170 03 Medical Imaging Informatics Introductory Comments n

#170. 03: Medical Imaging Informatics Introductory Comments n Goals: 1. Describe modern tools for processing and analyzing large amounts of imaging data. 2. Describe strategies to utilize effectively information from these large sets of imaging and metadata. 3. Provide examples of the use of these tools in a current research setting 4. Use easily obtainable and extensible open source tools (e. g. R and Weka) 5. Point to important literature in this field. 1

Medical Imaging Informatics Introductory Comments n Disclaimers: – Course aimed at broad audience so: § Technical types may expect more rigor § Non-technical types may occasionally feel overwhelmed – Course is experimental § Intent is to provide flavor of current research so there is no textbook (but based on how successful course is there may eventually be) § Direction can vary based on student input 2

Medical Imaging Informatics Introductory Comments n Disclaimers: – Course is team taught: § Satisfies individual teaching requirements ! § May experience some discontinuity but: – All lecturers are from same lab and have regular meetings re. course content – Individual lecturers bring particular expertise – Course will focus on MRI data § Data mining and statistical techniques introduced in course are data neutral but want to stress importance of knowing ones data so we will use data we know best for illustration 3

Medical Imaging Informatics Introductory Comments n Disclaimers: – Course is team taught: § Satisfies individual teaching requirements ! § May experience some discontinuity but: – All lecturers are from same lab and have regular meetings re. course content – Individual lecturers bring particular expertise – Course will focus on MRI data § Data mining and statistical techniques introduced in course are data neutral but want to stress importance of knowing ones data so we will use data we know best for illustration 4

Medical Imaging Informatics A Brief Example n MRI data and metadata on PTSD patients – Imaging: § § Hippocampal volume from structural MRI Intracranial volume from structural MRI Metabolite concentrations from Spectroscopic Imaging (SI) Tissue composition of SI voxels – Metadata: § § Age Gender Education CAPS score (estimated PTSD severity based on psychological exam) 5

Medical Imaging Informatics A Brief Example n Image Reconstruction and Processing – Reconstruction: § Formation from truncated sampling § Restoration for distortions and noise – Processing: § § § Segmentation and Classification Registration, i. e. between different image modalities Spatial normalization, i. e. for group analysis 6

Medical Imaging Informatics A Brief Example n Image – – – data Analysis Data Visualization Unbiased Quality Assessment Hypothesis driven and exploratory analyses Design of statistical models Data Mining 7

Medical Imaging Informatics A Brief Example n Decision Tree: <= 13 Education <=3. 01. Rt Hipp Vol >3. 01 Age PTSD- <=2. 66 > 13 Lt Hipp Vol >2. 66 PTSD+ Rt Hipp Vol <=43 PTSD- <=2. 43 >43 PTSD+ >2. 43 PTSD- Education <= 15 PTSD- > 15 Intracr Vol <= 1403 PTSD- > 1403 Age <= 33 PTSD- > 33 PTSD+ 8

Medical Imaging Informatics A Brief Example n Analysis using WEKA (10 times, 10 fold cross validation) – prediction accuracy: 9

Tentative Syllabus 10

Instructors 11

Teaching Material Course material will be available on the web or by handouts n http: //www. cind. research. va. gov/teaching/medical_imag ing. asp n 12