In this activity, students will use MATLAB to understand DICOM files, analyze them through MATLAB in both images and metadata. MATLAB Imaging Toolbox comes with its own DICOM files. However, students will also download some public DICOM studies from the Cancer Imaging Archive (TCIA). Using the DICOM images, students learn the richness of DICOM images and their metadata through MATLAB and the MATLAB Medical Imaging Toolbox. The goal of this activity is to prepare the students for their use of MATLAB with DICOM. However, more interesting complex tasks, such as data mining, digital image processing, and deep learning, on the images can be performed by other MATLAB libraries after processing these DICOM files using the Medical Imaging Toolbox.

Students are expected to remove artifacts and blurs/motion from the DICOM files using MATLAB Image Processing Toolbox in conjunction with MATLAB Medical Imaging Toolbox.

This teaching activity functions as a preparation for the students to use DICOM radiographic images in Data Mining. This assumes the students are already familiar with using MATLAB at a basic level. However, it does not assume prior experience with DICOM or the DICOM functionalities of MATLAB. While this proposed activity is run during a lab (that lasts 75 minutes), subsequent Data Mining tasks are offered as a take-home, multi-day task. This is a Data Mining, upper-level elective. Students are introduced to frameworks, including Python with Pandas, NumPy, scikit-learn, Matplotlib with its add-ons, and KNIME. The students then perform a peer-learning activity using one or more of these frameworks to present a data mining application. The students have a significant degree of freedom in this project-based learning activity. The artifacts in the DICOM images, in the context of data mining and data cleaning, are considered anomalies. Thus, this project is an anomaly detection task.

This lab serves as an introduction to DICOM imaging in radiology and processing the DICOM images using MATLAB. Additional pointers to MATLAB are provided as links under the Resources or can be found online.

This is part of the course, "Data Mining," an undergraduate upper-level elective course. The typical class size ranges from 10 to 15 students. It is also a stacked course (i.e., also offered to graduate students with some additional expectations), although usually only up to two grad students would join these courses, as there are not many grad students in the program.

This is a classroom activity, as the course does not have separate lab sessions. However, the class room activities are diverse in this course and this particular session is performed as a lab activity in a regular lecture room.

There is no prior experience with MATLAB expected from this class beyond a simple introduction provided to the students, with the expectation that they will install MATLAB and get familiarized with MATLAB and DICOM beforehand by going through the resources (included at the end of this activity). However, the students are senior-level undergraduates in computer science and computer systems engineering, and they are familiar with programming. The instructor will introduce DICOM and the associated toolkits.

MATLAB is provided to students and educators free of charge under the education license. The students should have installed MATLAB on their laptops and have a basic understanding of MATLAB. Students do not have access to MATLAB-installed machines in this course as the labs are run during regular lectures. But students are expected to bring in their personal laptops.

MATLAB Medical Imaging toolbox enables searching, reading, and writing to DICOM folders and DICOM volumes. MATLAB fully supports DICOM, including the storage and networking capabilities.

While this activity is designed for a Data Mining upper-level elective course, it can be easily adapted for other computer science upper-level courses, such as Machine Learning or Image Processing, or graduate school programs, such as Biomedical Informatics. As an extension, this activity can be extended into a multi-day take-home assignment by incorporating medical image analysis applications with deep learning, using the provided examples. The proposed task of removing artifacts in the DICOM images is only a sample activity. Students are free to choose alternative tasks.
The success of this activity depends on the instructor's familiarity with DICOM, TCIA, and the NBIA retriever, as well as some basic knowledge of radiology. However, depending on the comfort level of the instructor and the classroom, TCIA can be skipped, limiting the task to just existing DICOM images in MATLAB or from other open data archives where it is easier to retrieve the images. The provided resources can be used as a mandatory reading before this class to help with the in-class instructions.

Students are assessed at three levels. First, as part of the in-class lab activity, the instructor will evaluate the students informally on their progress in viewing, analyzing, and creating DICOM images from studies downloaded from the TCIA. The second part is their successful use of MATLAB in enhancing the quality of the DICOM images. The third part is the successful use of MATLAB as part of their Peer Learning Activity. There is no grading for this in-class activity. However, students will provide feedback in the form of a brief 1-minute survey to indicate what they learned from this session.

[1] DICOM Support in Image Processing Toolbox.
https://www.mathworks.com/help/images/dicom-support-in-the-image-processing-toolbox.html
This page introduces the DICOM Toolbox to the readers without expecting any experience or familiarity. It starts with working with DICOM using MATLAB for students at a basic level. But the links included in this page further go into advanced topics which the students will find particularly useful after getting familiar with processing DICOM files with MATLAB.

[2] The Cancer Imaging Archive (TCIA). https://www.cancerimagingarchive.net/
TCIA is an open and public archive of radiographic images. These images come from various modalities such as MRI, CT, X-Ray, and Ultrasound, and are curated into anonymized public and limited collections. Each collection contains the citation details and the associated research paper for further information. The images can be viewed online from the TCIA DICOM viewer. The collections can be downloaded in bulk, or each individual study in the collection can be downloaded individually. At the finer level, image series that make up the studies can be viewed and downloaded for quick assessments. Some images are marked as "Species: Phantom." They must not be used. Only those that are marked "Species: Human" are real patient images.

[3] DICOM Standard Browser https://dicom.innolitics.com/
This site provides a comprehensive list of DICOM attributes of each modality. The Tag uniquely identifies each of these metadata attributes, whereas a Keyword provides a human-readable brief description of the attribute. Each attribute is elaborated on its own page. For example, https://dicom.innolitics.com/ciods/computed-radiography-image/patient/00100020 details the Patient-level attribute, 00100020 (i.e., Tag = 0010,0020), which is the PatientID attribute.

[4] NBIA Data Retriever
https://wiki.cancerimagingarchive.net/display/NBIA/Downloading+TCIA+Images
This tool allows students to download studies separately and independently, rather than downloading a whole collection of hundreds of GBs. The students should download and install this toolkit to allow them to download DICOM images from the TCIA.

[5] Read, Process, and Write 3-D Medical Images with Spatial Referencing https://www.mathworks.com/help/medical-imaging/ug/read-process-and-write-3-d-medical-image-with-spatial-referencing.html
This page elaborates how to read, write, and process 3-D medical images.

[6] Medical Imaging Toolbox https://www.mathworks.com/products/medical-imaging.html This page provides an overview of the medical imaging toolbox, describing how to visualize, register, segment, and label 2D and 3D medical images.

[7] Image Processing Toolbox https://www.mathworks.com/products/image-processing.html This page provides an overview of the Image Processing Toolbox, describing how to perform image processing, visualization, and analysis.

[8] Using MATLAB with Python https://www.mathworks.com/products/matlab/matlab-and-python.html This page provides insights into MATLAB and Python integration. This page can be helpful for students who choose to develop their projects partially in Python and then integrate them with the MATLAB components. This is particularly useful since many students are familiar with Python and Python data mining and image processing libraries, and such an integration allows them to develop certain features in MATLAB while utilizing their existing code in Python.