Implement Enhanced Background Subtraction Method in Fluorescence Imag…

…ing Analysis (#169) * Update quick_start notebook with additional functionality The quick_start notebook has been updated with more in-depth functionality, including executing code cells and importing additional libraries. The changes also involve modifying the output, providing visual data clarification, and changing the sample path to retrieve other data. The adjustments aim to enhance the user's analytical experience. * Add code for delta calculations in astrocast notebook The changes includes importing necessary modules and adding code chunks for performing delta calculations, implementation of interpolation functions, data downsampling measures and peak removal operations. Added new function for delta calculation: get_subtracted, get_interpolated, and remove_peaks. The code edits are focused on handling and processing 3D datasets. * feat: Implement enhanced background subtraction in fluorescence imaging This commit introduces a `_subtract_delta_rbf`, designed to significantly improve background subtraction in time-series fluorescence imaging recordings. The primary objective is to enhance the signal-to-noise ratio and mitigate the effects of photo-bleaching. Key features of the function include: - Subdivision of video data into manageable chunks for efficient processing. - Identification and treatment of significant signal fluctuations (peaks) as missing data (NaN), which are then excluded from background calculations. - Utilization of the remaining pixels to interpolate the background, subsequently subtracting this from the original dataset to isolate the fluorescence signal from the near-static background. - Inclusion of operations like Gaussian blurring for noise reduction, scaled downsizing for computational efficiency, and 3D RBF interpolation for smoother data transitions. - Reassembly of processed chunks into a cohesive array, maintaining the original video dimensions and data type. * officially replacing _calculate_delta_min_filter with subtract_delta_rbf - add: testing the new delta subtraction - fix: switched function call and arguments in Delta.run - fix: updated documentation * Integration of new delta functionality into quick_start notebook * Updated CLI interface to include changes to Delta subtraction * removed Delta from the GUI. Needs to be updated with new subtraction function before it can be re-enabled * removed superfluous parameter max_tries * - fix: default parameter for rbf smoothing needs to be 0 in client as well * - refactor TestLocalCache * increase flakyness of subtraction performance testing * - add check if tensorflow is not available during import of denoising * moved another tensorflow import into try-except-clause * - fix: test functions for local caching * convert todo to GitHub issues * now adhere to PEP * rename 'Analysis GUI' to 'Exploratory Analysis GUI' * - fix: fix random output generator * reformat * move keras imports into tensorflow available try-except-clause
janreising · Dec 13, 2023 · c79d335 · c79d335
1 parent 1ed7e52
commit c79d335
Show file tree

Hide file tree

Showing 16 changed files with 6,833 additions and 5,382 deletions.
diff --git a/astrocast/analysis.py b/astrocast/analysis.py
diff --git a/astrocast/app_analysis.py b/astrocast/app_analysis.py
diff --git a/astrocast/app_preparation.py b/astrocast/app_preparation.py
diff --git a/astrocast/cli_interfaces.py b/astrocast/cli_interfaces.py
diff --git a/astrocast/denoising.py b/astrocast/denoising.py
diff --git a/astrocast/helper.py b/astrocast/helper.py
diff --git a/astrocast/preparation.py b/astrocast/preparation.py
diff --git a/...28_128_pre_post_frame_5-gap_frames_0-train_rotation_1_2_3-architecture_3_64_epochs-50.pth b/...28_128_pre_post_frame_5-gap_frames_0-train_rotation_1_2_3-architecture_3_64_epochs-50.pth
diff --git a/notebooks/exploring_delta.ipynb b/notebooks/exploring_delta.ipynb
diff --git a/notebooks/quick_start.ipynb b/notebooks/quick_start.ipynb
@@ -6,17 +6,17 @@
    "id": "initial_id",
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2023-11-15T13:23:02.649268383Z",
-     "start_time": "2023-11-15T13:23:00.374557839Z"
+     "end_time": "2023-12-13T01:46:47.302299673Z",
+     "start_time": "2023-12-13T01:46:46.489652956Z"
     }
    },
    "outputs": [],
    "source": [
-    "import shutil\n",
+    "import os\n",
+    "from pathlib import Path\n",
+    "from astrocast.analysis import IO, Video, Events, Plotting\n",
     "\n",
-    "from astrocast.analysis import IO, Video, Events\n",
-    "\n",
-    "from pathlib import Path"
+    "astroPlot = Plotting(None)"
    ]
   },
   {
@@ -45,13 +45,13 @@
    "id": "1332388990b34934",
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2023-11-15T13:23:05.045991906Z",
-     "start_time": "2023-11-15T13:23:02.848390645Z"
+     "end_time": "2023-12-13T01:40:46.705246932Z",
+     "start_time": "2023-12-13T01:40:44.377908284Z"
     }
    },
    "outputs": [],
    "source": [
-    "sample_path = Path(\"../testdata/sample_2.h5\")\n",
+    "sample_path = Path(\"../testdata/bigger_data.h5\")\n",
     "assert sample_path.exists()\n",
     "\n",
     "# we can use the visualize_h5 function from the command line interface (CLI)\n",
@@ -65,8 +65,8 @@
    "id": "530f135c103b42fa",
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2023-11-15T13:28:59.870710Z",
-     "start_time": "2023-11-15T13:28:58.958705846Z"
+     "end_time": "2023-12-13T01:40:47.555619679Z",
+     "start_time": "2023-12-13T01:40:46.698974442Z"
     }
    },
    "outputs": [],
@@ -100,14 +100,13 @@
    "id": "d29cde6734185ae9",
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2023-11-15T12:49:27.209811383Z",
-     "start_time": "2023-11-15T12:49:23.887395678Z"
+     "end_time": "2023-12-13T01:40:47.559949780Z",
+     "start_time": "2023-12-13T01:40:47.555929675Z"
     }
    },
    "outputs": [],
    "source": [
-    "from astrocast.preparation import MotionCorrection, Input, Delta\n",
-    "from astrocast.denoising import SubFrameGenerator\n"
+    "from astrocast.preparation import MotionCorrection, Input, Delta"
    ]
   },
   {
@@ -126,8 +125,8 @@
    "id": "df78f685eac21f86",
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2023-11-15T13:23:46.867382844Z",
-     "start_time": "2023-11-15T13:23:46.800282852Z"
+     "end_time": "2023-12-13T01:46:56.243495252Z",
+     "start_time": "2023-12-13T01:46:56.191593542Z"
     }
    },
    "outputs": [],
@@ -141,8 +140,8 @@
    "id": "c1fd2411849c47f8",
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2023-11-15T12:49:38.290035837Z",
-     "start_time": "2023-11-15T12:49:27.214509032Z"
+     "end_time": "2023-12-13T01:40:53.014833707Z",
+     "start_time": "2023-12-13T01:40:47.566361818Z"
     }
    },
    "outputs": [],
@@ -156,7 +155,10 @@
     "assert not test_path.exists()\n",
     "\n",
     "inp = Input()\n",
-    "inp.run(input_path=sample_path, output_path=test_path, loc_in=\"data/ch0\", loc_out=\"data/ch0\", compression=\"gzip\", infer_strategy=\"Z\")\n",
+    "inp.run(input_path=sample_path, output_path=test_path, loc_in=\"data/ch0\", loc_out=\"data\", compression=\"gzip\",\n",
+    "        chunk_strategy=\"Z\")\n",
+    "\n",
+    "!astrocast visualize-h5 {test_path.as_posix()}\n",
     "\n",
     "# alternatively with CLI:\n",
     "# !astrocast convert-h5 --output-path {test_path.as_posix()} --h5-loc-in data/ch0 --h5-loc-out data/ch0 --compression gzip --infer-chunks {sample_path.as_posix()}"
@@ -178,15 +180,15 @@
    "id": "963ae00bcfbff3b1",
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2023-11-15T12:50:03.346591303Z",
-     "start_time": "2023-11-15T12:49:38.271208670Z"
+     "end_time": "2023-12-13T01:41:34.337019648Z",
+     "start_time": "2023-12-13T01:40:53.027503559Z"
     }
    },
    "outputs": [],
    "source": [
     "MC = MotionCorrection()\n",
-    "MC.run(input_=test_path, loc=\"data/ch0\")\n",
-    "MC.save(output=test_path, loc=\"mc/ch0\", infer_strategy=\"Z\", compression=\"gzip\")"
+    "MC.run(path=test_path, loc=\"data/ch0\")\n",
+    "MC.save(output=test_path, loc=\"mc/ch0\", chunk_strategy=\"Z\", compression=\"gzip\")"
    ]
   },
   {
@@ -195,8 +197,8 @@
    "id": "6e2cbe26589901b0",
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2023-11-15T12:50:05.838657729Z",
-     "start_time": "2023-11-15T12:50:03.350712941Z"
+     "end_time": "2023-12-13T01:41:36.566852402Z",
+     "start_time": "2023-12-13T01:41:34.338136605Z"
     }
    },
    "outputs": [],
@@ -210,8 +212,8 @@
    "id": "e4ce35cc81668fe2",
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2023-11-15T12:50:05.844972520Z",
-     "start_time": "2023-11-15T12:50:05.840285718Z"
+     "end_time": "2023-12-13T01:41:36.601313254Z",
+     "start_time": "2023-12-13T01:41:36.574445364Z"
     }
    },
    "outputs": [],
@@ -229,58 +231,109 @@
     "## 1.3 Denoising"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "3f51ba37a0e325fe",
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2023-12-13T01:41:39.990036691Z",
+     "start_time": "2023-12-13T01:41:36.592564785Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "from astrocast.denoising import SubFrameDataset, PyTorchNetwork"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
    "id": "7958c8bd4688fe47",
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2023-11-15T12:50:05.852536566Z",
-     "start_time": "2023-11-15T12:50:05.843728331Z"
+     "end_time": "2023-12-13T01:41:39.994116932Z",
+     "start_time": "2023-12-13T01:41:39.990334737Z"
     }
    },
    "outputs": [],
    "source": [
-    "# Download pretrained models\n",
-    "# !astrocast download-models \"./models\"--"
+    "# Optional: Download pretrained models\n",
+    "# !astrocast download-models \"./models\"\n",
+    "# for f in os.listdir(\"./models\"):\n",
+    "#     print(f)"
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "6e40311858f09759",
+   "id": "c4714c207508281",
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2023-11-15T12:50:05.859394523Z",
-     "start_time": "2023-11-15T12:50:05.850467828Z"
+     "end_time": "2023-12-13T01:43:43.875067711Z",
+     "start_time": "2023-12-13T01:41:39.997752373Z"
     }
    },
    "outputs": [],
    "source": [
-    "# data_generator = SubFrameGenerator(paths=test_path, input_size=(64, 64), pre_post_frame=5, gap_frames=0, batch_size=64, loc=\"mc/ch0\",\n",
-    "#                                    overlap=5, padding=\"edge\",normalize=\"global\")\n",
-    "# \n",
-    "# denoising_model = Path(\"./models/1p_exvivo_input_size_64_64-pre_post_frame_5-gap_frames_0-train_rotation_1_2_3-architecture_3_64.h5\")\n",
-    "# assert denoising_model.exists()\n",
-    "# \n",
-    "# data_generator.infer(model=denoising_model, output=test_path, out_loc=\"dn/ch0\")"
+    "# Inference dataset\n",
+    "input_size = (128, 128)\n",
+    "loc_in = \"mc/ch0\"\n",
+    "loc_out = \"inf/ch0\"\n",
+    "use_cpu = True  # switch to False, if cuda is available\n",
+    "\n",
+    "infer_dataset = SubFrameDataset(paths=test_path, loc=loc_in, input_size=input_size, allowed_rotation=0, allowed_flip=-1,\n",
+    "                                shuffle=False, normalize=\"global\", overlap=10, padding=\"edge\")\n",
+    "\n",
+    "# Model parameters\n",
+    "model_path = \"../denoiser_models/1p_exvivo_input_size_128_128_pre_post_frame_5-gap_frames_0-train_rotation_1_2_3-architecture_3_64_epochs-50.pth\"\n",
+    "pre_post_frames = 5\n",
+    "gap_frames = 0\n",
+    "n_stacks, kernels = (3, 64)\n",
+    "\n",
+    "net = PyTorchNetwork(infer_dataset, load_model=model_path, val_dataset=None, batch_size=16, n_stacks=n_stacks,\n",
+    "                     kernels=kernels, kernel_size=3, batch_normalize=False,\n",
+    "                     use_cpu=use_cpu)\n",
+    "\n",
+    "# Denoise data\n",
+    "net.infer(dataset=infer_dataset, output=test_path, out_loc=loc_out, batch_size=1, dtype=float)"
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8a3975f809e908a8",
+   "id": "9ef7a72fa7b0e090",
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2023-11-15T12:50:05.859744528Z",
-     "start_time": "2023-11-15T12:50:05.855818082Z"
+     "end_time": "2023-12-13T01:43:43.876766363Z",
+     "start_time": "2023-12-13T01:43:43.874772842Z"
     }
    },
    "outputs": [],
    "source": [
     "# !astrocast visualize-h5 {test_path.as_posix()}"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "a001a57eddab48ad",
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2023-12-13T01:43:47.243894852Z",
+     "start_time": "2023-12-13T01:43:43.874895648Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "frame = 100\n",
+    "pixels = [(167, 153), (199, 306), (214, 109), (337, 72)]\n",
+    "\n",
+    "fig, axx = astroPlot.compare_pixels(pixels=pixels, frame=frame,\n",
+    "                                    data_1=test_path, loc_1=\"mc/ch0\", loc_2=\"inf/ch0\",\n",
+    "                                    title_1=\"before denoising\", title_2=\"after denoising\")"
+   ]
+  },
   {
    "cell_type": "markdown",
    "id": "c6652771275a017a",
@@ -297,15 +350,21 @@
    "id": "412edb17d8d727dc",
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2023-11-15T12:50:41.175675394Z",
-     "start_time": "2023-11-15T12:50:05.862139165Z"
+     "end_time": "2023-12-13T01:49:01.439766400Z",
+     "start_time": "2023-12-13T01:47:15.902855457Z"
     }
    },
    "outputs": [],
    "source": [
-    "delta = Delta(test_path, loc=\"mc/ch0\")\n",
-    "delta.run(window=10, method=\"dF\")\n",
-    "delta.save(output_path=test_path, loc=\"df/ch0\", infer_strategy=\"balanced\", compression=\"gzip\")\n",
+    "import astrocast.preparation as prep\n",
+    "import importlib\n",
+    "\n",
+    "importlib.reload(prep)\n",
+    "\n",
+    "delta = prep.Delta(test_path, loc=\"inf/ch0\")\n",
+    "res_delta = delta.run(method=\"dF\", scale_factor=0.25, neighbors=100, wlen=50, distance=5,\n",
+    "                      max_chunk_size_mb=10, width=5)\n",
+    "delta.save(output_path=test_path, loc=\"df/ch0\", chunk_strategy=\"balanced\", compression=\"gzip\")\n",
     "\n",
     "# TODO add terminal command"
    ]
@@ -316,13 +375,13 @@
    "id": "2b1f6630279f1fae",
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2023-11-15T12:50:43.678537712Z",
-     "start_time": "2023-11-15T12:50:41.170915936Z"
+     "end_time": "2023-12-13T01:45:25.051079053Z",
+     "start_time": "2023-12-13T01:45:25.043322800Z"
     }
    },
    "outputs": [],
    "source": [
-    "!astrocast visualize-h5 {test_path.as_posix()}"
+    "# !astrocast visualize-h5 {test_path.as_posix()}"
    ]
   },
   {
@@ -331,13 +390,18 @@
    "id": "4ee9f2f2330388f2",
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2023-11-15T12:50:43.684872030Z",
-     "start_time": "2023-11-15T12:50:43.680606803Z"
+     "end_time": "2023-12-13T01:49:11.858719123Z",
+     "start_time": "2023-12-13T01:49:09.002787740Z"
     }
    },
    "outputs": [],
    "source": [
-    "# TODO visualize results"
+    "frame = 100\n",
+    "pixels = [(167, 153), (199, 306), (214, 109), (337, 72)]\n",
+    "\n",
+    "fig, axx = astroPlot.compare_pixels(pixels=pixels, frame=frame,\n",
+    "                                    data_1=test_path, loc_1=\"inf/ch0\", loc_2=\"df/ch0\",\n",
+    "                                    title_1=\"before subtraction\", title_2=\"after subtraction\")"
    ]
   },
   {
@@ -356,8 +420,8 @@
    "id": "8f41c647cdca684e",
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2023-11-15T13:23:19.067051711Z",
-     "start_time": "2023-11-15T13:23:19.058555477Z"
+     "end_time": "2023-12-13T01:49:26.794352561Z",
+     "start_time": "2023-12-13T01:49:26.725874694Z"
     }
    },
    "outputs": [],
@@ -372,8 +436,8 @@
    "id": "c7161fbc46920ef",
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2023-11-15T13:25:36.386913586Z",
-     "start_time": "2023-11-15T13:23:53.540520527Z"
+     "end_time": "2023-12-13T01:50:48.088514635Z",
+     "start_time": "2023-12-13T01:49:28.543331996Z"
     }
    },
    "outputs": [],
@@ -392,14 +456,15 @@
    "id": "d9122fa8153591c9",
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2023-11-15T13:25:36.645356453Z",
-     "start_time": "2023-11-15T13:25:36.386641275Z"
+     "end_time": "2023-12-13T01:50:48.252272039Z",
+     "start_time": "2023-12-13T01:50:48.090799508Z"
     }
    },
    "outputs": [],
    "source": [
     "assert event_path.exists()\n",
-    "!ls -lthv {event_path.as_posix()}\n",
+    "for f in os.listdir(event_path):\n",
+    "    print(f)\n",
     "\n",
     "events = Events(event_path)\n",
     "events"
@@ -411,13 +476,12 @@
    "id": "c8840eb4703e630c",
    "metadata": {
     "ExecuteTime": {
-     "end_time": "2023-11-15T13:27:40.344269093Z",
-     "start_time": "2023-11-15T13:25:36.647233721Z"
+     "start_time": "2023-12-13T01:45:25.060165539Z"
     }
    },
    "outputs": [],
    "source": [
-    "#!astrocast view-detection-results --video-path {test_path.as_posix()} --h5-loc df/ch0 {event_path.as_posix()}"
+    "#!astrocast view-detection-results --video-path {test_path.as_posix()} --loc df/ch0 {event_path.as_posix()}"
    ]
   }
  ],