TwoPy 0.1.9__tar.gz → 0.2.1__tar.gz

{twopy-0.1.9 → twopy-0.2.1}/PKG-INFO

@@ -1,9 +1,11 @@
 Metadata-Version: 2.4
 Name: twopy
-Version: 0.1.9
+Version: 0.2.1
 Summary: Two-photon imaging analysis tool with a napari interface.
 Author-email: Gustavo Madeira Santana <fromweb1@gumadeiras.com>
 License-Expression: MIT
+Project-URL: Homepage, https://twopy.gumadeiras.com
+Project-URL: Repository, https://github.com/gumadeiras/twopy
 Requires-Python: <3.14,>=3.13
 Description-Content-Type: text/markdown
 Requires-Dist: h5py<4,>=3.16
@@ -73,8 +75,11 @@ When a source recording has not been converted yet, twopy converts it first. Con
 
 ## Docs
 
-- [GUI guide](docs/gui.md): napari loading, ROI editing, plotting, and saved outputs.
-- [Python API guide](docs/python_api.md): find recordings, convert data, run analysis, and open napari from scripts.
+- [Full documentation](https://twopy.gumadeiras.com)
+- [Getting started](docs/getting_started.md): install, configure, launch, and save your first analysis.
+- [GUI guide](docs/gui/index.md): one page per task — loading, ROIs, plots, custom workflows, group matching, saving.
+- [Python API guide](docs/python_api.md): find recordings, convert data, extract traces, fit kernels, and open napari from scripts.
+- [Writing custom workflows](docs/writing_custom_workflows.md): add your own analysis to the napari Custom tab.
 - [Development guide](docs/development.md): source setup and local checks.
 - [Input data spec](docs/input_data_spec.md): short recording folder contract.
 - [Recording file reference](docs/recording_file_schema.md): detailed source and converted file schema.

{twopy-0.1.9 → twopy-0.2.1}/README.md

@@ -53,8 +53,11 @@ When a source recording has not been converted yet, twopy converts it first. Con
 
 ## Docs
 
-- [GUI guide](docs/gui.md): napari loading, ROI editing, plotting, and saved outputs.
-- [Python API guide](docs/python_api.md): find recordings, convert data, run analysis, and open napari from scripts.
+- [Full documentation](https://twopy.gumadeiras.com)
+- [Getting started](docs/getting_started.md): install, configure, launch, and save your first analysis.
+- [GUI guide](docs/gui/index.md): one page per task — loading, ROIs, plots, custom workflows, group matching, saving.
+- [Python API guide](docs/python_api.md): find recordings, convert data, extract traces, fit kernels, and open napari from scripts.
+- [Writing custom workflows](docs/writing_custom_workflows.md): add your own analysis to the napari Custom tab.
 - [Development guide](docs/development.md): source setup and local checks.
 - [Input data spec](docs/input_data_spec.md): short recording folder contract.
 - [Recording file reference](docs/recording_file_schema.md): detailed source and converted file schema.

{twopy-0.1.9 → twopy-0.2.1}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "twopy"
-version = "0.1.9"
+version = "0.2.1"
 description = "Two-photon imaging analysis tool with a napari interface."
 readme = "README.md"
 requires-python = ">=3.13,<3.14"
@@ -29,6 +29,10 @@ dependencies = [
 [project.scripts]
 twopy = "twopy.napari:main"
 
+[project.urls]
+Homepage = "https://twopy.gumadeiras.com"
+Repository = "https://github.com/gumadeiras/twopy"
+
 [tool.setuptools.packages.find]
 where = ["src"]
 
@@ -43,11 +47,15 @@ target-version = "py313"
 [tool.ruff.lint]
 select = [
   "ANN",
+  "ARG",
   "B",
+  "C4",
   "D",
   "E",
   "F",
   "I",
+  "NPY",
+  "PTH",
   "SIM",
   "UP",
   "W",

{twopy-0.1.9 → twopy-0.2.1}/src/twopy/__init__.py

@@ -22,12 +22,17 @@ from twopy.analysis import (
     FrameWindow,
     GroupedRoiResponses,
     GroupedRoiResponseSummary,
-    InterpolatedEpochMapping,
+    Hemisphere,
+    KernelFitMethod,
+    KernelStimulusModality,
     LoadedAnalysisOutputs,
     LowPassFilterOptions,
     ManualFovGroupRow,
+    ManualRoiMatchGroup,
     ManualRoiMatchRow,
     ManualRoiMatchStatus,
+    RecordingKernelFit,
+    RecordingTiming,
     ResponseMapData,
     ResponseMapMode,
     ResponseMapOptions,
@@ -37,7 +42,10 @@ from twopy.analysis import (
     RoiResponseSummary,
     RoiResponseTrial,
     SmoothingOptions,
+    StimulusKernelOptions,
+    TimingSource,
     WindowedRoiResponse,
+    add_manual_roi_match_group,
     analyze_recording_responses,
     append_manual_roi_match_rows,
     apply_motion_artifact_mask_to_delta_f_over_f,
@@ -45,11 +53,13 @@ from twopy.analysis import (
     compute_recording_responses,
     compute_roi_delta_f_over_f,
     default_baseline_epoch_number,
+    default_kernel_stimulus_column,
     default_recording_baseline_epoch_number,
     extract_background_corrected_roi_traces,
-    frame_windows_from_photodiode_alignment,
+    finite_mean_and_sem,
+    fit_recording_stimulus_kernels,
+    fit_stimulus_kernel,
     group_delta_f_over_f_by_epoch,
-    interpolate_stimulus_epochs_to_frame_windows,
     is_baseline_epoch_name,
     load_analysis_outputs,
     load_manual_fov_group_rows,
@@ -58,10 +68,13 @@ from twopy.analysis import (
     make_frame_windows,
     make_manual_fov_group_rows,
     make_manual_roi_match_rows,
-    map_stimulus_epochs_to_frame_windows,
+    matched_manual_roi_groups,
     next_group_cell_id,
     no_baseline_epoch_frame_windows,
+    remove_manual_roi_match_group,
+    replace_manual_roi_match_group,
     resolve_baseline_frame_windows,
+    resolve_recording_timing,
     save_analysis_outputs,
     save_manual_fov_group_rows,
     save_manual_roi_match_rows,
@@ -76,7 +89,11 @@ from twopy.analysis import (
 )
 from twopy.api import find_recordings
 from twopy.conversion import convert_recording_to_twopy, load_source_conversion_inputs
-from twopy.converted import load_converted_recording, recording_frame_rate_hz
+from twopy.converted import (
+    load_converted_recording,
+    recording_frame_rate_hz,
+    recording_hemisphere,
+)
 from twopy.napari import (
     NapariRecordingView,
     NapariSidebarWidgets,
@@ -161,6 +178,7 @@ __all__ = [
     "AlignedPhotodiodeEvent",
     "AnalysisResponseRun",
     "AnalysisResponseComputation",
+    "add_manual_roi_match_group",
     "analyze_recording_responses",
     "add_twopy_magicgui_controls",
     "append_manual_roi_match_rows",
@@ -181,6 +199,7 @@ __all__ = [
     "DEFAULT_PIXEL_CALIBRATION_PATH",
     "DEFAULT_PIXEL_CALIBRATION_PROFILE_PATH",
     "default_baseline_epoch_number",
+    "default_kernel_stimulus_column",
     "default_recording_baseline_epoch_number",
     "detect_photodiode_events",
     "detect_recording_photodiode_events",
@@ -194,18 +213,20 @@ __all__ = [
     "extract_rois_from_image",
     "extract_roi_traces",
     "find_recordings",
-    "frame_windows_from_photodiode_alignment",
+    "fit_recording_stimulus_kernels",
+    "fit_stimulus_kernel",
     "FrameWindow",
     "full_frame_crop",
     "group_delta_f_over_f_by_epoch",
     "GroupedRoiResponseSummary",
     "GroupedRoiResponses",
+    "Hemisphere",
     "grid_roi_set",
     "grid_roi_set_microns",
     "grid_size_pixels_from_microns",
-    "InterpolatedEpochMapping",
-    "interpolate_stimulus_epochs_to_frame_windows",
     "is_baseline_epoch_name",
+    "KernelFitMethod",
+    "KernelStimulusModality",
     "launch_napari",
     "load_analysis_outputs",
     "load_response_map_data",
@@ -219,16 +240,19 @@ __all__ = [
     "LoadedAnalysisOutputs",
     "LowPassFilterOptions",
     "ManualFovGroupRow",
+    "ManualRoiMatchGroup",
     "ManualRoiMatchRow",
     "ManualRoiMatchStatus",
+    "matched_manual_roi_groups",
     "make_roi_set_from_label_image",
     "make_frame_windows",
     "make_manual_fov_group_rows",
     "make_manual_roi_match_rows",
     "make_roi_set",
-    "map_stimulus_epochs_to_frame_windows",
     "next_group_cell_id",
     "no_baseline_epoch_frame_windows",
+    "remove_manual_roi_match_group",
+    "replace_manual_roi_match_group",
     "map_stimulus_specific_column",
     "NapariRecordingView",
     "NapariSidebarWidgets",
@@ -255,10 +279,14 @@ __all__ = [
     "ResponseStatistic",
     "ResponseWatershedExtraction",
     "ResponseWatershedScoreImages",
+    "RecordingKernelFit",
+    "RecordingTiming",
     "recording_frame_rate_hz",
+    "recording_hemisphere",
     "resolve_pixel_size_um",
     "resolve_pixel_calibration_profile",
     "resolve_baseline_frame_windows",
+    "resolve_recording_timing",
     "response_watershed_roi_set",
     "RoiExtractionConfig",
     "RoiExtractionMethod",
@@ -282,10 +310,13 @@ __all__ = [
     "SmoothingOptions",
     "SpatialCrop",
     "SpatialDomain",
+    "StimulusKernelOptions",
     "StimulusSpecificColumnMapping",
     "stimulus_epoch_names_by_number",
+    "finite_mean_and_sem",
     "summarize_epoch_roi_responses",
     "summarize_grouped_responses",
+    "TimingSource",
     "validate_grouped_roi_responses",
     "watershed_roi_set",
     "WindowedRoiResponse",

{twopy-0.1.9 → twopy-0.2.1}/src/twopy/analysis/__init__.py

@@ -19,23 +19,34 @@ from twopy.analysis.dff import (
     compute_roi_delta_f_over_f,
 )
 from twopy.analysis.dff_options import DeltaFOverFBaselineMode, DeltaFOverFOptions
-from twopy.analysis.epoch_mapping import (
-    InterpolatedEpochMapping,
-    interpolate_stimulus_epochs_to_frame_windows,
-)
 from twopy.analysis.group_matching import (
     ManualFovGroupRow,
+    ManualRoiMatchGroup,
     ManualRoiMatchRow,
     ManualRoiMatchStatus,
+    add_manual_roi_match_group,
     append_manual_roi_match_rows,
     load_manual_fov_group_rows,
     load_manual_roi_match_rows,
     make_manual_fov_group_rows,
     make_manual_roi_match_rows,
+    matched_manual_roi_groups,
     next_group_cell_id,
+    remove_manual_roi_match_group,
+    replace_manual_roi_match_group,
     save_manual_fov_group_rows,
     save_manual_roi_match_rows,
 )
+from twopy.analysis.kernels import (
+    Hemisphere,
+    KernelFitMethod,
+    KernelStimulusModality,
+    RecordingKernelFit,
+    StimulusKernelOptions,
+    default_kernel_stimulus_column,
+    fit_recording_stimulus_kernels,
+    fit_stimulus_kernel,
+)
 from twopy.analysis.motion import apply_motion_artifact_mask_to_delta_f_over_f
 from twopy.analysis.persistence import (
     ANALYSIS_OUTPUT_FILE_FORMAT,
@@ -67,20 +78,24 @@ from twopy.analysis.responses import (
     GroupedRoiResponseSummary,
     RoiResponseSummary,
     RoiResponseTrial,
+    finite_mean_and_sem,
     group_delta_f_over_f_by_epoch,
     summarize_epoch_roi_responses,
     summarize_grouped_responses,
     validate_grouped_roi_responses,
 )
+from twopy.analysis.timing import (
+    RecordingTiming,
+    TimingSource,
+    resolve_recording_timing,
+)
 from twopy.analysis.trials import (
     EpochFrameWindow,
     FrameWindow,
     WindowedRoiResponse,
     default_baseline_epoch_number,
-    frame_windows_from_photodiode_alignment,
     is_baseline_epoch_name,
     make_frame_windows,
-    map_stimulus_epochs_to_frame_windows,
     no_baseline_epoch_frame_windows,
     resolve_baseline_frame_windows,
     select_baseline_frame_windows,
@@ -117,27 +132,33 @@ __all__ = [
     "EpochResponseMap",
     "EpochFrameWindow",
     "extract_background_corrected_roi_traces",
-    "frame_windows_from_photodiode_alignment",
+    "fit_recording_stimulus_kernels",
+    "fit_stimulus_kernel",
+    "default_kernel_stimulus_column",
+    "Hemisphere",
     "FrameWindow",
     "group_delta_f_over_f_by_epoch",
     "GroupedRoiResponseSummary",
     "GroupedRoiResponses",
-    "InterpolatedEpochMapping",
-    "interpolate_stimulus_epochs_to_frame_windows",
     "is_baseline_epoch_name",
+    "KernelFitMethod",
+    "KernelStimulusModality",
     "load_analysis_outputs",
     "load_response_map_data",
     "LoadedAnalysisOutputs",
     "LowPassFilterOptions",
     "ManualFovGroupRow",
+    "ManualRoiMatchGroup",
     "ManualRoiMatchRow",
     "ManualRoiMatchStatus",
+    "matched_manual_roi_groups",
     "make_frame_windows",
     "make_manual_fov_group_rows",
     "make_manual_roi_match_rows",
-    "map_stimulus_epochs_to_frame_windows",
     "no_baseline_epoch_frame_windows",
     "next_group_cell_id",
+    "RecordingKernelFit",
+    "RecordingTiming",
     "RoiDeltaFOverF",
     "ResponseMapData",
     "ResponseMapMode",
@@ -147,18 +168,25 @@ __all__ = [
     "RoiResponseSummary",
     "RoiResponseTrial",
     "resolve_baseline_frame_windows",
+    "resolve_recording_timing",
+    "add_manual_roi_match_group",
     "append_manual_roi_match_rows",
     "load_manual_fov_group_rows",
     "load_manual_roi_match_rows",
+    "remove_manual_roi_match_group",
+    "replace_manual_roi_match_group",
     "save_analysis_outputs",
     "save_response_map_data",
     "save_manual_fov_group_rows",
     "save_manual_roi_match_rows",
     "select_baseline_frame_windows",
+    "finite_mean_and_sem",
     "split_traces_by_frame_windows",
+    "StimulusKernelOptions",
     "SmoothingOptions",
     "summarize_epoch_roi_responses",
     "summarize_grouped_responses",
+    "TimingSource",
     "validate_grouped_roi_responses",
     "WindowedRoiResponse",
     "write_response_summary_grouped_csv",

{twopy-0.1.9 → twopy-0.2.1}/src/twopy/analysis/background_subtraction.py

@@ -15,6 +15,7 @@ from typing import Literal
 
 import numpy as np
 import numpy.typing as npt
+from scipy import sparse
 
 from twopy.converted import RecordingData
 from twopy.frame_ranges import normalize_frame_range
@@ -278,6 +279,7 @@ def _extract_mean_traces_from_indices(
     rather than total recording length.
     """
     traces = np.empty((frame_stop - frame_start, len(pixel_indices_by_trace)))
+    projection = _mean_projection(pixel_indices_by_trace)
     _check_cancelled(check_cancelled)
     for chunk_start, chunk_stop, frames in recording.movie.iter_frame_batches(
         chunk_frames=chunk_frames,
@@ -289,10 +291,7 @@ def _extract_mean_traces_from_indices(
         chunk_offset = chunk_start - frame_start
         chunk_slice = slice(chunk_offset, chunk_offset + (chunk_stop - chunk_start))
         chunk_flat = frames.reshape(frames.shape[0], -1)
-        for trace_index, pixel_indices in enumerate(pixel_indices_by_trace):
-            traces[chunk_slice, trace_index] = chunk_flat[:, pixel_indices].mean(
-                axis=1,
-            )
+        traces[chunk_slice, :] = _project_trace_pixels(chunk_flat, projection)
         _check_cancelled(check_cancelled)
 
     return traces
@@ -344,6 +343,7 @@ def _extract_movie_global_percentile_corrected_traces(
     )
     frame_count = frame_stop - frame_start
     roi_count = len(roi_pixel_indices)
+    projection = _mean_projection(roi_pixel_indices)
     raw_values = np.empty((frame_count, roi_count), dtype=np.float64)
     background_values = np.empty((frame_count, roi_count), dtype=np.float64)
     corrected_values = np.empty((frame_count, roi_count), dtype=np.float64)
@@ -359,14 +359,19 @@ def _extract_movie_global_percentile_corrected_traces(
         chunk_slice = slice(chunk_offset, chunk_offset + (chunk_stop - chunk_start))
         chunk_flat = frames.reshape(frames.shape[0], -1)
         background_trace = chunk_flat[:, background_indices].mean(axis=1)
-        for roi_index, pixel_indices in enumerate(roi_pixel_indices):
-            roi_pixels = chunk_flat[:, pixel_indices]
-            raw_values[chunk_slice, roi_index] = roi_pixels.mean(axis=1)
-            background_values[chunk_slice, roi_index] = background_trace
-            # Clamp pixels before averaging so dim pixels cannot pull the
-            # corrected ROI trace below zero.
-            corrected_pixels = np.maximum(roi_pixels - background_trace[:, None], 0.0)
-            corrected_values[chunk_slice, roi_index] = corrected_pixels.mean(axis=1)
+        roi_pixels = chunk_flat[:, projection.pixel_indices]
+        raw_values[chunk_slice, :] = _project_selected_pixels(roi_pixels, projection)
+        background_values[chunk_slice, :] = background_trace[:, np.newaxis]
+        # Clamp pixels before averaging so dim pixels cannot pull the
+        # corrected ROI trace below zero.
+        corrected_pixels = np.maximum(
+            roi_pixels - background_trace[:, np.newaxis],
+            0.0,
+        )
+        corrected_values[chunk_slice, :] = _project_selected_pixels(
+            corrected_pixels,
+            projection,
+        )
         _check_cancelled(check_cancelled)
 
     metadata: dict[str, BackgroundMetadataValue] = {
@@ -440,6 +445,7 @@ def _extract_movie_y_stripe_percentile_corrected_traces(
     )
     frame_count = frame_stop - frame_start
     roi_count = len(roi_pixel_indices)
+    projection = _mean_projection(roi_pixel_indices)
     raw_values = np.empty((frame_count, roi_count), dtype=np.float64)
     background_values = np.empty((frame_count, roi_count), dtype=np.float64)
     corrected_values = np.empty((frame_count, roi_count), dtype=np.float64)
@@ -460,15 +466,21 @@ def _extract_movie_y_stripe_percentile_corrected_traces(
         )
         chunk_flat = frames.reshape(frames.shape[0], -1)
         background_flat = background_field.reshape(background_field.shape[0], -1)
-        for roi_index, pixel_indices in enumerate(roi_pixel_indices):
-            roi_pixels = chunk_flat[:, pixel_indices]
-            roi_background_pixels = background_flat[:, pixel_indices]
-            raw_values[chunk_slice, roi_index] = roi_pixels.mean(axis=1)
-            background_values[chunk_slice, roi_index] = roi_background_pixels.mean(
-                axis=1,
-            )
-            corrected_pixels = np.maximum(roi_pixels - roi_background_pixels, 0.0)
-            corrected_values[chunk_slice, roi_index] = corrected_pixels.mean(axis=1)
+        roi_pixels = chunk_flat[:, projection.pixel_indices]
+        roi_background_pixels = background_flat[:, projection.pixel_indices]
+        raw_values[chunk_slice, :] = _project_selected_pixels(roi_pixels, projection)
+        background_values[chunk_slice, :] = _project_selected_pixels(
+            roi_background_pixels,
+            projection,
+        )
+        corrected_pixels = np.maximum(
+            roi_pixels - roi_background_pixels,
+            0.0,
+        )
+        corrected_values[chunk_slice, :] = _project_selected_pixels(
+            corrected_pixels,
+            projection,
+        )
         _check_cancelled(check_cancelled)
 
     metadata: dict[str, BackgroundMetadataValue] = {
@@ -593,6 +605,93 @@ def _roi_pixel_indices_from_masks(
     return tuple(np.flatnonzero(mask).astype(np.int64) for mask in flat_masks)
 
 
+@dataclass(frozen=True)
+class _MeanProjection:
+    """Sparse averaging matrix plus selected flattened pixel indices."""
+
+    pixel_indices: npt.NDArray[np.int64]
+    matrix: sparse.csc_matrix
+
+
+def _mean_projection(
+    pixel_indices_by_trace: tuple[npt.NDArray[np.int64], ...],
+) -> _MeanProjection:
+    """Return selected pixels and a sparse pixel-to-trace averaging matrix.
+
+    Args:
+        pixel_indices_by_trace: One flattened pixel-index array per trace.
+
+    Returns:
+        Selected flattened pixel indices plus a sparse matrix shaped
+        ``(selected_pixels, traces)`` whose columns average each trace.
+
+    Pixel indices are concatenated per trace instead of deduplicated. That
+    preserves the existing behavior for overlapping ROIs, where the same movie
+    pixel can contribute independently to more than one ROI mean.
+    """
+    selected_pixel_indices: list[npt.NDArray[np.int64]] = []
+    rows: list[npt.NDArray[np.int64]] = []
+    columns: list[npt.NDArray[np.int64]] = []
+    values: list[npt.NDArray[np.float64]] = []
+    row_start = 0
+    for trace_index, pixel_indices in enumerate(pixel_indices_by_trace):
+        if pixel_indices.size == 0:
+            continue
+        selected_pixel_indices.append(pixel_indices)
+        rows.append(
+            np.arange(
+                row_start,
+                row_start + pixel_indices.size,
+                dtype=np.int64,
+            ),
+        )
+        row_start += pixel_indices.size
+        columns.append(
+            np.full(pixel_indices.shape, trace_index, dtype=np.int64),
+        )
+        values.append(
+            np.full(
+                pixel_indices.shape,
+                1.0 / float(pixel_indices.size),
+                dtype=np.float64,
+            ),
+        )
+    if len(selected_pixel_indices) == 0:
+        return _MeanProjection(
+            pixel_indices=np.array([], dtype=np.int64),
+            matrix=sparse.csc_matrix((0, len(pixel_indices_by_trace))),
+        )
+    return _MeanProjection(
+        pixel_indices=np.concatenate(selected_pixel_indices),
+        matrix=sparse.csc_matrix(
+            (
+                np.concatenate(values),
+                (np.concatenate(rows), np.concatenate(columns)),
+            ),
+            shape=(row_start, len(pixel_indices_by_trace)),
+        ),
+    )
+
+
+def _project_trace_pixels(
+    chunk_flat: npt.NDArray[np.float64],
+    projection: _MeanProjection,
+) -> npt.NDArray[np.float64]:
+    """Return trace means from a flattened movie chunk and projection."""
+    return _project_selected_pixels(chunk_flat[:, projection.pixel_indices], projection)
+
+
+def _project_selected_pixels(
+    selected_pixels: npt.NDArray[np.float64],
+    projection: _MeanProjection,
+) -> npt.NDArray[np.float64]:
+    """Return trace means from selected pixel columns."""
+    return np.asarray(
+        selected_pixels @ projection.matrix,
+        dtype=np.float64,
+    )
+
+
 def _check_cancelled(check_cancelled: Callable[[], None] | None) -> None:
     """Raise ``CancelledError`` when an interactive caller superseded this work."""
     if check_cancelled is None:

{twopy-0.1.9 → twopy-0.2.1}/src/twopy/analysis/dff.py

@@ -115,6 +115,10 @@ def compute_roi_delta_f_over_f(
         data_rate_hz=data_rate_hz,
         baseline_sample_seconds=baseline_sample_seconds,
     )
+    _validate_baseline_samples_for_background_method(
+        traces=traces,
+        baseline_samples=baseline_samples,
+    )
     tau = _fit_shared_tau(
         frame_numbers=baseline_samples.frame_numbers,
         mean_fluorescence=baseline_samples.values.mean(axis=1),
@@ -274,6 +278,35 @@ def _baseline_samples(
     )
 
 
+def _validate_baseline_samples_for_background_method(
+    *,
+    traces: BackgroundCorrectedRoiTraces,
+    baseline_samples: _BaselineSamples,
+) -> None:
+    """Reject background-corrected baselines that cannot support dF/F."""
+    if traces.method != "roi_y_stripe_percentile":
+        return
+
+    invalid = ~np.isfinite(baseline_samples.values) | (baseline_samples.values <= 0.0)
+    if not np.any(invalid):
+        return
+
+    roi_indices = np.flatnonzero(np.any(invalid, axis=0))
+    labels = ", ".join(traces.labels[index] for index in roi_indices[:5])
+    if roi_indices.size > 5:
+        labels = f"{labels}, ..."
+    minimum = float(np.nanmin(baseline_samples.values[:, roi_indices]))
+    msg = (
+        "ROI y-stripe P% background subtraction over-subtracted baseline "
+        f"fluorescence for {labels}; minimum corrected baseline sample is "
+        f"{minimum:.6g}. dF/F needs positive baseline fluorescence, so the "
+        "local y-stripe pixels are not a valid additive background for those "
+        "ROIs. Use shared y-stripe P%, global percentile, or revise the ROIs "
+        "to leave dim local background pixels."
+    )
+    raise ValueError(msg)
+
+
 def _validate_baseline_local_range(
     *,
     window: FrameWindow,