In a first, scientists are developing an AI tool to help doctors distinguish between infectious diseases

Google Health also used an AI algorithm that looks at scans of the back of the eye to help predict patients’ risk of suffering a major cardiac event.

Now, a group of researchers from the Manipal College of Pharmaceutical Sciences in India have developed a new machine learning-based tool that could help doctors distinguish between different tropical diseases, including dengue fever and malaria.

The study is the first of its kind in which statistical and machine learning approaches have been explored simultaneously to differentiate between tropical infections.

Their results were published in OLPa non-profit research publisher based in San Francisco, California, USA.

The detection of tropical diseases has always been a challenge for doctors in emergency situations. Four of them, namely dengue fever, malaria, leptospirosis and scrub typhus, present similar clinical symptoms.

“Even an elaborate diagnosis could take three days to get results. That’s what prompted us to explore diagnosis with an AI tool,” said Girish Thunga, Senior Scale Assistant Professor in the Department of Pharmaceutical Practice at Manipal. College of Pharmaceutical Sciences, India. Interesting engineering (IE) in an interview.

A tool to distinguish dengue fever, malaria, leptospirosis and scrub typhus

Tropical infectious diseases such as dengue fever, malaria, leptospirosis, leishmaniasis, scrub typhus and rickettsia are major causes of acute febrile illnesses, which constitute a vital concern for the health of people in the countries where they are common.

The World Health Organization has classified dengue fever as a “neglected tropical disease” (a disease linked to poverty and lack of funding for research and development on treatments and cures), although nearly half of the world’s population live in areas where they are at risk of contracting the disease.

Leptospirosis, malaria, scrub typhus and endemic fever are among diseases commonly confused with dengue, according to a study of clinical and laboratory profiles of dengue-like illnesses at a tertiary care hospital in West Bengal, India , a region where dengue fever is endemic.

One of the main reasons for the difficulty in diagnosing tropical infections is that they have “similar laboratory values, overlapping symptomatology, early asymptomatic presentation, misunderstanding, and delayed diagnosis,” the article noted. of the PLOS.

The non-specific clinical presentations of these infectious diseases also make it difficult to predict larger outbreaks. In addition, overlapping symptoms and a delay in differential diagnosis could aggravate the situation.

The need for a tool that could help identify early symptoms and distinguish laboratory parameters of these infections is imperative to reduce their prevalence.

“Early and accurate diagnosis would lead to the right antibiotic [use]reducing antibiotic resistance and [lower] mortality rate in clinical settings,” Thunga said.

Our study aimed to create a tool that could distinguish between dengue fever, malaria, leptospirosis and scrub typhus in a tertiary care hospital for early prediction.”

WEKA Software for Machine Learning (ML) Modeling

Over a period of nine months, researchers conducted a needs analysis at a tertiary care center in southern India.

A nine-point self-administered questionnaire was developed, validated, distributed, and analyzed to estimate physicians’ need to differentiate tropical diseases in their context and components.

While the first part of the questionnaire included six disease-specific questions – such as the frequency of different tropical infections, the number of cases treated in a week, the barriers to treatment of tropical infections, the challenges of infection management and the perceived need for tool development.

The second part dealt with the development of the tool and included three questions regarding the parameters that clinicians would like to see included, as well as suggested formats for the tool and additional suggestions.

Data for the development of the prediction tool was then collected retrospectively from the medical records department. A prediction tool was then developed, which used multi-nominal regression analysis and a machine learning algorithm.

“We looked at a simple scoring system that would differentiate these infections through a simple decision tree. A total of 800 patients with 200 in each group for the four diseases were collected and analyzed accordingly,” Thunga explained.

Waikato Environment for Knowledge Analysis (WEKA) software was used for machine learning modeling. It was applied to test binary (one disease at a time) and multiclass (all four diseases) classification.

WEKA is an open source machine learning software in JAVA. It contains tools for data preparation, classification, regression, clustering, association rule exploration, and visualization.

The questionnaire that was circulated to 40 physicians and graduate students in the Department of Medicine was a godsend. Doctors said they were treating an average of 24 cases of tropical infection a week. They felt that the diagnosis was difficult and that the management of symptomatology was difficult.

According to the study, 35 doctors felt the need to develop a decision support tool. Thirty-four of them agreed to include laboratory parameters and [35 of them] clinical presentations as a primary criterion in the tool.

The questionnaire revealed that dengue fever, malaria, leptospirosis and scrub typhus were the most common tropical infections in this setting; with sodium, total bilirubin, albumin, lymphocytes and platelets, the most common laboratory parameters; and abdominal pain, arthralgia, myalgia, and urine output were the clinical presentation were the best predictors of disease.

The tool offered 60.7%, 62.5%, and 66% predictability for dengue, malaria, and leptospirosis, respectively, and 38% predictability for scrub typhus.

“The multiple classification machine learning model observed to have an overall predictability of 55-60%, whereas a binary classification machine learning algorithm showed an average of 79-84% for one versus the other and 69-88% for one versus one disease category,” the research documented.

Challenges and limitations

Thunga said the team faced many challenges. “Several permutations and combinations had to be done initially. Each patient would have varying degrees of the disease spectrum, and we were encountering false positives. Of course, eventually, we worked on it,” he said.

He said further studies were needed to provide detailed insight into the application of this study.

“The credibility of our findings could be further analyzed and improved depending on the respective clinical scenario,” he added.

The document indicates that future studies could focus on general aspects of the disease rather than the parameters considered in the study, based on a better knowledge of the geographical distribution.

The researchers also caution about the limitations of their study. Some of them include retrospective data collection, which means that clinical parameters were not recorded during the first visits to the emergency department or clinic, when they could have been more accurate.

Furthermore, “The results of these single-center data cannot be generalized to another part of the world because the nature and presentation of tropical diseases vary from place to place. And since we used the WEKA software for the machine learning does provide true negatives, false negatives, and specificity,” Thunga noted.

Leave a Reply

%d bloggers like this: