Imprimir Republish

Diagnosis by computer

The languages of psychosis

Mathematical approach shows the differences between the discourse of maniacs and schizophrenics

062-064_Psicose_194-1Psychiatrists and people in general find it relatively easy to distinguish a psychotic from someone who does not manifest signs of a diagnosed mental disorder. Psychotics report delirium and hallucinations and sometimes see themselves as the Messiah out to save the world. However, distinguishing mania from schizophrenia – two types of psychosis – is not that simple and requires extensive personal experience, knowledge, and intuition on the part of specialists. A mathematical approach, developed at the Brain Institute of the Federal University of Rio Grande do Norte (UFRN) might facilitate making this distinction, which is crucial to determine the most suitable treatment for each illness. This mathematical approach quantitatively evaluates differences in the verbal language structures used by patients diagnosed with mania or schizophrenia.

The analysis strategy – based on graph theory and representing words as nodes and the sequence of words in sentences or phrases as arrows – showed that people with mania are much more verbose and repetitive than people with schizophrenia, who are usually laconic and centered on a single topic, and do not let their minds roam. “Recurrence is the outstanding characteristic of a patient with mania, who repeats the same thing three or four times, while a patient with schizophrenia talks objectively, and to the point, though the discourse is poor in meanings,” says psychiatrist Natália Mota, a researcher at the Institute. Sidarta Ribeiro, director of the institute, says that “the number of words, the structure of the language, and other indicators are entirely different.”

The researchers believe that they have taken the first steps towards an objective way of distinguishing one form of psychosis from the other, akin to the way in which a blood test is used to check for an infectious disease, provided that the next tests – which will include a broader sample of participants – reinforce the consistency of this approach and physicians agree to work with an assistant of this kind. The comparative tests described in an article published recently in PLoS One indicated that this new approach has achieved a 93% diagnosis success rate, while currently used psychometric scales – based on symptom evaluation questionnaires – achieve a 67% success rate. “These are complementary methods,” says Natália. “The psychometric scales and the physicians’ experience are still necessary.”

“The result is quite simple, even for someone who doesn’t understand mathematics,” says physicist Mauro Copelli, from the Federal University of Pernambuco (UFPE), who participated in the project. The discourse of people with mania is portrayed as a tangle of nodes and lines, while the discourse of a schizophrenic is a straight line, with few nodes. Graph theory, the basis of these diagrams, has been used for centuries to examine the pathway through which a traveler could visit all the cities of a given region, for example. More recently, graph theory has been used to optimize air traffic, where airports are taken as a group of end points interconnected by airplanes.

“The first time I ran the graph programs, the language-related differences were immediately obvious,” says Natália. In 2007, after graduating from medical school and entering the medical residency program in psychiatry at the teaching hospital of UFRN, Natália noticed that many diagnoses distinguishing mania from schizophrenia depended on the physicians’ personal experience and subjective judgment – physicians who saw more schizophrenics tended to diagnose more cases of schizophrenia and fewer cases of mania and often there was no consensus. It was already widely known that maniacs talk more and deviate from the main topic more easily than schizophrenics, but this belief seemed to be too generic to Natália. Natália talked to Copelli at a science congress in the city of Fortaleza, State of Ceará, in 2008. At that time, Copelli had already been working with Ribeiro and he encouraged Natália to work with graphs. At first, she resisted the idea because she was unfamiliar with mathematics; soon, however, she found the new theory to be simple and practical.

To move forward with her research project, Natália recorded and transcribed interviews with 24 people. Nathália Lemos and Ana Cardina Pieretti helped her with this task. Of the 24 respondents, eight had been diagnosed as having mania, eight had schizophrenia, and eight did not have any diagnosed mental disorder. Natália asked the respondents to tell her about a dream; any comment that was not in line with this topic was considered as a flight of the imagination, a common characteristic of people with mania.

062-064_Psicose_194-2“The transcriptions of the mania patients’ reports showed that their stories were clearly longer than those of the patients with schizophrenia,” she says. She then eliminated the less important elements, such as articles and prepositions, and split the sentence into subject, verb, and objects, represented by the nodes. The sequence among the points in the sentence were represented by arrows joining two nodes. She also highlighted the sentences that were not related to the main topic of the report, i.e., the recent dream that she had asked the respondents to tell her about and which showed the digression common to people with mania.

A specific program for graphs, downloaded for free from the internet, specified the relevant characteristics – or attributes – for the analysis and represented the main differences in the patients’ reports, such as quantity of nodes, recurrence, verbosity, (or logorrhea) and digression from the main topic. “It’s very simple,” says Natália. When working on the validations and analyses of the results, she had the help of Osame Kinouchi, from the University of São Paulo (USP) at Ribeirão Preto, and of Guillermo Cecchi, from IBM’s Computational Biology Center in the United States.

The results showed that in most of the interview items, the respondents with mania scored higher than those with schizophrenia. “The logorrhea, which is typical of patients with mania, is the result not only of a very large number of words but also of a discourse that always goes back to the same topic, in comparison to the group with schizophrenia,” she observed. Strangely enough, the participants from the control group – the ones that had no diagnosed mental disorder – showed both discourse structures – sometimes redundant ones, similar to those of the patients with mania, and sometimes more streamlined ones, as in patients with schizophrenia, thus reflecting the differences in their personalities or the motivation to speak more or less at a given moment. “It is widely known that the pathology defines the discourse,” she says. Psychiatrists are trained to recognize these differences, but it is difficult for them to state that recurrence in a patient with mania is 28% lower, no matter how experienced they are.”

“The institute’s interdisciplinary environment was crucial to conduct this study, because every day I had the opportunity to exchange ideas with people from other professional fields. Nivaldo Vasconcelos, a computer engineer, helped me a lot,” she says. The Brain Institute, established in 2007, currently has 13 professors, 22 undergraduate students, 42 students enrolled in the post-graduate program, 8 students enrolled in post-doctoral programs, and 30 technicians. “Once we had overcome the initial difficulties, we managed to set up a group of talented young researchers,” says Ribeiro. “The facility we are using now has a big garden and very often we stay there until two or three o´clock in the morning talking about science and drinking mate tea.”

Scientific article
MOTA, N.B. et al. Speech graphs provide a quantitative measure of thought disorder in psychosis. PLoS ONE (forthcoming).