ianpan
/

mammoscreen

@@ -1,199 +1,96 @@
 ---
 library_name: transformers
-tags: []
 ---
-# Model Card for Model ID
-<!-- Provide a quick summary of what the model is/does. -->
-## Model Details
-### Model Description
-<!-- Provide a longer summary of what this model is. -->
-This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
-- **Developed by:** [More Information Needed]
-- **Funded by [optional]:** [More Information Needed]
-- **Shared by [optional]:** [More Information Needed]
-- **Model type:** [More Information Needed]
-- **Language(s) (NLP):** [More Information Needed]
-- **License:** [More Information Needed]
-- **Finetuned from model [optional]:** [More Information Needed]
-### Model Sources [optional]
-<!-- Provide the basic links for the model. -->
-- **Repository:** [More Information Needed]
-- **Paper [optional]:** [More Information Needed]
-- **Demo [optional]:** [More Information Needed]
-## Uses
-<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
-### Direct Use
-<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
-[More Information Needed]
-### Downstream Use [optional]
-<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
-[More Information Needed]
-### Out-of-Scope Use
-<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
-[More Information Needed]
-## Bias, Risks, and Limitations
-<!-- This section is meant to convey both technical and sociotechnical limitations. -->
-[More Information Needed]
-### Recommendations
-<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
-Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
-## How to Get Started with the Model
-Use the code below to get started with the model.
-[More Information Needed]
-## Training Details
-### Training Data
-<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
-[More Information Needed]
-### Training Procedure
-<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
-#### Preprocessing [optional]
-[More Information Needed]
-#### Training Hyperparameters
-- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
-#### Speeds, Sizes, Times [optional]
-<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
-[More Information Needed]
-## Evaluation
-<!-- This section describes the evaluation protocols and provides the results. -->
-### Testing Data, Factors & Metrics
-#### Testing Data
-<!-- This should link to a Dataset Card if possible. -->
-[More Information Needed]
-#### Factors
-<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
-[More Information Needed]
-#### Metrics
-<!-- These are the evaluation metrics being used, ideally with a description of why. -->
-[More Information Needed]
-### Results
-[More Information Needed]
-#### Summary
-## Model Examination [optional]
-<!-- Relevant interpretability work for the model goes here -->
-[More Information Needed]
-## Environmental Impact
-<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
-Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
-- **Hardware Type:** [More Information Needed]
-- **Hours used:** [More Information Needed]
-- **Cloud Provider:** [More Information Needed]
-- **Compute Region:** [More Information Needed]
-- **Carbon Emitted:** [More Information Needed]
-## Technical Specifications [optional]
-### Model Architecture and Objective
-[More Information Needed]
-### Compute Infrastructure
-[More Information Needed]
-#### Hardware
-[More Information Needed]
-#### Software
-[More Information Needed]
-## Citation [optional]
-<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
-**BibTeX:**
-[More Information Needed]
-**APA:**
-[More Information Needed]
-## Glossary [optional]
-<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
-[More Information Needed]
-## More Information [optional]
-[More Information Needed]
-## Model Card Authors [optional]
-[More Information Needed]
-## Model Card Contact
-[More Information Needed]

 ---
 library_name: transformers
+tags:
+- mammography
+- cancer
+- breast_cancer
+- radiology
+- breast_density
+license: apache-2.0
+base_model:
+- timm/tf_efficientnetv2_s.in21k_ft_in1k
+pipeline_tag: image-classification
 ---
+This is an ensemble model for predicting breast cancer and breast density based on screening mammography.
+The model uses 3 basic CNNs (`tf_efficientnetv2_s` backbone) and performs inference on each provided image (i.e., CC and MLO view).
+Each net in the ensemble uses a different resolution: 2048 x 1024, 1920 x 1280, and 1536 x 1536.
+The final outputs are averaged together across the provided views and the neural nets.
+The model can also perform inference on a single view (image), although performance will be decreased.
+A hybrid classification-segmentation model was first pretrained on the Curated Breast Imaging Subset of Digital Database for Screening Mammography
+[(CBIS-DDSM)](https://www.cancerimagingarchive.net/collection/cbis-ddsm/). This dataset contains film mammography studies
+(as opposed to digital) with accompanying ROI annotations for benign and malignant masses and calcifications.
+The resultant model was further trained on data from the [RSNA Screening Mammography Breast Cancer Detection challenge](https://www.kaggle.com/competitions/rsna-breast-cancer-detection/).
+The data was split into 80%/10%/10% train/val/test. Evaluation was performed on the 10% holdout test split.
+This procedure was repeated 3 separate times to better assess the model's performance.
+The provided weights are from the first data split.
+Exponential moving averaging was used during training and increased performance.
+Note that the model was trained using cropped images, and thus it is recommended to crop the image prior to inference.
+A cropping model is provided here: https://huggingface.co/ianpan/mammo-crop
+The primary evaluation metric is the area under the receiver operating characteristic curve (AUC/AUROC).
+Below are the average and standard deviation across the 3 splits.
+```
+Split 1: 0.9464
+Split 2: 0.9467
+Split 3: 0.9422
+Mean (std.): 0.9451 (0.002)
+```
+As this is a screening test, high sensitivity is desirable. We also calculate the specificity
+at varying sensitivities, shown below (averaged across 3 splits):
+```
+Sensitivity: 98.1%, Specificity: 65.4% +/- 7.2%
+Sensitivity: 94.3%, Specificity: 78.7% +/- 0.9%
+Sensitivity: 90.5%, Specificity: 84.8% +/- 2.7%
+```
+Example usage:
+```
+import cv2
+from transformers import AutoModel
+model = AutoModel.from_pretrained("ianpan/mammoscreen", trust_remote_code=True)
+model = model.eval().to("cuda:0")
+cc_img = cv2.imread("mammo_cc.png", cv2.IMREAD_GRAYSCALE)
+mlo_img = cv2.imread("mammo_mlo.png", cv2.IMREAD_GRAYSCALE)
+with torch.inference_mode():
+  output = model({"cc": cc_img, "mlo": mlo_img}, device="cuda:0")
+```
+Note that the model preprocesses the data within the `forward` function into the necessary format.
+`output` is a dictionary containing two keys: `cancer` and `density`. `output['cancer']` is a tensor of shape (N, 1) and `output['density']` is a tensor of shape (N, 4).
+If you want the predicted density class, take the argmax: `output['density'].argmax(1)`. If only a single study is provided, then N=1.
+You can also access each neural net separately using `model.net{i}`. However, you must apply the preprocessing outside of the `forward` function.
+```
+input_dict = model.net0.preprocess({"cc": cc_img, "mlo": mlo_img}, device="cuda:0")
+with torch.inference_mode():
+  out = model.net0(input_dict)
+```
+The model also supports batch inference. Construct a dictionary for each breast and pass a list of dictionaries to the model.
+For example, if you want to perform inference for each breast for 2 patients (`pt1`, `pt2`):
+```
+cc_images = ["rt_pt1_cc.png", "lt_pt1_cc.png", "rt_pt2_cc.png", "lt_pt2_cc.png"]
+mlo_images = ["rt_pt1_mlo.png", lt_pt1_mlo.png", "rt_pt2_mlo.png", "lt_pt2_mlo.png"]
+cc_images = [cv2.imread(_, cv2.IMREAD_GRAYSCALE) for _ in cc_images]
+mlo_images = [cv2.imread(_, cv2.IMREAD_GRAYSCALE) for _ in mlo_images]
+input_dict = [{"cc": cc_img, "mlo": mlo_img} for cc_img, mlo_img in zip(cc_images, mlo_images)]
+with torch.inference_mode():
+  output = model(input_dict, device="cuda:0")
+```