lesiondiffusion-mlx 0.3.0

Breaking changes

run returns the shared medmlx-core result with a new artifacts entry, and the output roles are renamed. outputs keeps the same files under image (was synthetic_image), mask (was target_mask) and raw_image (was raw_synthetic_image); organ_mask is unchanged. artifacts describes each output, and details keeps the run metadata.

-from lesiondiffusion_mlx.medmlx import run
+from lesiondiffusion_mlx import run
 result = run(image_path="ct.nii.gz", organ_mask_path="liver.nii.gz", output_dir="out", organ="liver", lesion_type="tumor")
-print(result["outputs"]["synthetic_image"], result["outputs"]["target_mask"])
+print(result["outputs"]["image"], result["outputs"]["mask"])

Requires medmlx-core>=0.3,<0.4 (was 0.2.x).

Fixes

Outputs are restored if publishing the new files fails part-way.

New

load_model() loads the weights once for repeated draws on one CT and lesion outline, for example many seeds or phenotype texts. Each model.run(...) takes the same arguments as run and returns the same result; the validated geometry and encoded CT latent are reused.

sampler="dpmpp_2m" selects deterministic DPM-Solver++ 2M and lmnet_diffusion_steps / linet_diffusion_steps set the step count of each stage. The default 200-step DDIM path is unchanged, and source mode keeps DDIM.

decoder_mode="scaled_fp16" decodes in scaled half precision on M5-class GPUs with NAX enabled. Lesion-region HU differs from the default FP32 decoder by about 0.12 HU on average (maximum below 0.5 HU).

The README example now uses output_region="lesion", which keeps the input CT outside the lesion neighborhood. The default stays "full".

Faster

With sampler="dpmpp_2m" at 10 steps, a generated-mask run takes 3.5 s warm at 2.60 GB peak MLX memory with decoder_mode="scaled_fp16", against 4.7 s at 5.19 GB with the default FP32 decoder (M5 Max, 160x160x128 crop). The VQ encoder, LINet and VQ decoder now load one after another, which lowers the scaled path's peak.

lesiondiffusion-mlx 0.2.8

Fixes

Conditioned inference decodes the VQ latent in FP32 to avoid FP16 overflow. Source mode retains FP16 arithmetic and reports a clear error if decoding produces non-finite values.

New

Keep the input CT with output_region="lesion". This mode adds the generated-minus-source VQ reconstruction within 33 voxels of the lesion on each axis and preserves the input elsewhere, avoiding the source reconstruction shift retained by box mode.

lesiondiffusion-mlx 0.2.7

Fixes

LesionDiffusion now runs on Macs with much less free memory. Its declared peak_memory_gb falls from 18.74 to 3.46, so MedMLX admits it on smaller Macs. On a 160×160×128 AMOS crop with 200 steps, peak MLX memory falls from 14.49 GB to 3.46 GB in source mode and from 5.14 GB to 3.29 GB in conditioned mode, with bitwise-identical outputs.

Outputs are staged in a private temporary folder before being moved into place, instead of beside the destination under predictable names. A dangling symlink at an output path now counts as an existing output, rejected unless overwrite is set.

Only MLX runtime errors are reported as ModelExecutionError. Other exceptions now reach the caller unchanged.

lesiondiffusion-mlx 0.2.6.post2

Updates the README on PyPI: the install command and links point at this release.

lesiondiffusion-mlx 0.2.6.post1

Updates the README on PyPI: removes links to a page that is no longer published.

lesiondiffusion-mlx 0.2.6

First release: add synthetic lesions to CT with native MLX, with controls for organ, size, placement and phenotype.