Affordable Access

Analysis of 31P MR spectroscopy data using artificial neural networks for longitudinal evaluation of muscle diseases: dermatomyositis.

Authors
  • Park, J H
  • Kari, S
  • King, L E Jr
  • Olsen, N J
Type
Published Article
Journal
NMR in biomedicine
Publication Date
Jan 01, 1998
Volume
11
Issue
4-5
Pages
245–256
Identifiers
PMID: 9719579
Source
Medline
License
Unknown

Abstract

Classical myopathic dermatomyositis (DM) is a chronic autoimmune disease characterized by an erythematous rash and severe, proximal muscle weakness. A disease variant, amyopathic DM, presents with the typical rash but without clinical evidence of muscle weakness. Prednisone and immunosuppressive drugs alleviate symptoms in many patients. Accurate longitudinal evaluations of patients are important to limit serious side effects of these drugs, including osteoporosis, cataracts, and growth inhibition. Metabolic abnormalities detected with 31P magnetic resonance spectroscopy (MRS) provide the best quantitative data for evaluating these patients. With 31P MRS, the levels of inorganic phosphate (Pi), phosphocreatine (PCr), ATP, and phosphodiesters (PDE) were determined in the quadricep muscles of patients during rest and exercise. Artificial neural network (ANN) analyses of these data were previously used for accurate classification of patients with myopathic or amyopathic DM and normal controls. In the present investigation, an artificial neural network was employed for further analysis of the 31P metabolite levels in quantitative, longitudinal evaluations of the extent (percent) of clinical improvement or deterioration during treatment with prednisone and immunosuppressive drugs. The ANN results showed that adult patients in a severe myopathic state could improve with treatment to a clinical status of amyopathic DM. In contrast, severely weak juvenile patients in the myopathic state recovered to normal status. One juvenile patient did not improve and remained in the myopathic state. Additionally, a serious clinical relapse in an amyopathic patient was predicted with serial ANN analyses well in advance of the actual clinical event. These network analyses show potential utility for clinical applications in muscle diseases.

Report this publication

Statistics

Seen <100 times