
-
-    Returns
-    -------
-    list[numpy.ndarray]
-        Smoothed arrays A1..AJ (float32), same shape as image.
-    """
-    scales: list[np.ndarray] = []
-    current = image.astype(np.float32, copy=False)
-
-    for j in range(1, J + 1):
-        step = 2 ** (j - 1)
-        kernel = _b3_kernel(step)
-
-        filtered = current
-        for axis in range(current.ndim):
-            filtered = ndi.convolve1d(filtered, kernel, axis=axis, mode="mirror")
-
-        scales.append(filtered.astype(np.float32, copy=False))
-        current = filtered
-
-    return scales
-
-
-def _wavelet_planes(A0: np.ndarray, scales: list[np.ndarray]) -> list[np.ndarray]:
-    """Compute detail planes Wi = A_{i-1} - Ai.
-
-    Parameters
-    ----------
-    A0 : numpy.ndarray
-        Original image.
-    scales : list[numpy.ndarray]
-        Smoothing scales A1..AJ.
-
-    Returns
-    -------
-    list[numpy.ndarray]
-        Detail planes W1..WJ (float32), same shape as A0.
-    """
-    W: list[np.ndarray] = []
-    prev = A0.astype(np.float32, copy=False)
-    for Ai in scales:
-        W.append((prev - Ai).astype(np.float32, copy=False))
-        prev = Ai
-    return W
-
-
-def _lambda_coeffs(num_scales: int, width: int, height: int) -> np.ndarray:
-    """Compute WAT scale-dependent lambda coefficients."""
-    lambdac = np.empty(num_scales + 2, dtype=np.float32)
-    for i in range(num_scales + 2):
-        denom = 1 << (2 * i)
-        val = (width * height) / denom
-        lambdac[i] = np.sqrt(2 * np.log(val)) if val > 0 else 0.0
-    return lambdac
-
-
-def _mean_abs_dev(data: np.ndarray, axis=None, keepdims: bool = False) -> np.ndarray:
-    """Mean absolute deviation from the mean."""
-    mean = data.mean(axis=axis, keepdims=True)
-    return np.mean(np.abs(data - mean), axis=axis, keepdims=keepdims)
-
-
-def _wat_threshold_inplace(
-    Wi: np.ndarray, scale_index: int, lambdac: np.ndarray, sensitivity: float
-) -> None:
-    """Apply WAT thresholding in-place for a single scale."""
-    dcoeff = max(sensitivity / 100.0, EPS)
-    if Wi.ndim == 2:
-        mad = _mean_abs_dev(Wi, axis=None, keepdims=False)
-        coeff_thr = (lambdac[scale_index + 1] * mad) / dcoeff
-        if coeff_thr <= EPS:
-            return
-        Wi[Wi < coeff_thr] = 0.0
-        return
-
-    mad = _mean_abs_dev(Wi, axis=(1, 2), keepdims=True)
-    coeff_thr = (lambdac[scale_index + 1] * mad) / dcoeff
-    Wi[Wi < coeff_thr] = 0.0
-
-
-def _multiscale_product(W_planes: list[np.ndarray]) -> np.ndarray:
-    """Multiply thresholded planes to form the multiscale product image.
-
-    Parameters
-    ----------
-    W_planes : list[numpy.ndarray]
-        Thresholded wavelet planes, all the same shape.
-
-    Returns
-    -------
-    numpy.ndarray
-        Multiscale product image (float32).
-    """
-    if not W_planes:
-        raise ValueError("W_planes must be a non-empty list.")
-    P = np.ones_like(W_planes[0], dtype=np.float32)
-    for Wi in W_planes:
-        P *= Wi.astype(np.float32, copy=False)
-    return P
-
-
-def _detect_from_product(P: np.ndarray, ld: float) -> np.ndarray:
-    """Binary mask where |P| > ld.
-
-    Parameters
-    ----------
-    P : numpy.ndarray
-        Multiscale product image.
-    ld : float
-        Detection threshold on |P|.
-
-    Returns
-    -------
-    numpy.ndarray
-        Boolean mask of detections.
-    """
-    return np.abs(P) > ld
-
-
-def _detect_2d(image: np.ndarray, params: _Params) -> np.ndarray:
-    """Detect spots in a 2D image and return labeled instances.
-
-    Parameters
-    ----------
-    image : numpy.ndarray
-        Input 2D image (Y, X).
-    params : _Params
-        Detector parameters.
-
-    Returns
-    -------
-    numpy.ndarray
-        Labeled instance mask (int32), background 0.
-    """
-    if params.scale_enabled:
-        num_scales = max(
-            (i + 1 for i, enabled in enumerate(params.scale_enabled) if enabled),
-            default=1,
-        )
-    else:
-        num_scales = max(1, params.num_scales)
-
-    _ensure_min_size(image.shape, num_scales)
-
-    scales = _atrous_smoothing_scales(image, num_scales)
-    W = _wavelet_planes(image, scales)
-    lambdac = _lambda_coeffs(num_scales, image.shape[1], image.shape[0])
-
-    enabled_indices = [
-        i for i, enabled in enumerate(params.scale_enabled) if enabled and i < len(W)
-    ]
-    if not enabled_indices:
-        return np.zeros(image.shape, dtype=np.int32)
-
-    enabled_planes: list[np.ndarray] = []
-    for i in enabled_indices:
-        sensitivity = (
-            params.scale_sensitivity[i]
-            if i < len(params.scale_sensitivity)
-            else 100.0
-        )
-        _wat_threshold_inplace(W[i], i, lambdac, sensitivity)
-        enabled_planes.append(W[i])
-
-    P = _multiscale_product(enabled_planes)
-    binary = _detect_from_product(P, params.ld)
-    return label(binary, connectivity=2).astype(np.int32, copy=False)
-
-
-def _detect_2d_stack(stack: np.ndarray, params: _Params) -> np.ndarray:
-    """Apply the 2D detector per Z-slice and keep labels unique.
-
-    Parameters
-    ----------
-    stack : numpy.ndarray
-        Input 3D stack (Z, Y, X).
-    params : _Params
-        Detector parameters.
-
-    Returns
-    -------
-    numpy.ndarray
-        Labeled mask for each slice (int32).
-    """
-    labels = np.zeros(stack.shape, dtype=np.int32)
-    next_label = 1
-
-    for z in range(stack.shape[0]):
-        slice_labels = _detect_2d(stack[z], params)
-        m = int(slice_labels.max())
-        if m > 0:
-            slice_labels = slice_labels + (next_label - 1)
-            next_label += m
-        labels[z] = slice_labels
-
-    return labels
-
-
-def _detect_3d(volume: np.ndarray, params: _Params) -> np.ndarray:
-    """Detect spots in a 3D volume using the direct 3D extension.
-
-    Parameters
-    ----------
-    volume : numpy.ndarray
-        Input 3D volume (Z, Y, X).
-    params : _Params
-        Detector parameters.
-
-    Returns
-    -------
-    numpy.ndarray
-        Labeled instance mask (int32).
-    """
-    if params.scale_enabled:
-        num_scales = max(
-            (i + 1 for i, enabled in enumerate(params.scale_enabled) if enabled),
-            default=1,
-        )
-    else:
-        num_scales = max(1, params.num_scales)
-
-    _ensure_min_size(volume.shape, num_scales)
-
-    scales = _atrous_smoothing_scales(volume, num_scales)
-    W = _wavelet_planes(volume, scales)
-    lambdac = _lambda_coeffs(num_scales, volume.shape[2], volume.shape[1])
-
-    enabled_indices = [
-        i for i, enabled in enumerate(params.scale_enabled) if enabled and i < len(W)
-    ]
-    if not enabled_indices:
-        return np.zeros(volume.shape, dtype=np.int32)
-
-    enabled_planes: list[np.ndarray] = []
-    for i in enabled_indices:
-        sensitivity = (
-            params.scale_sensitivity[i]
-            if i < len(params.scale_sensitivity)
-            else 100.0
-        )
-        _wat_threshold_inplace(W[i], i, lambdac, sensitivity)
-        enabled_planes.append(W[i])
-
-    P = _multiscale_product(enabled_planes)
-    binary = _detect_from_product(P, params.ld)
-    return label(binary, connectivity=3).astype(np.int32, copy=False)
-
-
-class UDWTDetector(SenoQuantSpotDetector):
-    """Undecimated B3-spline wavelet spot detector.
-
-    Settings: ld, force_2d, scale_*_enabled, scale_*_sensitivity.
-
-    Higher ld yields fewer detections (stricter thresholds).
-    Higher sensitivity yields more detections on that scale.
-    """
-
-    def __init__(self, models_root=None) -> None:
-        """Initialize the detector.
-
-        Parameters
-        ----------
-        models_root : pathlib.Path or None, optional
-            Root folder for detector resources (unused).
-        """
-        super().__init__("udwt", models_root=models_root)
-
-    def run(self, **kwargs) -> dict:
-        """Run the detector on a napari Image layer.
-
-        Parameters
-        ----------
-        **kwargs
-            layer : napari.layers.Image or None
-                Image layer (single-channel).
-            settings : dict, optional
-                Keys:
-                - ld (float): higher is stricter final detection threshold
-                - force_2d (bool): for 3D, detect per slice (ignores Z context)
-                - scale_N_enabled (bool): include scale N in the product
-                - scale_N_sensitivity (float): higher is more permissive
-
-        Returns
-        -------
-        dict
-            Output with keys "mask" (labeled int32 or None) and "points" (None).
-        """
-        layer = kwargs.get("layer")
-        if layer is None:
-            return {"mask": None, "points": None}
-
-        if getattr(layer, "rgb", False):
-            raise ValueError("UDWT requires single-channel images (rgb=False).")
-
-        settings = kwargs.get("settings", {}) or {}
-
-        ld = float(settings.get("ld", 1.0))
-        force_2d = bool(settings.get("force_2d", False))
-
-        default_enabled = (True, True, True, False, False)
-        enabled_all: list[bool] = []
-        sensitivity_all: list[float] = []
-        for i in range(MAX_SCALES):
-            enabled_key = f"scale_{i + 1}_enabled"
-            sens_key = f"scale_{i + 1}_sensitivity"
-            enabled_val = bool(settings.get(enabled_key, default_enabled[i]))
-            sens_val = float(settings.get(sens_key, 100.0))
-            sens_val = max(1.0, min(100.0, sens_val))
-            enabled_all.append(enabled_val)
-            sensitivity_all.append(sens_val)
-
-        if any(enabled_all):
-            num_scales = max(i + 1 for i, enabled in enumerate(enabled_all) if enabled)
-        else:
-            num_scales = 1
-
-        params = _Params(
-            num_scales=num_scales,
-            ld=ld,
-            force_2d=force_2d,
-            scale_enabled=tuple(enabled_all),
-            scale_sensitivity=tuple(sensitivity_all),
-        )
-
-        data = layer_data_asarray(layer)
-        if data.ndim not in (2, 3):
-            raise ValueError("UDWT expects 2D images or 3D stacks.")
-
-        image = np.asarray(data, dtype=np.float32)
-
-        if image.ndim == 2:
-            return {"mask": _detect_2d(image, params), "points": None}
-
-        # 3D input
-        if params.force_2d:
-            return {"mask": _detect_2d_stack(image, params), "points": None}
-
-        return {"mask": _detect_3d(image, params), "points": None}

senoquant-1.0.0b2.dist-info/METADATA

@@ -1,193 +0,0 @@
-Metadata-Version: 2.4
-Name: senoquant
-Version: 1.0.0b2
-Summary: Napari plugin for spatial quantification of senescence markers in tissue imaging
-Author: SenoQuant Contributors
-Maintainer: SenoQuant Contributors
-License: BSD-3-Clause
-Project-URL: Homepage, https://github.com/HaamsRee/senoquant-dev
-Project-URL: Documentation, https://haamsree.github.io/senoquant-dev/
-Project-URL: Repository, https://github.com/HaamsRee/senoquant-dev
-Project-URL: Bug Tracker, https://github.com/HaamsRee/senoquant-dev/issues
-Keywords: napari,plugin,senescence,quantification,microscopy,image analysis,segmentation,spot detection
-Classifier: Development Status :: 4 - Beta
-Classifier: Framework :: napari
-Classifier: Intended Audience :: Science/Research
-Classifier: License :: OSI Approved :: BSD License
-Classifier: Operating System :: OS Independent
-Classifier: Programming Language :: Python :: 3
-Classifier: Programming Language :: Python :: 3.11
-Classifier: Programming Language :: Python :: 3.12
-Classifier: Topic :: Scientific/Engineering :: Bio-Informatics
-Classifier: Topic :: Scientific/Engineering :: Image Processing
-Requires-Python: >=3.11
-Description-Content-Type: text/markdown
-License-File: LICENSE
-Requires-Dist: bioio>=3.2.0
-Requires-Dist: bioio-czi>=2.4.2
-Requires-Dist: bioio-dv>=1.2.0
-Requires-Dist: bioio-imageio>=1.3.0
-Requires-Dist: bioio-lif>=1.4.0
-Requires-Dist: bioio-nd2>=1.6.0
-Requires-Dist: bioio-ome-tiff>=1.4.0
-Requires-Dist: bioio-ome-zarr>=3.2.1
-Requires-Dist: bioio-sldy>=1.4.0
-Requires-Dist: bioio-tifffile>=1.3.0
-Requires-Dist: bioio-tiff-glob>=1.2.0
-Requires-Dist: numpy<=1.26.4,>=1.23
-Requires-Dist: cellpose==4.0.8
-Requires-Dist: onnx>=1.16
-Requires-Dist: onnxruntime>=1.16; platform_system == "Darwin"
-Requires-Dist: onnxruntime-gpu>=1.16; platform_system != "Darwin"
-Requires-Dist: openpyxl>=3.1
-Requires-Dist: huggingface_hub>=0.23.0
-Requires-Dist: scikit-image<0.25,>=0.22
-Requires-Dist: scipy>=1.8
-Requires-Dist: senoquant-stardist-ext>=0.1.0
-Requires-Dist: dask[array]>=2024.4
-Requires-Dist: dask[distributed]>=2024.4
-Provides-Extra: gpu
-Requires-Dist: cupy-cuda12x>=12.0; extra == "gpu"
-Requires-Dist: cucim>=23.4; extra == "gpu"
-Provides-Extra: all
-Requires-Dist: napari[all]; extra == "all"
-Requires-Dist: bioio>=3.2.0; extra == "all"
-Requires-Dist: bioio-czi>=2.4.2; extra == "all"
-Requires-Dist: bioio-dv>=1.2.0; extra == "all"
-Requires-Dist: bioio-imageio>=1.3.0; extra == "all"
-Requires-Dist: bioio-lif>=1.4.0; extra == "all"
-Requires-Dist: bioio-nd2>=1.6.0; extra == "all"
-Requires-Dist: bioio-ome-tiff>=1.4.0; extra == "all"
-Requires-Dist: bioio-ome-zarr>=3.2.1; extra == "all"
-Requires-Dist: bioio-sldy>=1.4.0; extra == "all"
-Requires-Dist: bioio-tifffile>=1.3.0; extra == "all"
-Requires-Dist: bioio-tiff-glob>=1.2.0; extra == "all"
-Requires-Dist: numpy<=1.26.4,>=1.23; extra == "all"
-Requires-Dist: cellpose==4.0.8; extra == "all"
-Requires-Dist: onnx>=1.16; extra == "all"
-Requires-Dist: onnxruntime>=1.16; platform_system == "Darwin" and extra == "all"
-Requires-Dist: onnxruntime-gpu>=1.16; platform_system != "Darwin" and extra == "all"
-Requires-Dist: openpyxl>=3.1; extra == "all"
-Requires-Dist: huggingface_hub>=0.23.0; extra == "all"
-Requires-Dist: scikit-image<0.25,>=0.22; extra == "all"
-Requires-Dist: scipy>=1.8; extra == "all"
-Requires-Dist: senoquant-stardist-ext>=0.1.0; extra == "all"
-Requires-Dist: dask[array]>=2024.4; extra == "all"
-Requires-Dist: dask[distributed]>=2024.4; extra == "all"
-Requires-Dist: cupy-cuda12x>=12.0; extra == "all"
-Requires-Dist: cucim>=23.4; extra == "all"
-Dynamic: license-file
-
-# SenoQuant
-
-![tests](https://github.com/HaamsRee/senoquant-dev/actions/workflows/tests.yml/badge.svg)
-[![PyPI version](https://badge.fury.io/py/senoquant.svg)](https://badge.fury.io/py/senoquant)
-[![Python 3.11+](https://img.shields.io/badge/python-3.11+-blue.svg)](https://www.python.org/downloads/)
-[![License](https://img.shields.io/badge/License-BSD_3--Clause-blue.svg)](https://opensource.org/licenses/BSD-3-Clause)
-
-SenoQuant is a versatile Napari plugin designed for comprehensive, accurate,
-and unbiased spatial quantification and prediction of senescence markers
-across diverse tissue contexts.
-
-## Features
-
-- **Multi-Model Segmentation**: Nuclear and cytoplasmic segmentation with 5 built-in models
-  - StarDist ONNX (2D/3D)
-  - Cellpose SAM
-  - Morphological operations (dilation, perinuclear rings)
-- **Spot Detection**: Detect and quantify punctate senescence markers
-  - Undecimated B3-spline wavelet (UDWT)
-  - Rotational morphological processing (RMP)
-- **Quantification**: Extract intensity, morphology, and spot metrics
-  - Per-cell marker intensities
-  - Morphological descriptors
-  - Spot counting and colocalization
-- **Batch Processing**: Automated analysis of entire image folders
-  - Profile save/load for reproducibility
-  - Multi-scene file support
-- **File Format Support**: Microscopy formats via BioIO
-  - OME-TIFF, ND2, LIF, CZI, Zarr, and more
-
-## Installation
-
-### Prerequisites
-
-SenoQuant requires Python 3.11+ and napari:
-
-```bash
-conda create -n senoquant python=3.11
-conda activate senoquant
-pip install "napari[all]"
-```
-
-Or using `uv` (faster installer):
-
-```bash
-conda create -n senoquant python=3.11
-conda activate senoquant
-pip install uv
-uv pip install "napari[all]"
-```
-
-### Install SenoQuant
-
-```bash
-pip install senoquant
-```
-
-Or with `uv`:
-
-```bash
-uv pip install senoquant
-```
-
-Model files are downloaded automatically on first use from Hugging Face.
-To override the model repository, set `SENOQUANT_MODEL_REPO` environment variable.
-
-For GPU acceleration (Windows/Linux with CUDA):
-
-```bash
-pip install senoquant[gpu]
-```
-
-**Note:** The first launch of napari and the SenoQuant plugin will be slower as napari initializes and SenoQuant downloads model files (~1.3 GB) from Hugging Face. Subsequent launches will be faster as models are cached locally.
-
-## Quick Start
-
-1. **Launch napari and open your image:**
-   ```bash
-   napari
-   ```
-   File → Open File(s)... → Select your image
-
-2. **Open SenoQuant plugin:**  
-   Plugins → SenoQuant
-
-3. **Run segmentation:**  
-   Segmentation tab → Select nuclear channel → Choose model → Run
-
-4. **Detect spots (optional):**  
-   Spots tab → Select channel → Choose detector → Run
-
-5. **Export quantification:**  
-   Quantification tab → Configure features → Export
-
-6. **Batch process (optional):**  
-   Batch tab → Configure settings → Run Batch
-
-## Documentation
-
-Full documentation is available at [https://haamsree.github.io/senoquant-dev/](https://haamsree.github.io/senoquant-dev/)
-
-- [Installation Guide](https://haamsree.github.io/senoquant-dev/user/installation/)
-- [Quick Start Tutorial](https://haamsree.github.io/senoquant-dev/user/quickstart/)
-- [Segmentation Models](https://haamsree.github.io/senoquant-dev/user/segmentation/)
-- [Spot Detection](https://haamsree.github.io/senoquant-dev/user/spots/)
-- [Quantification Features](https://haamsree.github.io/senoquant-dev/user/quantification/)
-- [Batch Processing](https://haamsree.github.io/senoquant-dev/user/batch/)
-- [API Reference](https://haamsree.github.io/senoquant-dev/api/)
-
-## Development
-
-See the [Contributing Guide](https://haamsree.github.io/senoquant-dev/developer/contributing/) for development setup instructions.
-