dslighting 1.3.9__py3-none-any.whl → 1.7.6__py3-none-any.whl

dslighting/registry/random-acts-of-pizza/prepare_val.py

@@ -0,0 +1,144 @@
+import json
+import shutil
+from pathlib import Path
+from typing import List, Dict, Any
+
+import pandas as pd
+from sklearn.model_selection import train_test_split
+
+
+def _create_split(
+    data_to_split: List[Dict],
+    test_size: float,
+    test_fields: List[str],
+    public_path: Path,
+    private_path: Path,
+    random_state: int,
+) -> List[Dict]:
+    """
+    Helper function to perform a data split and create all required files.
+
+    Args:
+        data_to_split: The list of data samples to be split.
+        test_size: The proportion of the dataset to allocate to the test split.
+        test_fields: The list of fields to keep in the test set.
+        public_path: The directory for public-facing files (train set, unlabeled test set).
+        private_path: The directory for private files (test set labels).
+        random_state: The seed for the random number generator.
+
+    Returns:
+        The training portion of the split, to be used for subsequent splits if needed.
+    """
+    # Create output directories if they don't exist
+    public_path.mkdir(parents=True, exist_ok=True)
+    private_path.mkdir(parents=True, exist_ok=True)
+
+    # Create train, test from the provided data
+    train_set, test_set = train_test_split(
+        data_to_split, test_size=test_size, random_state=random_state
+    )
+
+    test_set_without_labels = []
+    for sample in test_set:
+        # Keep only the fields that should be in the test set
+        sample_unlabeled = {key: sample[key] for key in test_fields}
+        test_set_without_labels.append(sample_unlabeled)
+
+    # Save the new train and test splits to the public directory
+    with open(public_path / "train.json", "w") as f:
+        json.dump(train_set, f, indent=4)
+    with open(public_path / "test.json", "w") as f:
+        json.dump(test_set_without_labels, f, indent=4)
+    # Also save zipped versions
+    shutil.make_archive(public_path / "train", "zip", public_path, "train.json")
+    shutil.make_archive(public_path / "test", "zip", public_path, "test.json")
+
+    # Create answers for the private directory
+    answers_rows = []
+    for sample in test_set:
+        answers_rows.append(
+            {
+                "request_id": sample["request_id"],
+                "requester_received_pizza": int(sample["requester_received_pizza"]),
+            }
+        )
+    answers = pd.DataFrame(answers_rows)
+    answers.to_csv(private_path / "test.csv", index=False)
+
+    # Create sample submission for the public directory
+    sample_submission = answers.copy()
+    sample_submission["requester_received_pizza"] = 0
+    sample_submission.to_csv(public_path / "sampleSubmission.csv", index=False)
+
+    # Perform checks
+    assert len(train_set) + len(test_set) == len(
+        data_to_split
+    ), f"Expected {len(data_to_split)} total samples, but got {len(train_set)} in train and {len(test_set)} in test"
+    assert len(test_set) == len(
+        test_set_without_labels
+    ), "Test set and unlabeled test set must have the same length"
+    assert len(answers) == len(test_set), "Answers must have the same length as the test set"
+    assert len(sample_submission) == len(
+        test_set
+    ), "Sample submission must have the same length as the test set"
+    assert set(answers.columns) == set(
+        ["request_id", "requester_received_pizza"]
+    ), "Answers must have 'request_id' and 'requester_received_pizza' columns"
+    assert set(sample_submission.columns) == set(
+        ["request_id", "requester_received_pizza"]
+    ), "Sample submission must have 'request_id' and 'requester_received_pizza' columns"
+
+    return train_set
+
+
+def prepare(raw: Path, public: Path, private: Path):
+    """
+    Splits the data in raw into public and private datasets with appropriate test/train splits.
+    Then, it creates a second, parallel validation split from the first training set.
+    """
+
+    # Load data
+    with open(raw / "train.json") as f:
+        old_train = json.load(f)
+    with open(raw / "test.json") as f:
+        old_test = json.load(f)
+
+    test_ratio = len(old_test) / (len(old_train) + len(old_test))
+
+    all_fields = list([key for key in old_train[0].keys()])
+    assert all(set(all_fields) == set([key for key in sample.keys()]) for sample in old_train)
+    test_fields = list([key for key in old_test[0].keys()])
+    assert all(set(test_fields) == set([key for key in sample.keys()]) for sample in old_test)
+
+    # --- Original Split ---
+    # This split creates the primary `public` and `private` competition data.
+    # The returned `new_train` set will be used for the subsequent validation split.
+    new_train = _create_split(
+        data_to_split=old_train,
+        test_size=test_ratio,
+        test_fields=test_fields,
+        public_path=public,
+        private_path=private,
+        random_state=0,
+    )
+
+    # --- New Validation Split ---
+    # Define new directories for the validation set, parallel to the original ones.
+    public_val = public.parent / "public_val"
+    private_val = private.parent / "private_val"
+
+    # To keep the new test set (`test_val`) size consistent with the original test set,
+    # we adjust the test ratio for the second split.
+    # val_test_size = original_test_size / original_train_size
+    val_test_size = test_ratio / (1.0 - test_ratio)
+
+    # Create the validation split using the same logic, but on the `new_train` data
+    # and saving to the new `_val` directories.
+    _create_split(
+        data_to_split=new_train,
+        test_size=val_test_size,
+        test_fields=test_fields,
+        public_path=public_val,
+        private_path=private_val,
+        random_state=0,
+    )

dslighting/registry/ranzcr-clip-catheter-line-classification/checksums.yaml

@@ -0,0 +1,7 @@
+private:
+  test.csv: e2959bb0ca0ca9fc983b02e01caf31fa
+public:
+  sample_submission.csv: fade90160bce8a12138711b20fa89584
+  train.csv: 620fa6759729f96539ed66ae860a84f9
+  train_annotations.csv: 1929aab17c1206719cc7edf6d7e9e57b
+zip: f8a117e7ba1b5527c99c80b54beddeb5

dslighting/registry/ranzcr-clip-catheter-line-classification/classes.py

@@ -0,0 +1,11 @@
+CLASSES = [
+    "ETT - Abnormal",
+    "ETT - Borderline",
+    "ETT - Normal",
+    "NGT - Abnormal",
+    "NGT - Borderline",
+    "NGT - Incompletely Imaged",
+    "NGT - Normal",
+    "CVC - Abnormal",
+    "CVC - Borderline",
+]

dslighting/registry/ranzcr-clip-catheter-line-classification/config.yaml

@@ -0,0 +1,31 @@
+id: ranzcr-clip-catheter-line-classification
+name: RANZCR CLiP - Catheter and Line Position Challenge
+competition_type: code
+awards_medals: true
+prizes:
+  - position: 1
+    value: 16000
+  - position: 2
+    value: 12000
+  - position: 3
+    value: 6000
+  - position: 4
+    value: 4000
+  - position: 5
+    value: 4000
+  - position: 6
+    value: 4000
+  - position: 7
+    value: 4000
+    
+description: mlebench/competitions/ranzcr-clip-catheter-line-classification/description.md
+
+dataset:
+  answers: ranzcr-clip-catheter-line-classification/prepared/private/test.csv
+  sample_submission: ranzcr-clip-catheter-line-classification/prepared/public/sample_submission.csv
+
+grader:
+  name: auc-roc
+  grade_fn: mlebench.competitions.ranzcr-clip-catheter-line-classification.grade:grade
+
+preparer: mlebench.competitions.ranzcr-clip-catheter-line-classification.prepare:prepare

dslighting/registry/ranzcr-clip-catheter-line-classification/description.md

@@ -0,0 +1,172 @@
+# Overview
+
+## Description
+
+Serious complications can occur as a result of malpositioned lines and tubes in patients. Doctors and nurses frequently use checklists for placement of lifesaving equipment to ensure they follow protocol in managing patients. Yet, these steps can be time consuming and are still prone to human error, especially in stressful situations when hospitals are at capacity.
+
+Hospital patients can have catheters and lines inserted during the course of their admission and serious complications can arise if they are positioned incorrectly. Nasogastric tube malpositioning into the airways has been reported in up to 3% of cases, with up to 40% of these cases demonstrating complications [1-3]. Airway tube malposition in adult patients intubated outside the operating room is seen in up to 25% of cases [4,5]. The likelihood of complication is directly related to both the experience level and specialty of the proceduralist. Early recognition of malpositioned tubes is the key to preventing risky complications (even death), even more so now that millions of COVID-19 patients are in need of these tubes and lines.
+
+The gold standard for the confirmation of line and tube positions are chest radiographs. However, a physician or radiologist must manually check these chest x-rays to verify that the lines and tubes are in the optimal position. Not only does this leave room for human error, but delays are also common as radiologists can be busy reporting other scans. Deep learning algorithms may be able to automatically detect malpositioned catheters and lines. Once alerted, clinicians can reposition or remove them to avoid life-threatening complications.
+
+The Royal Australian and New Zealand College of Radiologists (RANZCR) is a not-for-profit professional organisation for clinical radiologists and radiation oncologists in Australia, New Zealand, and Singapore. The group is one of many medical organisations around the world (including the NHS) that recognizes malpositioned tubes and lines as preventable. RANZCR is helping design safety systems where such errors will be caught.
+
+In this competition, you’ll detect the presence and position of catheters and lines on chest x-rays. Use machine learning to train and test your model on 40,000 images to categorize a tube that is poorly placed.
+
+The dataset has been labelled with a set of definitions to ensure consistency with labelling. The normal category includes lines that were appropriately positioned and did not require repositioning. The borderline category includes lines that would ideally require some repositioning but would in most cases still function adequately in their current position. The abnormal category included lines that required immediate repositioning.
+
+If successful, your efforts may help clinicians save lives. Earlier detection of malpositioned catheters and lines is even more important as COVID-19 cases continue to surge. Many hospitals are at capacity and more patients are in need of these tubes and lines. Quick feedback on catheter and line placement could help clinicians better treat these patients. Beyond COVID-19, detection of line and tube position will ALWAYS be a requirement in many ill hospital patients.
+
+This is a Code Competition. Refer to Code Requirements for details.
+
+1. Koopmann MC, Kudsk KA, Sz otkowski MJ, Rees SM. A Team-Based Protocol and Electromagnetic Technology Eliminate Feeding Tube Placement Complications [Internet]. Vol. 253, Annals of Surgery. 2011. p. 297–302. Available from: [http://dx.doi.org/10.1097/sla.0b013e318208f550](http://dx.doi.org/10.1097/sla.0b013e318208f550)
+2. Sorokin R, Gottlieb JE. Enhancing patient safety during feeding-tube insertion: a review of more than 2,000 insertions. JPEN J Parenter Enteral Nutr. 2006 Sep;30(5):440–5.
+3. Marderstein EL, Simmons RL, Ochoa JB. Patient safety: effect of institutional protocols on adverse events related to feeding tube placement in the critically ill. J Am Coll Surg. 2004 Jul;199(1):39–47; discussion 47–50.
+4. Jemmett ME. Unrecognized Misplacement of Endotracheal Tubes in a Mixed Urban to Rural Emergency Medical Services Setting [Internet]. Vol. 10, Academic Emergency Medicine. 2003. p. 961–5. Available from: [http://dx.doi.org/10.1197/s1069-6563(03)00315-4](http://dx.doi.org/10.1197/s1069-6563(03)00315-4)
+5. Lotano R, Gerber D, Aseron C, Santarelli R, Pratter M. Utility of postintubation chest radiographs in the intensive care unit. Crit Care. 2000 Jan 24;4(1):50–3.
+
+## Evaluation
+
+Submissions are evaluated on [area under the ROC curve](http://en.wikipedia.org/wiki/Receiver_operating_characteristic) between the predicted probability and the observed target.
+
+To calculate the final score, AUC is calculated for each of the 11 labels, then averaged. The score is then the average of the individual AUCs of each predicted column.
+
+### Submission File
+
+For each ID in the test set, you must predict a probability for all target variables. The file should contain a header and have the following format:
+```
+StudyInstanceUID,ETT - Abnormal,ETT - Borderline,ETT - Normal,NGT - Abnormal,NGT - Borderline,NGT - Incompletely Imaged,NGT - Normal,CVC - Abnormal,CVC - Borderline,CVC - Normal,Swan Ganz Catheter Present
+1.2.826.0.1.3680043.8.498.62451881164053375557257228990443168843,0,0,0,0,0,0,0,0,0,0,0
+1.2.826.0.1.3680043.8.498.83721761279899623084220697845011427274,0,0,0,0,0,0,0,0,0,0,0
+1.2.826.0.1.3680043.8.498.12732270010839808189235995393981377825,0,0,0,0,0,0,0,0,0,0,0
+1.2.826.0.1.3680043.8.498.11769539755086084996287023095028033598,0,0,0,0,0,0,0,0,0,0,0
+1.2.826.0.1.3680043.8.498.87838627504097587943394933987052577153,0,0,0,0,0,0,0,0,0,0,0
+1.2.826.0.1.3680043.8.498.53211840524738036417560823327351887819,0,0,0,0,0,0,0,0,0,0,0
+1.2.826.0.1.3680043.8.498.93555795394184819372299157360228027866,0,0,0,0,0,0,0,0,0,0,0
+1.2.826.0.1.3680043.8.498.52241894131170494723503100795076463919,0,0,0,0,0,0,0,0,0,0,0
+1.2.826.0.1.3680043.8.498.36500167484503936720548852591033878284,0,0,0,0,0,0,0,0,0,0,0
+1.2.826.0.1.3680043.8.498.86199852603457900780565655267977637728,0,0,0,0,0,0,0,0,0,0,0
+```
+
+## Timeline
+- March 8, 2021 - Entry deadline. You must accept the competition rules before this date in order to compete.
+- March 8, 2021 - Team Merger deadline. This is the last day participants may join or merge teams.
+- March 16, 2021 - Final submission deadline.
+
+All deadlines are at 11:59 PM UTC on the corresponding day unless otherwise noted. The competition organizers reserve the right to update the contest timeline if they deem it necessary.
+
+## Prizes
+- 1st Place - $16,000
+- 2nd Place - $12,000
+- 3rd Place - $6,000
+- 4th-7th Place - $4,000
+
+## Code Requirements
+
+**This is a Code Competition**
+
+Submissions to this competition must be made through Notebooks. In order for the "Submit to Competition" button to be active after a commit, the following conditions must be met:
+
+- CPU Notebook <= 9 hours run-time
+- GPU Notebook <= 9 hours run-time
+- [TPUs](https://www.kaggle.com/docs/tpu) will not be available for making submissions to this competition. You are still welcome to use them for training models. For a walk-through on how to train on TPUs and run inference/submit on GPUs, see our [TPU Docs](https://www.kaggle.com/docs/tpu#tpu6).
+- No internet access enabled on submission
+- External data, freely & publicly available, is allowed. This includes pre-trained models.
+- Submission file must be named submission.csv
+
+Please see the [Code Competition FAQ](https://www.kaggle.com/docs/competitions#notebooks-only-FAQ) for more information on how to submit. And review the [code debugging doc](https://www.kaggle.com/code-competition-debugging) if you are encountering submission errors.
+
+## RANZCR 2021
+
+[RANZCR’s](https://www.ranzcr2021.com/) 71st Annual Scientific Meeting: Elysium Diagnostics and Therapeutics (Now and in Future) with Artificial Intelligence from 16- 19 September 2021 will take place at the Melbourne Convention and Exhibition Centre. Not only does the meeting promise to deliver a thought-provoking program, you will also have the opportunity to enjoy the cosmopolitan city of Melbourne.
+
+RANZCR2021 will showcase emerging trends in radiology and radiation oncology AI software, hardware, regulatory, industry perspectives, impact on healthcare, and patient outcomes. A varied and engaging program of international and national speakers and an interactive program will highlight the opportunities in radiology and radiation oncology in the future. There will be a series of multi-disciplinary sessions to discuss current clinical topics with affiliated organisations. In 2021, we will have an increased number of international experts to deliver captivating sessions for attendees.
+
+**RANZCR 2021 PRESENTING PARTNER**
+
+RANZCR and [Intelerad](https://bit.ly/2Xkd4E1) have a strong partnership that will emerge from the COVID-19 crisis stronger than ever. With the postponement of the Annual Scientific Meeting this year, [Intelerad](https://bit.ly/2Xkd4E1) will proudly serve as a Presenting Partner of RANZCR2021. Our roots run deep in Australia and New Zealand. We’re proud and excited to be Presenting Partners of RANZCR 2021. Through this partnership, we are reconfirming our ongoing commitment to the medical imaging and radiology community of Oceania.
+
+## Acknowledgements
+
+### Challenge Organizing Team
+
+- Professor Meng Law, Alfred Health, Monash University
+- Dr. Jennifer Tang, Melbourne Health
+- Dr. Jarrel Seah, Alfred Health
+- Mr. Adil Zia, Alfred Health
+- Dr. George Shih, Weill Cornell Medicine
+- Dr. Anouk Stein, MD - MD.ai
+- Dr. Andrew Dixon, Alfred Health
+- Associate Professor Frank Gaillard, Melbourne Health, University of Melbourne
+- Sarah Dowd and RANZCR staff
+
+### Acknowledgements
+
+Thank you to the National Institutes of Health Clinical Center for publicly providing the Chest X-Ray14 dataset [1].
+
+### Data Annotators
+
+The challenge organizers wish to thank the doctors who volunteered their time to create the dataset labelling over 40,000 examinations.
+
+The following individuals labeled data for the challenge:
+
+More than 5,000 exams:
+
+Dr. Jay Gajera, Dr. Richard Schlegel, Dr. Dayu Gai, Dr. Aaron Wong, Dr. Shu Su, Dr. Tony Bose
+
+**1000 - 5000 exams:**
+
+Dr. Marcus Kok, Dr. Alex Jarema, Dr. George Harisis, Dr. Chris-tin Cheng, Dr. Brian Ngo, Dr. David Ng, Dr. Matthew Lukies, Dr. Christina Giudice, Dr. Sameer Dave
+
+**500 - 1000 exams:**
+
+Dr. Jamie Cooke, Dr. Cassie Williams, Dr. Amy Reid, Dr. Wayland Wang, Dr. Joel Hng
+
+**250 - 500 exams:**
+
+Dr. Tracy Liu, Dr. Jake Gipson, Dr. Neel Gore, Dr. Irene Bryan, Dr. Maryam Shekarforoush, Dr. Calvin Gan, Dr. Sabrina Yeh, Dr. Madeleine Schicchitano, Dr. Shalini Bigwood
+
+Special thanks to [MD.ai](https://www.kaggle.com/competitions/ranzcr-clip-catheter-line-classification/md.ai) for providing tooling for the data annotation process and to Anouk Stein, MD and Dr. George Shih for user and technical support throughout the process.
+
+1. Wang X, Peng Y, Lu L, Lu Z, Bagheri M, Summers RM. ChestX-ray8: Hospital-scale Chest X-ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseases. IEEE CVPR 2017, [http://openaccess.thecvf.com/content_cvpr_2017/papers/Wang_ChestX-ray8_Hospital-Scale_Chest_CVPR_2017_paper.pdf](http://openaccess.thecvf.com/content_cvpr_2017/papers/Wang_ChestX-ray8_Hospital-Scale_Chest_CVPR_2017_paper.pdf)
+
+## Citation
+Jarrel Seah, Jen, Maggie, Meng Law, Phil Culliton, Sarah Dowd. (2020). RANZCR CLiP - Catheter and Line Position Challenge. Kaggle. https://kaggle.com/competitions/ranzcr-clip-catheter-line-classification
+
+
+# Dataset Description
+
+In this competition, you’ll detect the presence and position of catheters and lines on chest x-rays. Use machine learning to train and test your model on 40,000 images to categorize a tube that is poorly placed.
+
+## What files do I need?
+
+You will need the `train` and `test` images. This is a code-only competition so there is a hidden test set (approximately 4x larger, with ~14k images) as well.
+
+`train.csv` contains image IDs, binary labels, and patient IDs.
+
+TFRecords are available for both train and test. (They are also available for the hidden test set.)
+
+We've also included `train_annotations.csv`. These are segmentation annotations for training samples that have them. They are included solely as additional information for competitors.
+
+## Files
+
+- train.csv - contains image IDs, binary labels, and patient IDs.
+- sample_submission.csv - a sample submission file in the correct format
+- test - test images
+- train - training images
+Columns
+
+## StudyInstanceUID - unique ID for each image
+
+- `StudyInstanceUID` - unique ID for each image
+- `ETT - Abnormal` - endotracheal tube placement abnormal
+- `ETT - Borderline` - endotracheal tube placement borderline abnormal
+- `ETT - Normal` - endotracheal tube placement normal
+- `NGT - Abnormal` - nasogastric tube placement abnormal
+- `NGT - Borderline` - nasogastric tube placement borderline abnormal
+- `NGT - Incompletely Imaged` - nasogastric tube placement inconclusive due to imaging
+- `NGT - Normal` - nasogastric tube placement borderline normal
+- `CVC - Abnormal` - central venous catheter placement abnormal
+- `CVC - Borderline` - central venous catheter placement borderline abnormal
+- `CVC - Normal` - central venous catheter placement normal
+- `Swan Ganz Catheter Present`
+- `PatientID` - unique ID for each patient in the dataset

dslighting/registry/ranzcr-clip-catheter-line-classification/description_obfuscated.md

@@ -0,0 +1,53 @@
+# Task
+
+Detect the presence and position of catheters and lines on chest x-rays.
+
+# Metric
+
+Area under the ROC curve for each label, with the final score being the average of the individual AUCs of each predicted column.
+
+# Submission Format
+
+For each ID in the test set, you must predict a probability for all target variables. The file should contain a header and have the following format:
+```
+StudyInstanceUID,ETT - Abnormal,ETT - Borderline,ETT - Normal,NGT - Abnormal,NGT - Borderline,NGT - Incompletely Imaged,NGT - Normal,CVC - Abnormal,CVC - Borderline,CVC - Normal,Swan Ganz Catheter Present
+1.2.826.0.1.3680043.8.498.62451881164053375557257228990443168843,0,0,0,0,0,0,0,0,0,0,0
+1.2.826.0.1.3680043.8.498.83721761279899623084220697845011427274,0,0,0,0,0,0,0,0,0,0,0
+1.2.826.0.1.3680043.8.498.12732270010839808189235995393981377825,0,0,0,0,0,0,0,0,0,0,0
+1.2.826.0.1.3680043.8.498.11769539755086084996287023095028033598,0,0,0,0,0,0,0,0,0,0,0
+1.2.826.0.1.3680043.8.498.87838627504097587943394933987052577153,0,0,0,0,0,0,0,0,0,0,0
+1.2.826.0.1.3680043.8.498.53211840524738036417560823327351887819,0,0,0,0,0,0,0,0,0,0,0
+1.2.826.0.1.3680043.8.498.93555795394184819372299157360228027866,0,0,0,0,0,0,0,0,0,0,0
+1.2.826.0.1.3680043.8.498.52241894131170494723503100795076463919,0,0,0,0,0,0,0,0,0,0,0
+1.2.826.0.1.3680043.8.498.36500167484503936720548852591033878284,0,0,0,0,0,0,0,0,0,0,0
+1.2.826.0.1.3680043.8.498.86199852603457900780565655267977637728,0,0,0,0,0,0,0,0,0,0,0
+```
+
+# Dataset
+
+`train.csv` contains image IDs, binary labels, and patient IDs.
+
+TFRecords are available for both train and test.
+
+We've also included `train_annotations.csv`. These are segmentation annotations for training samples that have them. They are included solely as additional information for competitors.
+
+- train.csv - contains image IDs, binary labels, and patient IDs.
+- sample_submission.csv - a sample submission file in the correct format
+- test - test images
+- train - training images
+
+## Columns
+
+- `StudyInstanceUID` - unique ID for each image
+- `ETT - Abnormal` - endotracheal tube placement abnormal
+- `ETT - Borderline` - endotracheal tube placement borderline abnormal
+- `ETT - Normal` - endotracheal tube placement normal
+- `NGT - Abnormal` - nasogastric tube placement abnormal
+- `NGT - Borderline` - nasogastric tube placement borderline abnormal
+- `NGT - Incompletely Imaged` - nasogastric tube placement inconclusive due to imaging
+- `NGT - Normal` - nasogastric tube placement borderline normal
+- `CVC - Abnormal` - central venous catheter placement abnormal
+- `CVC - Borderline` - central venous catheter placement borderline abnormal
+- `CVC - Normal` - central venous catheter placement normal
+- `Swan Ganz Catheter Present`
+- `PatientID` - unique ID for each patient in the dataset

dslighting/registry/ranzcr-clip-catheter-line-classification/grade.py

@@ -0,0 +1,31 @@
+import numpy as np
+import pandas as pd
+from sklearn.metrics import roc_auc_score
+
+from mlebench.competitions.utils import prepare_for_auroc_metric
+from mlebench.grade_helpers import InvalidSubmissionError
+
+from .classes import CLASSES
+
+
+def grade(submission: pd.DataFrame, answers: pd.DataFrame) -> float:
+    for class_name in CLASSES:
+        assert class_name in answers.columns, f"Missing class {class_name} in answers."
+        if class_name not in submission.columns:
+            raise InvalidSubmissionError(f"Class {class_name} is not in the submission.")
+    assert len(submission) == len(
+        answers
+    ), f"Expected {len(answers)} rows in submission, got {len(submission)}."
+
+    roc_auc_scores = []
+    for class_name in CLASSES:
+        roc_auc_inputs = prepare_for_auroc_metric(
+            submission=submission,
+            answers=answers,
+            id_col="StudyInstanceUID",
+            target_col=class_name,
+        )
+        roc_auc_scores.append(roc_auc_score(roc_auc_inputs["y_true"], roc_auc_inputs["y_score"]))
+
+    score = np.average(roc_auc_scores)
+    return score

dslighting/registry/ranzcr-clip-catheter-line-classification/prepare.py

@@ -0,0 +1,53 @@
+import shutil
+from pathlib import Path
+
+from sklearn.model_selection import train_test_split
+
+from mlebench.utils import read_csv
+
+from .classes import CLASSES
+
+
+def prepare(raw: Path, public: Path, private: Path):
+    # Create train, test from train split
+    old_train = read_csv(raw / "train.csv")
+    new_train, new_test = train_test_split(old_train, test_size=0.1, random_state=0)
+
+    old_train_annotations = read_csv(raw / "train_annotations.csv")
+    old_train_uids = old_train_annotations["StudyInstanceUID"]
+    new_train_uids = new_train["StudyInstanceUID"]
+    is_in_new_train = old_train_uids.isin(new_train_uids)
+
+    new_train_annotations = old_train_annotations[is_in_new_train]
+
+    (public / "train").mkdir(exist_ok=True)
+    (public / "test").mkdir(exist_ok=True)
+
+    for file_id in new_train["StudyInstanceUID"]:
+        shutil.copyfile(
+            src=raw / "train" / f"{file_id}.jpg",
+            dst=public / "train" / f"{file_id}.jpg",
+        )
+
+    for file_id in new_test["StudyInstanceUID"]:
+        shutil.copyfile(
+            src=raw / "train" / f"{file_id}.jpg",
+            dst=public / "test" / f"{file_id}.jpg",
+        )
+
+    assert len(list(public.glob("train/*.jpg"))) == len(
+        new_train
+    ), f"Expected {len(new_train)} files in public train, got {len(list(public.glob('train/*.jpg')))}"
+    assert len(list(public.glob("test/*.jpg"))) == len(
+        new_test
+    ), f"Expected {len(new_test)} files in public test, got {len(list(public.glob('test/*.jpg')))}"
+
+    # Create a sample submission file
+    submission_df = new_test[["StudyInstanceUID"] + CLASSES]
+    submission_df[CLASSES] = 0
+
+    # Copy over files
+    new_train.to_csv(public / "train.csv", index=False)
+    new_train_annotations.to_csv(public / "train_annotations.csv", index=False)
+    new_test.to_csv(private / "test.csv", index=False)
+    submission_df.to_csv(public / "sample_submission.csv", index=False)

dslighting/registry/ranzcr-clip-catheter-line-classification/prepare_val.py

@@ -0,0 +1,113 @@
+import shutil
+from pathlib import Path
+
+import pandas as pd
+from sklearn.model_selection import train_test_split
+
+from mlebench.utils import read_csv
+
+from .classes import CLASSES
+
+
+def _create_split_files(
+    train_df: pd.DataFrame,
+    test_df: pd.DataFrame,
+    all_annotations_df: pd.DataFrame,
+    public_dir: Path,
+    private_dir: Path,
+    raw_image_dir: Path,
+):
+    """
+    Helper function to populate public and private directories for a given data split.
+
+    This function handles directory creation, image copying, and CSV file generation,
+    ensuring a consistent output structure.
+    """
+    # Ensure output directories exist
+    public_dir.mkdir(parents=True, exist_ok=True)
+    private_dir.mkdir(parents=True, exist_ok=True)
+    (public_dir / "train").mkdir(exist_ok=True)
+    (public_dir / "test").mkdir(exist_ok=True)
+
+    # Filter annotations to only include those for the current training set
+    train_uids = train_df["StudyInstanceUID"]
+    is_in_train = all_annotations_df["StudyInstanceUID"].isin(train_uids)
+    split_train_annotations = all_annotations_df[is_in_train]
+
+    # Copy image files for the current train and test sets
+    for file_id in train_df["StudyInstanceUID"]:
+        shutil.copyfile(
+            src=raw_image_dir / f"{file_id}.jpg",
+            dst=public_dir / "train" / f"{file_id}.jpg",
+        )
+
+    for file_id in test_df["StudyInstanceUID"]:
+        shutil.copyfile(
+            src=raw_image_dir / f"{file_id}.jpg",
+            dst=public_dir / "test" / f"{file_id}.jpg",
+        )
+
+    # Assert that the correct number of images were copied
+    assert len(list(public_dir.glob("train/*.jpg"))) == len(
+        train_df
+    ), f"Expected {len(train_df)} files in {public_dir}/train, got {len(list(public_dir.glob('train/*.jpg')))}"
+    assert len(list(public_dir.glob("test/*.jpg"))) == len(
+        test_df
+    ), f"Expected {len(test_df)} files in {public_dir}/test, got {len(list(public_dir.glob('test/*.jpg')))}"
+
+    # Create a sample submission file for the current test set
+    submission_df = test_df[["StudyInstanceUID"] + CLASSES].copy()
+    submission_df[CLASSES] = 0
+
+    # Save all required CSV files with the required standard filenames
+    train_df.to_csv(public_dir / "train.csv", index=False)
+    split_train_annotations.to_csv(public_dir / "train_annotations.csv", index=False)
+    submission_df.to_csv(public_dir / "sample_submission.csv", index=False)
+    test_df.to_csv(private_dir / "test.csv", index=False)
+
+
+def prepare(raw: Path, public: Path, private: Path):
+    # Load raw data once
+    old_train = read_csv(raw / "train.csv")
+    old_train_annotations = read_csv(raw / "train_annotations.csv")
+    raw_image_dir = raw / "train"
+
+    # --- 1. Original Split: Create main train and test sets ---
+    # This split creates the primary train/test sets for the competition.
+    # The outputs in `public` and `private` must remain identical to the original script.
+    new_train, new_test = train_test_split(old_train, test_size=0.1, random_state=0)
+
+    # Use the helper to create the original public/private directory contents
+    _create_split_files(
+        train_df=new_train,
+        test_df=new_test,
+        all_annotations_df=old_train_annotations,
+        public_dir=public,
+        private_dir=private,
+        raw_image_dir=raw_image_dir,
+    )
+
+    # --- 2. New Validation Split: Create a validation set from the main train set ---
+    # Define paths for the new validation split directories.
+    public_val = public.parent / "public_val"
+    private_val = private.parent / "private_val"
+
+    # Split the main training set (`new_train`) again to create a smaller
+    # training set and a validation set.
+    # The validation set (`test_val`) size should be ~10% of the *original* data.
+    # new_train size = 0.9 * total. test_val_size = (0.1 * total) / (0.9 * total) = 1/9.
+    test_val_size = 1 / 9
+    train_val, test_val = train_test_split(
+        new_train, test_size=test_val_size, random_state=0
+    )
+
+    # Use the same helper to create the new validation directories (`public_val`, `private_val`)
+    # The helper ensures filenames and structure are identical to the original outputs.
+    _create_split_files(
+        train_df=train_val,
+        test_df=test_val,
+        all_annotations_df=old_train_annotations,
+        public_dir=public_val,
+        private_dir=private_val,
+        raw_image_dir=raw_image_dir,
+    )

dslighting/registry/rsna-2022-cervical-spine-fracture-detection/checksums.yaml

@@ -0,0 +1,8 @@
+private:
+  answers.csv: 9d13e08d12211a39b8c5d2c29a9e7e8f
+public:
+  sample_submission.csv: 2690b6c1b8c2fbc8d090e8fb00c1d79f
+  test.csv: d5c91b7191186743a2aa74b904207931
+  train.csv: 812aecfc08948d621decc684759e243a
+  train_bounding_boxes.csv: d74270bdb517e79f9f2ac6a0f90825fd
+zip: 2b38d6c8f3f357d56a247b09769e862b

dslighting/registry/rsna-2022-cervical-spine-fracture-detection/config.yaml

@@ -0,0 +1,32 @@
+id: rsna-2022-cervical-spine-fracture-detection
+name: RSNA 2022 Cervical Spine Fracture Detection
+competition_type: code
+awards_medals: true
+prizes:
+  - position: 1
+    value: 6000
+  - position: 2
+    value: 5000
+  - position: 3
+    value: 4000
+  - position: 4
+    value: 3000
+  - position: 5
+    value: 3000
+  - position: 6
+    value: 3000
+  - position: 7
+    value: 3000
+  - position: 8
+    value: 3000
+description: mlebench/competitions/rsna-2022-cervical-spine-fracture-detection/description.md
+
+dataset:
+  answers: rsna-2022-cervical-spine-fracture-detection/prepared/private/answers.csv
+  sample_submission: rsna-2022-cervical-spine-fracture-detection/prepared/public/sample_submission.csv
+
+grader:
+  name: weighted-multi-label-log-loss
+  grade_fn: mlebench.competitions.rsna-2022-cervical-spine-fracture-detection.grade:grade
+
+preparer: mlebench.competitions.rsna-2022-cervical-spine-fracture-detection.prepare:prepare