JavaScript/TypeScript SDK for Legit.Health Dermatology API

Official SDK for integrating the Dermatology API offered by Legit.Health 🩺🤖

Instructions

If you want to start sending requests to Legit.Health's Dermatology API, you have to create an instance of the class @legit.health/dapi-sdk/MediaAnalyzer. It requires two arguments:

• The URL of the API. The preproduction enviroment uses the following value: https://ia-pre.legit.health:8443/v2/legit_health/predict.
• The API Key we have provided to you. Without it, you won't be able to send requests to the API.

The class MediaAnalyzer exposes two methods:

• predict, to send a predict request to the API, in case you need to analyze an image to obtain the probability of the detected pathologies.

• followUp, to send a follow up request to get the evolution information about a diagnosed image.

Predict requests

The predict method of our MediaAnalyzer class receives one argument of the class MediaAnalyzerArguments containing a PredictData object. The constructor of this class receives several arguments, so you can specify the image itself and information about the patient or the body site:

import MediaAnalyzer from '@legit.health/dapi-sdk/MediaAnalyzer';
import PredictData from '@legit.health/dapi-sdk/PredictData';
import MediaAnalyzerArguments from '@legit.health/dapi-sdk/MediaAnalyzerArguments';
import BodySiteCode from '@legit.health/dapi-sdk/BodySiteCode';
import Operator from '@legit.health/dapi-sdk/Operator';
import Subject from '@legit.health/dapi-sdk/Subject';
import Gender from '@legit.health/dapi-sdk/Gender';
import Company from '@legit.health/dapi-sdk/Company';

const predictData = new PredictData(
    image,
    BodySiteCode.ArmLeft,
    Operator.Patient,
    new Subject(
        'subject identifier',
        Gender.Male,
        '1.75',
        '69.00',
        new Date(),
        'practitioner identifier'
        new Company('company identifier', 'Company Name')
    )
);
const mediaAnalyzerArguments = new MediaAnalyzerArguments(
    generateRandomString(15),
    predictData
);

Once you've created a PredictData object, you can send the request in this way:

const mediaAnalyzer = new MediaAnalyzer(apiUrl, apiKey);
const response = await mediaAnalyzer.predict(mediaAnalyzerArguments);

The response object contains several properties with the information returned by the API about the analyzed image:

• preliminaryFindings is an object of the class @legit.health/dapi-sdk/PreliminaryFindingsValue with the probability of the different suspicions that the algorithm has about the image.

• modality is the modality of the image detected.

• mediaValidity is an object that contains information about whether the image sent contains relevant dermatological information

• metricsValue contains the sensitivity and specificity values.

• conclusions is an array of @legit.health/dapi-sdk/Conclusion objects with the detected pathologies and its probability. The total probability is distributed among each of the pathologies detected.

• iaSeconds is the time spent by the algorithms analyzying the image.

Follow up requests

The followUp method of the MediaAnalyzer class receives one object of the class MediaAnalyzerArguments containing a FollowUpData object. The constructor of this class receives several arguments, so you can specify the image itself and information about a well known condition.

Example. Follow up request for psoriasis

Let's see how to send a follow-up request for a patient diagnosed with psoriasis.

Firstly, we will create the different objects that represents the questionnaires used to track the evolution of psoriasis:

import ApasiLocalQuestionnaire from '@legit.health/dapi-sdk/ApasiLocalQuestionnaire';
import PasiLocalQuestionnaire from '@legit.health/dapi-sdk/PasiLocalQuestionnaire';
import Pure4Questionnaire from '@legit.health/dapi-sdk/Pure4Questionnaire';
import DlqiQuestionnaire from '@legit.health/dapi-sdk/DlqiQuestionnaire';
import Questionnaires from '@legit.health/dapi-sdk/Questionnaires';

// ...

const apasiLocal = new ApasiLocalQuestionnaire(3);
const pasiLocal = new PasiLocalQuestionnaire(3, 2, 1, 1);
const pure4 = new Pure4Questionnaire(0, 0, 0, 1);
const dlqi = new DlqiQuestionnaire(1, 1, 2, 0, 0, 0, 1, 2, 2, 0);
const questionnaires = new Questionnaires([apasiLocal, pasiLocal, pure4, dlqi]);

Then, we will create an object of the class @legit.health/dapi-sdk/FollowUpArguments:

// ...
import FollowUpData from '@legit.health/dapi-sdk/FollowUpData';
import MediaAnalyzerArguments from '@legit.health/dapi-sdk/MediaAnalyzerArguments';

// ...
const previousMedias = [new PreviousMedia(previousImage, previousImageDate)];
const scoringSystems = questionnaires.questionnaires.map((questionnaire) => questionnaire.getName());
// const scoringSystems = ['APASI_LOCAL', 'PASI_LOCAL', 'PURE4', 'DLQI'];
const followUpData = new FollowUpData(
    generateRandomString(15),
    image,
    'Psoriasis',
    BodySiteCode.ArmLeft,
    previousMedias,
    Operator.Patient,
    new Subject(
        generateRandomString(15),
        Gender.Male,
        '1.75',
        '69.00',
        new Date(),
        generateRandomString(15),
        new Company(generateRandomString(), 'Company Name')
    ),
    questionnaires.questionnaires.map((questionnaire) => questionnaire.getName()),
    questionnaires
);
const mediaAnalyzerArguments = new MediaAnalyzerArguments(
    generateRandomString(15),
    followUpData
);

Unlike diagnostic support requests, follow-up requests supports the following additional arguments:

• previousMedias is an array of objects of the class @legit.health/dapi-sdk/MediaAnalyzerArguments/PreviousMedia/PreviousMedia with a list of previous images taken of the current pathology.

• scoringSystems is an array of strings with the name of the scoring systems to be evaluated. It supports the following values:

[ ASCORAD_LOCAL, APASI_LOCAL, AUAS_LOCAL, AIHS4_LOCAL, DLQI, SCOVID, ALEGI, PURE4, UCT, AUAS7, APULSI, SCORAD_LOCAL, PASI_LOCAL, UAS_LOCAL, IHS4_LOCAL]

• questionnaires is an object of the class @legit.health/dapi-sdkQuestionnaires with the values of the scoring systems to be evaluated.

Once you've created a MediaAnalyzerArguments object, you can send the request in this way:

const mediaAnalyzer = new MediaAnalyzer(apiUrl, apiKey);
const response = await mediaAnalyzer.followUp(mediaAnalyzerArguments);

The response object contains several properties with the information returned by the API about the analyzed image:

• preliminaryFindings is an object of the class @legit.health/dapi-sdk/PreliminaryFindingsValue with the probability of the different suspicions that the algorithm has about the image.

• modality is the modality of the image detected.

• mediaValidity is an object that contains information about whether the image sent contains relevant dermatological information

• metricsValue contains the sensitivity and specificity values.

• iaSeconds is the time spent by the algorithms analyzying the image.

Besides, it contains two extra properties:

• explainabilityMedia, with the image containing the surface of the injury detected by our AI algorithms.

• scoringSystemsValues, an object of the class @legit.health/dapi-sdk/ScoringSystemValues with the values calculated for each scoring system included in the array scoringSystems of the arguments.

The ScoringSystemValues object

The ScoringSystemValues property of MediaAnalyzerResponse contains all the information about a scoring system, for example, APASI_LOCAL.

You can access to the value of one scoring system using the method getScoringSystemValues:

const apasiLocalScoringSystemValue = response.getScoringSystemValues('APASI_LOCAL');

Once you have one object of the class ScoringSystemValues, you can perform the following actions:

• Access to the value of each facet using the method getFacetCalculatedValue(facetCode: string): ScoringSystemFacetCalculatedValue.

• Access to the score of the scoring system through its property score. It is an object with three properties:

  • category, which represents the severity of the score.
  • calculatedScore, for those scoring systems whose calculation depends on facets that are computed by our AI algorithm: APASI_LOCAL, APULSI, ASCORAD_LOCAL and AUAS_LOCAL.
  • questionnaire, for those scoring systems whose calculations not depends on facet computed by our AI algorithm, for example, DLQI.

Full example:

const apasiLocalScoringSystemValue = response.getScoringSystemValues('APASI_LOCAL');

const apasiScore = apasiLocalScoringSystemValue->getScore()->calculatedScore;
const apasiSeverityCategory = $apasiLocalScoringSystemValue->getScore()->category;

const apasiLocalScoringSystemValue = response.getScoringSystemValues('APASI_LOCAL');
const desquamation = response.getFacetCalculatedValue('desquamation');
const desquamationValue = desquamation.value; // A value between 0 and 4 as the PASI states
const desquamationIntensity = desquamation.intensity; // A value between 0 and 100 reflecting the intensity of the desquamation