Klinische Neurophysiologie 2014; 45 - P49
DOI: 10.1055/s-0034-1371262

Diagnosis of neuromuscular diseases using a novel questionnaire and data mining applications. Science or fiction?

K Kollewe 1, S Petri 1, R Dengler 1, U Schuhmacher 2, W Lechner 3, F Klawonn 4, 5, L Grigull 6
  • 1Medizinische Hochschule Hannover, Neurologie, Hannover, Deutschland
  • 2DRK Clementinenkrankenhaus, Hannover, Deutschland
  • 3Improved Medical Diagnostics, Singapore, Deutschland
  • 4Helmholtz centre for infection research, Braunschweig, Deutschland
  • 5Ostfalia university of applied sciences, Wolfenbüttel, Deutschland
  • 6Department of pediatric hematology and oncology, Hannover, Deutschland

Introduction: The diagnosis of neuromuscular diseases is challenging. Especially rare diseases such as Pompe disease are frequently diagnosed with delay or are even misdiagnosed. We therefore developed a questionnaire based and data mining supported tool for diagnosing selected neuromuscular diseases (e.g. muscular dystrophy, storage diseases, amyotrophic lateral sclerosis (ALS)) to facilitate the diagnosis and conducted a monocentric pilot study.

Material and methods: First, interviews with patients were conducted focussing on their pre-diagnostic experiences. Likewise the patients' view and expertise was collected, extracted and categorised. Based on these observations we developed a new questionnaire containing 46 questions. In a second step, patients with neuromuscular diseases were asked to answer this questionnaire. Accordingly, the patients' observations during the pre-diagnostic phase were collected in a data base. The patients were contacted via self-help groups and the neuromuscular outpatient clinic of Hannover Medical School.

Results: A total of 172 patients answered the questionnaire, 28 healthy individuals served as controls. With this set of data, a novel computer algorithm was trained to classify answer-patterns. Using 8 different data mining applications and a fusion algorithm, the computer made a diagnostic suggestion. This diagnostic tool worked well for patients with Pompe disease (98% correct diagnoses), for patients with spinal muscular atrophy (98% correct diagnoses) and ALS-patients (95% correct diagnoses). All questions or rather all answers increased the diagnostic accuracy of the system. Receiver operating characteristic (ROC) analyses confirmed the excellent sensitivity of this diagnostic tool.

Discussion: For patients with muscular weakness, a questionnaire based diagnostic support tool using data mining application exhibited good results in making a correct diagnostic suggestion. This tool might be a valuable aid for earlier diagnosis of rare neuro-muscular diseases such as Pompe disease when patients present with rather unspecific symptoms at the general practitioner.

Funding: This work was in part supported by Genzyme Sanofi