Cancer evolution assessment using artificial neural networks.

Artificial Neural Networks have demonstrated good utility to establish disease prognostic, if there are correctly trained. Our aim was to realize Artificial Neural Networks in different cancer types, to evaluate time survival. For the beginning this was applied for breast cancer. The conclusion is that the homogeneity of data entrance sets, the number, and their coding in relation with their importance for cancer prognostic are decisive for the results of the trained ANN. These results can be useful in the physician clinical decision.

[A retrospective study of bone metastases distribution on 420 whole body scans]. / Studiu retrospectiv al distributiei metastazelor osoase pe 420 de scintigrame "corp intreg".

UNLABELLED: 420 patients with cancer of breast, prostate, lung and colon were investigated through 99mTc Methylene-diphosphonate whole body scintigraphy. The presence of pathologic radiotracer uptake was qualitatively and quantitatively analyzed in order to establish the metastases distribution. Patient selection was realized over 2455 whole body scintigraphies effectuated between 1998 and 2001 in our Nuclear Medicine Service. All selected cases were metastases with known origin primary cancer. RESULTS: Using the Qui-square Test we have compared the frequency of nine well delimited skeleton regions involved like metastatic site in the different cancer types. We have found a significant statistic difference of the range of frequency only between breast and prostate cancers as well as between pulmonary and prostate cancers. The mean number of the metastatic involved skeletal regions was significantly greater in breast and prostate cancers in comparison with lung and colon cancers (p < 0.0001). The higher metastases site frequency was the rachis, than the pelvis, the ribs and the sternum. The skull metastases localization is more frequent in breast cancer in comparison with all other cancers (7.67% versus less than 4% in other cancer types). The pelvis was more involved for the metastatic process in prostate neoplasm. On the other hand, the highest mean anatomic sites number per patient was found in breast cancer (5.7) and prostate cancer (4.8) related to colon (3.3) and lung (3.0) cancers. CONCLUSIONS: Even some particularities were evident, in our study, between the metastases distribution in these four cancer types, the data are not sufficient to sustain the existence of a characteristic pattern related to the primary cancer origin. Metastases localization could be, however, related to the metastazation mechanism.

Image processing in myocardial SPECT acquisition.

Image processing for the imagistic medicine domain nuclear medicine have long time being realised through specifical protocols which rarely allowed the intervention of other informatical modalities. Processing depends first of all on the acquisition type. This paper presents the steps of one of the most complex image processing protocols, in the case of myocardial SPECT acquisition (Single Photon Emission Computed Tomography).

[18F FDG PET-Applications in Oncology]. / 18F FDG PET--aplicatii în oncologie (II).

In the first part our intention was, essentially, to present the particularities of glucose tumoral cells metabolism, PET components, the synthesis of 18F FDG and the detection of unknown cancers. This second part makes reference about mainly types of tumors who benefit by FDG-PET indications. Clinical PET has a rapid growth because of its use in cancer diagnosis and management. According with published studies all over the world, the sensibility and specificity of FDG-PET, noninvasive method, is higher than that of the conventional methods like CT, IRM, ultrasonography. PET is en excellent detection method of most of common cancer types and depends not on the histological neoplasm type; the more aggressive is the tumor, more it will uptake the radiotracer. The cost is significant, so the indications must be very precise: evaluating the malignity of solitary pulmonary nodules, evaluating the recurrences of melanoma, colon cancer diagnosis, differentiation between recurrent brain tumor and radiation injury, differential diagnosis of the benign lymph and malign lymph nodes, staging of Hodgkin's and non-Hodgkin's lymphoma, evaluation the response to therapy. Because the PET images are difficult to interpret, appears the necessity of correlation with anatomic images: this was the fusion images beginnings (the PET and CT images combination); now the physiologic information has precise anatomic localization. The growing of this method is very probably, both using 18F FDG -thanks to its highly favorable physical characteristics- and other new radiopharmaceuticals. The clinical cases that illustrate the applications are investigated at CERMEP, Lyon, France.