Generated API

Return a shallow copy with a processing function appended to its queue.

The function is called as fn(intensity, mz, **fn_kwargs). It may return a new intensity vector or a pair (new_mz, new_intensity).

Apply queued processing steps and return a new dataset.

Each step is applied to every spectrum in queue order. Shared-domain data must retain the same m/z axis across all pixels. File output is deferred to the I/O phase and is not accepted here.

Return a shallow copy with all pending processing steps removed.

Queue baseline estimation and subtraction for every spectrum.

Methods are locmin, hull, snip, and median. The returned dataset is a copy with a deferred step; call process to apply it. Output intensities are clipped at zero after baseline subtraction.

Eagerly bin spectra onto shared m/z values.

Aggregation methods sum, mean, max, and min are supported. Interpolation names from the R API are recognized but not yet implemented.

Estimate a regular shared domain from per-spectrum m/z vectors.

Return the shared m/z axis or estimate one from ragged spectra.

Detect reference peaks from the mean spectrum of a dataset.

Apply queued steps and align detected peaks to a shared sparse matrix.

When ref is omitted, local maxima are detected, merged by tolerance, and used as the reference axis. This is an eager operation and returns centroided feature-by-pixel data.

Queue local-maximum peak picking, optionally extracting a reference.

Methods diff, sd, mad, quantile, filter, and cwt select noise estimates/detection backends. type_ is height or area. Without ref, ragged spectra become per-spectrum peak lists; shared-axis experiments keep the shared axis and zero non-peak samples.

Convenience pipeline for peak picking followed by peak alignment.

If sample_size is set and ref is omitted, evenly spaced spectra are peak-picked to estimate a reference before extracting peaks from the full dataset. A fraction in (0, 1) is treated as a proportion; values at least 1 are interpreted as a count.

Return zero-based feature indices matching metadata predicates.

query is a pandas expression, such as "mz >= 500 and quality > 0.8". Keyword conditions use exact column names for equality, or column__op with eq, ne, lt, le, gt, ge, in, or notin. A callable condition receives the metadata Series and returns a boolean mask. Multiple predicates are combined with logical AND.

Return zero-based pixel indices matching pixel metadata predicates.

Query strings and keyword conditions follow :func:features semantics.

Return a shared-domain experiment with matching feature rows.

Return a dataset containing pixels matching metadata predicates.

Subset shared-domain data by feature and pixel selectors.

select indexes feature rows and subset indexes pixel columns. For ragged MSImagingArrays, only pixel selection is defined.

Rank features by colocalization against a reference image or feature.

i selects a feature index, mz selects matching feature m/z values, and ref may be a numeric pixel vector or a logical mask. The returned DataFrame contains the ranking columns plus feature metadata.

Extract selected feature intensities as 2D ion-image rasters.

Coordinates are rasterized with rows ordered by ascending y and columns by ascending x. Missing grid positions are NaN. With simplify=True, results are stacked as (feature, run, y, x) and singleton dimensions are removed when drop=True. Otherwise a feature-major list of per-run image lists is returned. Three-dimensional coordinates are not yet rasterized.

Find each point's neighbors within radius r.

The point itself is included. metric="maximum" uses Chebyshev distance; euclidean and minkowski use the supplied Minkowski p. Group labels prevent neighbors from crossing run or other group boundaries.

Return neighborhood-weighted distances from observations in x to y.

The result has one row per center coordinate and one column per observation in y. Set byrow=False for a feature-by-pixel matrix.

Calculate Gaussian spatial weights, optionally modulated by data similarity.

For an imaging experiment, pixel coordinates and feature vectors are inferred. Otherwise x supplies coordinates unless coord is explicitly provided. Adaptive weights multiply the spatial Gaussian by a per-neighborhood Gaussian based on Euclidean distances between data vectors.

Reduce each row of a two-dimensional dense or sparse matrix.

Reduce each column of a two-dimensional dense or sparse matrix.

Add per-feature summaries across pixels to feature metadata.

Group labels refer to pixels. Grouped columns are named group.stat.

Add per-pixel summaries across features to pixel metadata.

Group labels refer to features. Grouped columns are named group.stat.

Fit nonnegative matrix factorization with the selected solver.

Fit the available OPLS-compatible supervised baseline.

Fit principal components to observations by features.

Fit a supervised PLS baseline for regression or classification.

Return simple pairwise contrasts from a fitted means test.

Fit and score a model independently on each validation fold.

Fit a formula-based means model to per-pixel experiment summaries.

Run a means test using the first spatial segmentation class.

Fit independent Gaussian mixtures for selected feature columns.

Fit a neighborhood-smoothed low-dimensional spatial projection.

Cluster observations with K-means for one or more cluster counts.

Fit a supervised or unsupervised shrunken-centroid baseline.

Rank features for a supported Phase 6 result object.

Plot a model's spatial values when pixel coordinates are attached.

Plot one or more ion images and return the Matplotlib axes.

Plot common result-object fields such as scores, centers, or clusters.

Plot selected spectra and return the Matplotlib axes.

Combine named boolean masks into a first-match categorical factor.

Return a pixel mask from a polygon or selected coordinate points.

polygon and points provide a deterministic, noninteractive path for scripts and tests. When neither is provided, an interactive Matplotlib polygon selector is opened for mode="region".

Bin ragged spectra onto a shared m/z axis and return an experiment.

If all spectra already share exactly the same axis, it is reused. For varying axes, supply mz or both mass_range and resolution. Nearest-axis bins outside tolerance are omitted; default tolerance is half the requested resolution (or half a local axis interval).

Return nonzero per-pixel spectra from a shared-domain experiment.

Read an imzML file into a mass-imaging dataset.

This first implementation reads spectra eagerly. memory=False is reserved for future file-backed loading and currently raises NotImplementedError rather than silently loading the full file.

Write a dataset and its binary sidecar as an imzML pair.

With bundle=True, file names a directory and output XML uses that directory's name. With bundle=False, file is the output XML path.

Read a supported mass-spectrometry imaging file by extension.

Write a supported mass-spectrometry imaging file by extension.

Read Analyze 7.5 data (not implemented yet).

Write Analyze 7.5 data (not implemented yet).

Generate one or more reproducible profile or centroided spectra.

Returns a MassDataFrame with one m/z value per row. A single spectrum uses an intensity column; multiple spectra use intensity_1 through intensity_n. Profile output is sampled on a shared domain. Centroided output uses the supplied/theoretical peak positions as its shared domain.

Return pixel metadata with a circle or square ROI mask column added.

Build pixel/feature designs for one of nine example image presets.

Presets provide compact, deterministic-in-design examples rather than reproductions of Cardinal's R random-number stream. Presets 1-8 are 2D; preset 9 creates a 3D lattice when dim has three entries.

Simulate a complete imaging experiment from a design or preset.

Matching boolean columns in pixel_data and numeric columns in feature_data define where each feature is present and its mean signal. If preset is supplied, preset_image_def creates that design. random_state accepts an integer seed or a NumPy Generator.