Search Plan Generation-Based Segmentation of Medical Radiographic Images
| Project/Area Number |
07680408
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | Kyushu Institute of Design (1997)
Kyushu Institute of Technology (1995-1996) |
|---|
Principal Investigator |
FUKUSHIMA Shigehiro Kyushu Institute of Design, Graduate School, Professor, 芸術工学研究科, 教授 (60027927)
|
|---|
| Project Period (FY) |
1995 – 1997
|
| Project Status |
Completed (Fiscal Year 1997)
|
| Budget Amount *help |
¥2,300,000 (Direct Cost: ¥2,300,000)
Fiscal Year 1997: ¥700,000 (Direct Cost: ¥700,000)
Fiscal Year 1996: ¥600,000 (Direct Cost: ¥600,000)
Fiscal Year 1995: ¥1,000,000 (Direct Cost: ¥1,000,000)
|
|---|
| Keywords | Gastric radiograms / Radiograms of temporomandubular joint / Image understanding / Region segmentation / Search / Active contour / Optimization / Dynamic programming / 画像認識 / コンピュータ支援画像診断 / 医用画像 / システム構築 / ハフ変換 / 画像理解 / 詳細化 |
|---|
| Research Abstract |
Methods have been developed for region segmentation on the basis of search plan generation. The objects are the organs in the gastric radiograms and in the radiographic images of the temporomandibular joint. The gastric radiograms are of good quality but its structure is complex, while the radiograms of the mandibular joint are rather of a simple structure but their quality is poor. The methods adopted the search strategy of "Global-to-Fine" : i.e. generating the search plan flrst, and then finding the detail. The concrete methods of segmentation were designed by problem solving approach adapted to the properties of the respective regions. For the gastric radiograms, the region of the stomach was found by a line-tracking method which refers to the guide image of extracted feature elements. It was also found by an optimization method which uses the active contour and the dynamic programming. The barium pool region was found by an edge-based method which uses the zero-crossing, without using the search plan. It was also found by first grouping the concave points of the gray-level and then by region growing. The spinal region was found by excluding the interfererence from the barium pool region and by applying the Hough transformation to coarsely extracting the spinal sides. The recognition methods for the individual regions were integrated to construct an image understanding system. The methods achieved good performance of recognition of the regions. For the radiograms of the temporomandibular joint, the models were positively used from the early stages of the process of preprocessing, edge enhancement, and the following optimization by dynamic programming. The methods achieved good performance of tracing of the region of interest. An indication was given for the measurement of the region of interest.
|
Report
(4 results)
Research Products
(28 results)
