Our lab has developed an artificial pancreas system called the McGill Artificial Pancreas (MAP) for automating insulin delivery. Using patient's basal-bolus parameters (basal rates and ICRs), the artificial pancreas involves a control algorithm that modulates insulin infusion based on the sensor readings and meal information. However, because basal-bolus parameters are difficult to optimize, proper glycemic control is not always achieved. Therefore, we have developed a learning algorithm that estimates optimal basal-bolus parameters using data over several days. The algorithm examines daily glucose, insulin, and meal data to make changes in patients' basal rates and ICRs.

The objective of this project is to test our artificial pancreas system with and without the learning algorithm using a randomized crossover design in between 31 and 67 children and adolescents at camp Carowanis. We hypothesize that adding a learning algorithm to the artificial pancreas will improve the performance of our artificial pancreas system by increasing the time spent in target glucose range (4mmol/L - 10mmol/L) compared with the artificial pancreas system alone.

This is an open-label, randomized, two-way, cross-over study to compare the glucose control between closed-loop strategy with and without a learning module. Children and adolescent type 1 diabetes patients at Camp Carowanis will be enrolled in the study, where they will undergo two randomly ordered interventions:

1. Closed-loop therapy: participants will undergo a closed-loop therapy where insulin delivery is determined by the MAP system. The study parameters (basal rates and ICRs) will be determined by the camp's physicians on day 1 of camp. The research staff will update the pump's settings to reflect the physician's recommendations at the beginning of the closed-loop therapy, and each time the physicians update the study parameters. Camp physicians will review participants' sensor and insulin data daily, and if necessary, adjust participant basal rates and ICRs. The research staff members will likewise adjust the pump's basal rates and ICR settings as per physician's recommendations. The closed-loop therapy will last 2 days (48 hours).
2. Closed-loop therapy with learning module: participants will undergo a closed-loop therapy where insulin delivery is determined by the MAP system. The study parameters (basal rates and ICRs) will be computed by the learning algorithm and updated daily. The learning algorithm runs on a computer of the research staff members and requires patient data to calculate the optimal basal rates and ICR. Each morning, the research staff members will upload patient data onto the computer, run the leaning algorithm, and update the pump parameters to reflect the recommendations computed by the learning algorithm. Camp physicians will be to review the algorithm's recommendations before they are entered into the patient's pump. The closed-loop therapy with the learning module will last 8 days (192 hours).

Inclusion Criteria:

1. Males and females between 8 and 21 years old.
2. Clinical diagnosis of type 1 diabetes for at least 12 months. The diagnosis of type 1 diabetes is based on the investigator's judgment; C peptide level and antibody determinations are not needed.
3. The participant will have been on insulin pump therapy for at least 3 months.
4. HbA1c ≤ 11%.

Exclusion Criteria:

1. Participants who cannot or are unwilling to use NovoRapid (Aspart) insulin or Humalog (Lispro) insulin for the duration of the study.
2. Serious medical illness likely to interfere with study participation or with the ability to complete the trial by the judgment of the investigator.
3. Failure to comply with the study protocol or with team's recommendations (e.g. not willing to use trial pump, etc.).