PITA Fiscal Year 2009 Projects - Biomedical and Health Engineering

Towards noninvasive radiological diagnosis by automated learning from MRI and histology image data

Principal Investigators: Gustavo K. Rohde, Jelena Kovacevic

Within the realm of clinical medicine, pathology is the closest approximation to the true nature of disease and its manifestation. An organ can be removed and its very finest detail examined at the cellular, subcellular, and molecular levels. The truth, however, comes at a price---an “invasive procedure;” tissue needs to be removed for examination. Removal of tissue is usually performed after a noninvasive radiological imaging procedure (typically MRI) has demonstrated an abnormality and the radiologist has generated a list of diagnostic possibilities. While several procedures for extracting tissue samples exist, they share several similarities: They are invasive, have morbidity that includes hemorrhage, pain and in rare instances organ loss, and are invariably associated with significant patient and family anxiety. Thus:

We propose to develop computational and mathematical approaches to recapitulate histological presentation of tissue from magnetic resonance images, with the ultimate goal of developing accurate classifiers capable of detecting malignant and benign tissue-types directly from MR image data, avoiding the invasive procedures used to procure tissue for histopathology studies.

By understanding the relationship between the signal information of histological section and MRIs through automated learning approaches we propose to develop a method for “virtual” histology whereby an estimate of what a histological image would “look like” will be produced based entirely on MRIs. In this proposal, we describe an important step towards achieving our aim: spatially aligning MR and histology images so that the corresponding information can be mined. The proposed research will be conducted by using embryonic stem cell derived teratomas as a testbed, in collaboration with clinicians from the University of Pittsburgh Medical School.