A Recommender System Based on Binary Matrix Representations for Cognitive Disorders
Raoul H. Kutil, Georg Zimmermann, Christian Borgelt
TL;DR
The paper tackles differential diagnosis in cognitive disorders with overlapping symptoms by proposing a machine-actionable binary matrix framework and abductive reasoning to suggest plausible disorders and the most informative follow-up symptoms. It introduces dual binary representations (MP for initial filtering and AP for exhaustive profiling) and profile generators that encode DSM-5 criteria into scalable structures, along with conditional generators that adapt to observed symptoms. A Python prototype demonstrates feasible identification of candidate disorders and informative symptoms using synthetic and limited real data, and an extension enables on-demand profile generation to handle sparse inputs. The work lays a foundation for a scalable clinical decision-support tool with potential applications beyond mental health, and outlines future steps toward real-world deployment and validation.
Abstract
Diagnosing cognitive (mental health) disorders is a delicate and complex task. Identifying the next most informative symptoms to assess, in order to distinguish between possible disorders, presents an additional challenge. This process requires comprehensive knowledge of diagnostic criteria and symptom overlap across disorders, making it difficult to navigate based on symptoms alone. This research aims to develop a recommender system for cognitive disorder diagnosis using binary matrix representations. The core algorithm utilizes a binary matrix of disorders and their symptom combinations. It filters through the rows and columns based on the patient's current symptoms to identify potential disorders and recommend the most informative next symptoms to examine. A prototype of the recommender system was implemented in Python. Using synthetic test and some real-life data, the system successfully identified plausible disorders from an initial symptom set and recommended further symptoms to refine the diagnosis. It also provided additional context on the symptom-disorder relationships. Although this is a prototype, the recommender system shows potential as a clinical support tool. A fully-developed application of this recommender system may assist mental health professionals in identifying relevant disorders more efficiently and guiding symptom-specific follow-up investigations to improve diagnostic accuracy.